41 patches for repository http://tahoe-lafs.org/source/tahoe/trunk: Thu Aug 25 01:32:17 BST 2011 david-sarah@jacaranda.org * interfaces.py: 'which -> that' grammar cleanup. Tue Sep 20 00:29:26 BST 2011 david-sarah@jacaranda.org * Pluggable backends -- new and moved files, changes to moved files. refs #999 Tue Sep 20 00:32:56 BST 2011 david-sarah@jacaranda.org * Pluggable backends -- all other changes. refs #999 Tue Sep 20 04:38:03 BST 2011 david-sarah@jacaranda.org * Work-in-progress, includes fix to bug involving BucketWriter. refs #999 Tue Sep 20 18:17:37 BST 2011 david-sarah@jacaranda.org * docs/backends: document the configuration options for the pluggable backends scheme. refs #999 Wed Sep 21 04:12:07 BST 2011 david-sarah@jacaranda.org * Fix some incorrect attribute accesses. refs #999 Wed Sep 21 04:16:25 BST 2011 david-sarah@jacaranda.org * docs/backends/S3.rst: remove Issues section. refs #999 Wed Sep 21 04:17:05 BST 2011 david-sarah@jacaranda.org * docs/backends/S3.rst, disk.rst: describe type of space settings as 'quantity of space', not 'str'. refs #999 Wed Sep 21 19:46:49 BST 2011 david-sarah@jacaranda.org * More fixes to tests needed for pluggable backends. refs #999 Wed Sep 21 23:14:21 BST 2011 david-sarah@jacaranda.org * Fix more shallow bugs, mainly FilePathification. Also, remove the max_space_per_bucket parameter from BucketWriter since it can be obtained from the _max_size attribute of the share (via a new get_allocated_size() accessor). refs #999 Wed Sep 21 23:20:38 BST 2011 david-sarah@jacaranda.org * uri.py: resolve a conflict between trunk and the pluggable-backends patches. refs #999 Thu Sep 22 05:54:51 BST 2011 david-sarah@jacaranda.org * Fix some more test failures. refs #999 Thu Sep 22 19:30:08 BST 2011 david-sarah@jacaranda.org * Fix most of the crawler tests. refs #999 Thu Sep 22 19:33:23 BST 2011 david-sarah@jacaranda.org * Reinstate the cancel_lease methods of ImmutableDiskShare and MutableDiskShare, since they are needed for lease expiry. refs #999 Fri Sep 23 02:20:44 BST 2011 david-sarah@jacaranda.org * Blank line cleanups. Fri Sep 23 05:08:25 BST 2011 david-sarah@jacaranda.org * mutable/publish.py: elements should not be removed from a dictionary while it is being iterated over. refs #393 Fri Sep 23 05:10:03 BST 2011 david-sarah@jacaranda.org * A few comment cleanups. refs #999 Fri Sep 23 05:11:15 BST 2011 david-sarah@jacaranda.org * Move advise_corrupt_share to allmydata/storage/backends/base.py, since it will be common to the disk and S3 backends. refs #999 Fri Sep 23 05:13:14 BST 2011 david-sarah@jacaranda.org * Add incomplete S3 backend. refs #999 Fri Sep 23 21:37:23 BST 2011 david-sarah@jacaranda.org * interfaces.py: add fill_in_space_stats method to IStorageBackend. refs #999 Fri Sep 23 21:44:25 BST 2011 david-sarah@jacaranda.org * Remove redundant si_s argument from check_write_enabler. refs #999 Fri Sep 23 21:46:11 BST 2011 david-sarah@jacaranda.org * Implement readv for immutable shares. refs #999 Fri Sep 23 21:49:14 BST 2011 david-sarah@jacaranda.org * The cancel secret needs to be unique, even if it isn't explicitly provided. refs #999 Fri Sep 23 21:49:45 BST 2011 david-sarah@jacaranda.org * Make EmptyShare.check_testv a simple function. refs #999 Fri Sep 23 21:52:19 BST 2011 david-sarah@jacaranda.org * Update the null backend to take into account interface changes. Also, it now records which shares are present, but not their contents. refs #999 Fri Sep 23 21:53:45 BST 2011 david-sarah@jacaranda.org * Update the S3 backend. refs #999 Fri Sep 23 21:55:10 BST 2011 david-sarah@jacaranda.org * Minor cleanup to disk backend. refs #999 Fri Sep 23 23:09:35 BST 2011 david-sarah@jacaranda.org * Add 'has-immutable-readv' to server version information. refs #999 Tue Sep 27 08:09:47 BST 2011 david-sarah@jacaranda.org * util/deferredutil.py: add some utilities for asynchronous iteration. refs #999 Tue Sep 27 08:14:03 BST 2011 david-sarah@jacaranda.org * test_storage.py: fix test_status_bad_disk_stats. refs #999 Tue Sep 27 08:15:44 BST 2011 david-sarah@jacaranda.org * Cleanups to disk backend. refs #999 Tue Sep 27 08:18:55 BST 2011 david-sarah@jacaranda.org * Cleanups to S3 backend (not including Deferred changes). refs #999 Tue Sep 27 08:28:48 BST 2011 david-sarah@jacaranda.org * test_storage.py: fix test_no_st_blocks. refs #999 Tue Sep 27 08:35:30 BST 2011 david-sarah@jacaranda.org * mutable/publish.py: resolve conflicting patches. refs #999 Tue Sep 27 08:39:03 BST 2011 david-sarah@jacaranda.org * Work in progress for asyncifying the backend interface (necessary to call txaws methods that return Deferreds). This is incomplete so lots of tests fail. refs #999 Wed Sep 28 02:37:29 BST 2011 david-sarah@jacaranda.org * Undo an incompatible change to RIStorageServer. refs #999 Wed Sep 28 02:38:57 BST 2011 david-sarah@jacaranda.org * test_system.py: incorrect arguments were being passed to the constructor for MutableDiskShare. refs #999 Wed Sep 28 02:40:19 BST 2011 david-sarah@jacaranda.org * test_system.py: more debug output for a failing check in test_filesystem. refs #999 Wed Sep 28 02:40:49 BST 2011 david-sarah@jacaranda.org * scripts/debug.py: fix incorrect arguments to dump_immutable_share. refs #999 Wed Sep 28 02:41:26 BST 2011 david-sarah@jacaranda.org * mutable/publish.py: don't crash if there are no writers in _report_verinfo. refs #999 Wed Sep 28 06:23:24 BST 2011 david-sarah@jacaranda.org * Use factory functions to create share objects rather than their constructors, to allow the factory to return a Deferred. Also change some methods on IShareSet and IStoredShare to return Deferreds. Refactor some constants associated with mutable shares. refs #999 New patches: [interfaces.py: 'which -> that' grammar cleanup. david-sarah@jacaranda.org**20110825003217 Ignore-this: a3e15f3676de1b346ad78aabdfb8cac6 ] { hunk ./src/allmydata/interfaces.py 38 the StubClient. This object doesn't actually offer any services, but the announcement helps the Introducer keep track of which clients are subscribed (so the grid admin can keep track of things like the size of - the grid and the client versions in use. This is the (empty) + the grid and the client versions in use). This is the (empty) RemoteInterface for the StubClient.""" class RIBucketWriter(RemoteInterface): hunk ./src/allmydata/interfaces.py 276 (binary) storage index string, and 'shnum' is the integer share number. 'reason' is a human-readable explanation of the problem, probably including some expected hash values and the computed ones - which did not match. Corruption advisories for mutable shares should + that did not match. Corruption advisories for mutable shares should include a hash of the public key (the same value that appears in the mutable-file verify-cap), since the current share format does not store that on disk. hunk ./src/allmydata/interfaces.py 413 remote_host: the IAddress, if connected, otherwise None This method is intended for monitoring interfaces, such as a web page - which describes connecting and connected peers. + that describes connecting and connected peers. """ def get_all_peerids(): hunk ./src/allmydata/interfaces.py 515 # TODO: rename to get_read_cap() def get_readonly(): - """Return another IURI instance, which represents a read-only form of + """Return another IURI instance that represents a read-only form of this one. If is_readonly() is True, this returns self.""" def get_verify_cap(): hunk ./src/allmydata/interfaces.py 542 passing into init_from_string.""" class IDirnodeURI(Interface): - """I am a URI which represents a dirnode.""" + """I am a URI that represents a dirnode.""" class IFileURI(Interface): hunk ./src/allmydata/interfaces.py 545 - """I am a URI which represents a filenode.""" + """I am a URI that represents a filenode.""" def get_size(): """Return the length (in bytes) of the file that I represent.""" hunk ./src/allmydata/interfaces.py 553 pass class IMutableFileURI(Interface): - """I am a URI which represents a mutable filenode.""" + """I am a URI that represents a mutable filenode.""" def get_extension_params(): """Return the extension parameters in the URI""" hunk ./src/allmydata/interfaces.py 856 """ class IFileNode(IFilesystemNode): - """I am a node which represents a file: a sequence of bytes. I am not a + """I am a node that represents a file: a sequence of bytes. I am not a container, like IDirectoryNode.""" def get_best_readable_version(): """Return a Deferred that fires with an IReadable for the 'best' hunk ./src/allmydata/interfaces.py 905 multiple versions of a file present in the grid, some of which might be unrecoverable (i.e. have fewer than 'k' shares). These versions are loosely ordered: each has a sequence number and a hash, and any version - with seqnum=N was uploaded by a node which has seen at least one version + with seqnum=N was uploaded by a node that has seen at least one version with seqnum=N-1. The 'servermap' (an instance of IMutableFileServerMap) is used to hunk ./src/allmydata/interfaces.py 1014 as a guide to where the shares are located. I return a Deferred that fires with the requested contents, or - errbacks with UnrecoverableFileError. Note that a servermap which was + errbacks with UnrecoverableFileError. Note that a servermap that was updated with MODE_ANYTHING or MODE_READ may not know about shares for all versions (those modes stop querying servers as soon as they can fulfil their goals), so you may want to use MODE_CHECK (which checks hunk ./src/allmydata/interfaces.py 1073 """Upload was unable to satisfy 'servers_of_happiness'""" class UnableToFetchCriticalDownloadDataError(Exception): - """I was unable to fetch some piece of critical data which is supposed to + """I was unable to fetch some piece of critical data that is supposed to be identically present in all shares.""" class NoServersError(Exception): hunk ./src/allmydata/interfaces.py 1085 exists, and overwrite= was set to False.""" class NoSuchChildError(Exception): - """A directory node was asked to fetch a child which does not exist.""" + """A directory node was asked to fetch a child that does not exist.""" class ChildOfWrongTypeError(Exception): """An operation was attempted on a child of the wrong type (file or directory).""" hunk ./src/allmydata/interfaces.py 1403 if you initially thought you were going to use 10 peers, started encoding, and then two of the peers dropped out: you could use desired_share_ids= to skip the work (both memory and CPU) of - producing shares for the peers which are no longer available. + producing shares for the peers that are no longer available. """ hunk ./src/allmydata/interfaces.py 1478 if you initially thought you were going to use 10 peers, started encoding, and then two of the peers dropped out: you could use desired_share_ids= to skip the work (both memory and CPU) of - producing shares for the peers which are no longer available. + producing shares for the peers that are no longer available. For each call, encode() will return a Deferred that fires with two lists, one containing shares and the other containing the shareids. hunk ./src/allmydata/interfaces.py 1535 required to be of the same length. The i'th element of their_shareids is required to be the shareid of the i'th buffer in some_shares. - This returns a Deferred which fires with a sequence of buffers. This + This returns a Deferred that fires with a sequence of buffers. This sequence will contain all of the segments of the original data, in order. The sum of the lengths of all of the buffers will be the 'data_size' value passed into the original ICodecEncode.set_params() hunk ./src/allmydata/interfaces.py 1582 Encoding parameters can be set in three ways. 1: The Encoder class provides defaults (3/7/10). 2: the Encoder can be constructed with an 'options' dictionary, in which the - needed_and_happy_and_total_shares' key can be a (k,d,n) tuple. 3: + 'needed_and_happy_and_total_shares' key can be a (k,d,n) tuple. 3: set_params((k,d,n)) can be called. If you intend to use set_params(), you must call it before hunk ./src/allmydata/interfaces.py 1780 produced, so that the segment hashes can be generated with only a single pass. - This returns a Deferred which fires with a sequence of hashes, using: + This returns a Deferred that fires with a sequence of hashes, using: tuple(segment_hashes[first:last]) hunk ./src/allmydata/interfaces.py 1796 def get_plaintext_hash(): """OBSOLETE; Get the hash of the whole plaintext. - This returns a Deferred which fires with a tagged SHA-256 hash of the + This returns a Deferred that fires with a tagged SHA-256 hash of the whole plaintext, obtained from hashutil.plaintext_hash(data). """ hunk ./src/allmydata/interfaces.py 1856 be used to encrypt the data. The key will also be hashed to derive the StorageIndex. - Uploadables which want to achieve convergence should hash their file + Uploadables that want to achieve convergence should hash their file contents and the serialized_encoding_parameters to form the key (which of course requires a full pass over the data). Uploadables can use the upload.ConvergentUploadMixin class to achieve this hunk ./src/allmydata/interfaces.py 1862 automatically. - Uploadables which do not care about convergence (or do not wish to + Uploadables that do not care about convergence (or do not wish to make multiple passes over the data) can simply return a strongly-random 16 byte string. hunk ./src/allmydata/interfaces.py 1872 def read(length): """Return a Deferred that fires with a list of strings (perhaps with - only a single element) which, when concatenated together, contain the + only a single element) that, when concatenated together, contain the next 'length' bytes of data. If EOF is near, this may provide fewer than 'length' bytes. The total number of bytes provided by read() before it signals EOF must equal the size provided by get_size(). hunk ./src/allmydata/interfaces.py 1919 def read(length): """ - Returns a list of strings which, when concatenated, are the next + Returns a list of strings that, when concatenated, are the next length bytes of the file, or fewer if there are fewer bytes between the current location and the end of the file. """ hunk ./src/allmydata/interfaces.py 1932 class IUploadResults(Interface): """I am returned by upload() methods. I contain a number of public - attributes which can be read to determine the results of the upload. Some + attributes that can be read to determine the results of the upload. Some of these are functional, some are timing information. All of these may be None. hunk ./src/allmydata/interfaces.py 1965 class IDownloadResults(Interface): """I am created internally by download() methods. I contain a number of - public attributes which contain details about the download process.:: + public attributes that contain details about the download process.:: .file_size : the size of the file, in bytes .servers_used : set of server peerids that were used during download hunk ./src/allmydata/interfaces.py 1991 class IUploader(Interface): def upload(uploadable): """Upload the file. 'uploadable' must impement IUploadable. This - returns a Deferred which fires with an IUploadResults instance, from + returns a Deferred that fires with an IUploadResults instance, from which the URI of the file can be obtained as results.uri .""" def upload_ssk(write_capability, new_version, uploadable): hunk ./src/allmydata/interfaces.py 2041 kind of lease that is obtained (which account number to claim, etc). TODO: any problems seen during checking will be reported to the - health-manager.furl, a centralized object which is responsible for + health-manager.furl, a centralized object that is responsible for figuring out why files are unhealthy so corrective action can be taken. """ hunk ./src/allmydata/interfaces.py 2056 will be put in the check-and-repair results. The Deferred will not fire until the repair is complete. - This returns a Deferred which fires with an instance of + This returns a Deferred that fires with an instance of ICheckAndRepairResults.""" class IDeepCheckable(Interface): hunk ./src/allmydata/interfaces.py 2141 that was found to be corrupt. Each share locator is a list of (serverid, storage_index, sharenum). - count-incompatible-shares: the number of shares which are of a share + count-incompatible-shares: the number of shares that are of a share format unknown to this checker list-incompatible-shares: a list of 'share locators', one for each share that was found to be of an unknown hunk ./src/allmydata/interfaces.py 2148 format. Each share locator is a list of (serverid, storage_index, sharenum). servers-responding: list of (binary) storage server identifiers, - one for each server which responded to the share + one for each server that responded to the share query (even if they said they didn't have shares, and even if they said they did have shares but then didn't send them when asked, or hunk ./src/allmydata/interfaces.py 2345 will use the data in the checker results to guide the repair process, such as which servers provided bad data and should therefore be avoided. The ICheckResults object is inside the - ICheckAndRepairResults object, which is returned by the + ICheckAndRepairResults object that is returned by the ICheckable.check() method:: d = filenode.check(repair=False) hunk ./src/allmydata/interfaces.py 2436 methods to create new objects. I return synchronously.""" def create_mutable_file(contents=None, keysize=None): - """I create a new mutable file, and return a Deferred which will fire + """I create a new mutable file, and return a Deferred that will fire with the IMutableFileNode instance when it is ready. If contents= is provided (a bytestring), it will be used as the initial contents of the new file, otherwise the file will contain zero bytes. keysize= is hunk ./src/allmydata/interfaces.py 2444 usual.""" def create_new_mutable_directory(initial_children={}): - """I create a new mutable directory, and return a Deferred which will + """I create a new mutable directory, and return a Deferred that will fire with the IDirectoryNode instance when it is ready. If initial_children= is provided (a dict mapping unicode child name to (childnode, metadata_dict) tuples), the directory will be populated hunk ./src/allmydata/interfaces.py 2452 class IClientStatus(Interface): def list_all_uploads(): - """Return a list of uploader objects, one for each upload which + """Return a list of uploader objects, one for each upload that currently has an object available (tracked with weakrefs). This is intended for debugging purposes.""" def list_active_uploads(): hunk ./src/allmydata/interfaces.py 2462 started uploads.""" def list_all_downloads(): - """Return a list of downloader objects, one for each download which + """Return a list of downloader objects, one for each download that currently has an object available (tracked with weakrefs). This is intended for debugging purposes.""" def list_active_downloads(): hunk ./src/allmydata/interfaces.py 2689 def provide(provider=RIStatsProvider, nickname=str): """ - @param provider: a stats collector instance which should be polled + @param provider: a stats collector instance that should be polled periodically by the gatherer to collect stats. @param nickname: a name useful to identify the provided client """ hunk ./src/allmydata/interfaces.py 2722 class IValidatedThingProxy(Interface): def start(): - """ Acquire a thing and validate it. Return a deferred which is + """ Acquire a thing and validate it. Return a deferred that is eventually fired with self if the thing is valid or errbacked if it can't be acquired or validated.""" } [Pluggable backends -- new and moved files, changes to moved files. refs #999 david-sarah@jacaranda.org**20110919232926 Ignore-this: ec5d2d1362a092d919e84327d3092424 ] { adddir ./src/allmydata/storage/backends adddir ./src/allmydata/storage/backends/disk move ./src/allmydata/storage/immutable.py ./src/allmydata/storage/backends/disk/immutable.py move ./src/allmydata/storage/mutable.py ./src/allmydata/storage/backends/disk/mutable.py adddir ./src/allmydata/storage/backends/null addfile ./src/allmydata/storage/backends/__init__.py addfile ./src/allmydata/storage/backends/base.py hunk ./src/allmydata/storage/backends/base.py 1 + +from twisted.application import service + +from allmydata.storage.common import si_b2a +from allmydata.storage.lease import LeaseInfo +from allmydata.storage.bucket import BucketReader + + +class Backend(service.MultiService): + def __init__(self): + service.MultiService.__init__(self) + + +class ShareSet(object): + """ + This class implements shareset logic that could work for all backends, but + might be useful to override for efficiency. + """ + + def __init__(self, storageindex): + self.storageindex = storageindex + + def get_storage_index(self): + return self.storageindex + + def get_storage_index_string(self): + return si_b2a(self.storageindex) + + def renew_lease(self, renew_secret, new_expiration_time): + found_shares = False + for share in self.get_shares(): + found_shares = True + share.renew_lease(renew_secret, new_expiration_time) + + if not found_shares: + raise IndexError("no such lease to renew") + + def get_leases(self): + # Since all shares get the same lease data, we just grab the leases + # from the first share. + try: + sf = self.get_shares().next() + return sf.get_leases() + except StopIteration: + return iter([]) + + def add_or_renew_lease(self, lease_info): + # This implementation assumes that lease data is duplicated in + # all shares of a shareset, which might not be true for all backends. + for share in self.get_shares(): + share.add_or_renew_lease(lease_info) + + def make_bucket_reader(self, storageserver, share): + return BucketReader(storageserver, share) + + def testv_and_readv_and_writev(self, storageserver, secrets, + test_and_write_vectors, read_vector, + expiration_time): + # The implementation here depends on the following helper methods, + # which must be provided by subclasses: + # + # def _clean_up_after_unlink(self): + # """clean up resources associated with the shareset after some + # shares might have been deleted""" + # + # def _create_mutable_share(self, storageserver, shnum, write_enabler): + # """create a mutable share with the given shnum and write_enabler""" + + # secrets might be a triple with cancel_secret in secrets[2], but if + # so we ignore the cancel_secret. + write_enabler = secrets[0] + renew_secret = secrets[1] + + si_s = self.get_storage_index_string() + shares = {} + for share in self.get_shares(): + # XXX is it correct to ignore immutable shares? Maybe get_shares should + # have a parameter saying what type it's expecting. + if share.sharetype == "mutable": + share.check_write_enabler(write_enabler, si_s) + shares[share.get_shnum()] = share + + # write_enabler is good for all existing shares + + # now evaluate test vectors + testv_is_good = True + for sharenum in test_and_write_vectors: + (testv, datav, new_length) = test_and_write_vectors[sharenum] + if sharenum in shares: + if not shares[sharenum].check_testv(testv): + self.log("testv failed: [%d]: %r" % (sharenum, testv)) + testv_is_good = False + break + else: + # compare the vectors against an empty share, in which all + # reads return empty strings + if not EmptyShare().check_testv(testv): + self.log("testv failed (empty): [%d] %r" % (sharenum, + testv)) + testv_is_good = False + break + + # gather the read vectors, before we do any writes + read_data = {} + for shnum, share in shares.items(): + read_data[shnum] = share.readv(read_vector) + + ownerid = 1 # TODO + lease_info = LeaseInfo(ownerid, renew_secret, + expiration_time, storageserver.get_serverid()) + + if testv_is_good: + # now apply the write vectors + for shnum in test_and_write_vectors: + (testv, datav, new_length) = test_and_write_vectors[shnum] + if new_length == 0: + if shnum in shares: + shares[shnum].unlink() + else: + if shnum not in shares: + # allocate a new share + share = self._create_mutable_share(storageserver, shnum, write_enabler) + shares[shnum] = share + shares[shnum].writev(datav, new_length) + # and update the lease + shares[shnum].add_or_renew_lease(lease_info) + + if new_length == 0: + self._clean_up_after_unlink() + + return (testv_is_good, read_data) + + def readv(self, wanted_shnums, read_vector): + """ + Read a vector from the numbered shares in this shareset. An empty + shares list means to return data from all known shares. + + @param wanted_shnums=ListOf(int) + @param read_vector=ReadVector + @return DictOf(int, ReadData): shnum -> results, with one key per share + """ + datavs = {} + for share in self.get_shares(): + shnum = share.get_shnum() + if not wanted_shnums or shnum in wanted_shnums: + datavs[shnum] = share.readv(read_vector) + + return datavs + + +def testv_compare(a, op, b): + assert op in ("lt", "le", "eq", "ne", "ge", "gt") + if op == "lt": + return a < b + if op == "le": + return a <= b + if op == "eq": + return a == b + if op == "ne": + return a != b + if op == "ge": + return a >= b + if op == "gt": + return a > b + # never reached + + +class EmptyShare: + def check_testv(self, testv): + test_good = True + for (offset, length, operator, specimen) in testv: + data = "" + if not testv_compare(data, operator, specimen): + test_good = False + break + return test_good + addfile ./src/allmydata/storage/backends/disk/__init__.py addfile ./src/allmydata/storage/backends/disk/disk_backend.py hunk ./src/allmydata/storage/backends/disk/disk_backend.py 1 + +import re + +from twisted.python.filepath import UnlistableError + +from zope.interface import implements +from allmydata.interfaces import IStorageBackend, IShareSet +from allmydata.util import fileutil, log, time_format +from allmydata.storage.common import si_b2a, si_a2b +from allmydata.storage.bucket import BucketWriter +from allmydata.storage.backends.base import Backend, ShareSet +from allmydata.storage.backends.disk.immutable import ImmutableDiskShare +from allmydata.storage.backends.disk.mutable import MutableDiskShare, create_mutable_disk_share + +# storage/ +# storage/shares/incoming +# incoming/ holds temp dirs named $START/$STORAGEINDEX/$SHARENUM which will +# be moved to storage/shares/$START/$STORAGEINDEX/$SHARENUM upon success +# storage/shares/$START/$STORAGEINDEX +# storage/shares/$START/$STORAGEINDEX/$SHARENUM + +# Where "$START" denotes the first 10 bits worth of $STORAGEINDEX (that's 2 +# base-32 chars). +# $SHARENUM matches this regex: +NUM_RE=re.compile("^[0-9]+$") + + +def si_si2dir(startfp, storageindex): + sia = si_b2a(storageindex) + newfp = startfp.child(sia[:2]) + return newfp.child(sia) + + +def get_share(fp): + f = fp.open('rb') + try: + prefix = f.read(32) + finally: + f.close() + + if prefix == MutableDiskShare.MAGIC: + return MutableDiskShare(fp) + else: + # assume it's immutable + return ImmutableDiskShare(fp) + + +class DiskBackend(Backend): + implements(IStorageBackend) + + def __init__(self, storedir, readonly=False, reserved_space=0, discard_storage=False): + Backend.__init__(self) + self._setup_storage(storedir, readonly, reserved_space, discard_storage) + self._setup_corruption_advisory() + + def _setup_storage(self, storedir, readonly, reserved_space, discard_storage): + self._storedir = storedir + self._readonly = readonly + self._reserved_space = int(reserved_space) + self._discard_storage = discard_storage + self._sharedir = self._storedir.child("shares") + fileutil.fp_make_dirs(self._sharedir) + self._incomingdir = self._sharedir.child('incoming') + self._clean_incomplete() + if self._reserved_space and (self.get_available_space() is None): + log.msg("warning: [storage]reserved_space= is set, but this platform does not support an API to get disk statistics (statvfs(2) or GetDiskFreeSpaceEx), so this reservation cannot be honored", + umid="0wZ27w", level=log.UNUSUAL) + + def _clean_incomplete(self): + fileutil.fp_remove(self._incomingdir) + fileutil.fp_make_dirs(self._incomingdir) + + def _setup_corruption_advisory(self): + # we don't actually create the corruption-advisory dir until necessary + self._corruption_advisory_dir = self._storedir.child("corruption-advisories") + + def _make_shareset(self, sharehomedir): + return self.get_shareset(si_a2b(sharehomedir.basename())) + + def get_sharesets_for_prefix(self, prefix): + prefixfp = self._sharedir.child(prefix) + try: + sharesets = map(self._make_shareset, prefixfp.children()) + def _by_base32si(b): + return b.get_storage_index_string() + sharesets.sort(key=_by_base32si) + except EnvironmentError: + sharesets = [] + return sharesets + + def get_shareset(self, storageindex): + sharehomedir = si_si2dir(self._sharedir, storageindex) + incominghomedir = si_si2dir(self._incomingdir, storageindex) + return DiskShareSet(storageindex, sharehomedir, incominghomedir, discard_storage=self._discard_storage) + + def fill_in_space_stats(self, stats): + stats['storage_server.reserved_space'] = self._reserved_space + try: + disk = fileutil.get_disk_stats(self._sharedir, self._reserved_space) + writeable = disk['avail'] > 0 + + # spacetime predictors should use disk_avail / (d(disk_used)/dt) + stats['storage_server.disk_total'] = disk['total'] + stats['storage_server.disk_used'] = disk['used'] + stats['storage_server.disk_free_for_root'] = disk['free_for_root'] + stats['storage_server.disk_free_for_nonroot'] = disk['free_for_nonroot'] + stats['storage_server.disk_avail'] = disk['avail'] + except AttributeError: + writeable = True + except EnvironmentError: + log.msg("OS call to get disk statistics failed", level=log.UNUSUAL) + writeable = False + + if self._readonly: + stats['storage_server.disk_avail'] = 0 + writeable = False + + stats['storage_server.accepting_immutable_shares'] = int(writeable) + + def get_available_space(self): + if self._readonly: + return 0 + return fileutil.get_available_space(self._sharedir, self._reserved_space) + + def advise_corrupt_share(self, sharetype, storageindex, shnum, reason): + fileutil.fp_make_dirs(self._corruption_advisory_dir) + now = time_format.iso_utc(sep="T") + si_s = si_b2a(storageindex) + + # Windows can't handle colons in the filename. + name = ("%s--%s-%d" % (now, si_s, shnum)).replace(":", "") + f = self._corruption_advisory_dir.child(name).open("w") + try: + f.write("report: Share Corruption\n") + f.write("type: %s\n" % sharetype) + f.write("storage_index: %s\n" % si_s) + f.write("share_number: %d\n" % shnum) + f.write("\n") + f.write(reason) + f.write("\n") + finally: + f.close() + + log.msg(format=("client claims corruption in (%(share_type)s) " + + "%(si)s-%(shnum)d: %(reason)s"), + share_type=sharetype, si=si_s, shnum=shnum, reason=reason, + level=log.SCARY, umid="SGx2fA") + + +class DiskShareSet(ShareSet): + implements(IShareSet) + + def __init__(self, storageindex, sharehomedir, incominghomedir=None, discard_storage=False): + ShareSet.__init__(self, storageindex) + self._sharehomedir = sharehomedir + self._incominghomedir = incominghomedir + self._discard_storage = discard_storage + + def get_overhead(self): + return (fileutil.get_disk_usage(self._sharehomedir) + + fileutil.get_disk_usage(self._incominghomedir)) + + def get_shares(self): + """ + Generate IStorageBackendShare objects for shares we have for this storage index. + ("Shares we have" means completed ones, excluding incoming ones.) + """ + try: + for fp in self._sharehomedir.children(): + shnumstr = fp.basename() + if not NUM_RE.match(shnumstr): + continue + sharehome = self._sharehomedir.child(shnumstr) + yield self.get_share(sharehome) + except UnlistableError: + # There is no shares directory at all. + pass + + def has_incoming(self, shnum): + if self._incominghomedir is None: + return False + return self._incominghomedir.child(str(shnum)).exists() + + def make_bucket_writer(self, storageserver, shnum, max_space_per_bucket, lease_info, canary): + sharehome = self._sharehomedir.child(str(shnum)) + incominghome = self._incominghomedir.child(str(shnum)) + immsh = ImmutableDiskShare(self.get_storage_index(), shnum, sharehome, incominghome, + max_size=max_space_per_bucket, create=True) + bw = BucketWriter(storageserver, immsh, max_space_per_bucket, lease_info, canary) + if self._discard_storage: + bw.throw_out_all_data = True + return bw + + def _create_mutable_share(self, storageserver, shnum, write_enabler): + fileutil.fp_make_dirs(self._sharehomedir) + sharehome = self._sharehomedir.child(str(shnum)) + serverid = storageserver.get_serverid() + return create_mutable_disk_share(sharehome, serverid, write_enabler, storageserver) + + def _clean_up_after_unlink(self): + fileutil.fp_rmdir_if_empty(self._sharehomedir) + hunk ./src/allmydata/storage/backends/disk/immutable.py 1 -import os, stat, struct, time hunk ./src/allmydata/storage/backends/disk/immutable.py 2 -from foolscap.api import Referenceable +import struct from zope.interface import implements hunk ./src/allmydata/storage/backends/disk/immutable.py 5 -from allmydata.interfaces import RIBucketWriter, RIBucketReader -from allmydata.util import base32, fileutil, log + +from allmydata.interfaces import IStoredShare +from allmydata.util import fileutil from allmydata.util.assertutil import precondition hunk ./src/allmydata/storage/backends/disk/immutable.py 9 +from allmydata.util.fileutil import fp_make_dirs from allmydata.util.hashutil import constant_time_compare hunk ./src/allmydata/storage/backends/disk/immutable.py 11 +from allmydata.util.encodingutil import quote_filepath +from allmydata.storage.common import si_b2a, UnknownImmutableContainerVersionError, DataTooLargeError from allmydata.storage.lease import LeaseInfo hunk ./src/allmydata/storage/backends/disk/immutable.py 14 -from allmydata.storage.common import UnknownImmutableContainerVersionError, \ - DataTooLargeError + # each share file (in storage/shares/$SI/$SHNUM) contains lease information # and share data. The share data is accessed by RIBucketWriter.write and hunk ./src/allmydata/storage/backends/disk/immutable.py 41 # then the value stored in this field will be the actual share data length # modulo 2**32. -class ShareFile: - LEASE_SIZE = struct.calcsize(">L32s32sL") +class ImmutableDiskShare(object): + implements(IStoredShare) + sharetype = "immutable" hunk ./src/allmydata/storage/backends/disk/immutable.py 45 + LEASE_SIZE = struct.calcsize(">L32s32sL") + hunk ./src/allmydata/storage/backends/disk/immutable.py 48 - def __init__(self, filename, max_size=None, create=False): - """ If max_size is not None then I won't allow more than max_size to be written to me. If create=True and max_size must not be None. """ + def __init__(self, storageindex, shnum, finalhome=None, incominghome=None, max_size=None, create=False): + """ If max_size is not None then I won't allow more than + max_size to be written to me. If create=True then max_size + must not be None. """ precondition((max_size is not None) or (not create), max_size, create) hunk ./src/allmydata/storage/backends/disk/immutable.py 53 - self.home = filename + self._storageindex = storageindex self._max_size = max_size hunk ./src/allmydata/storage/backends/disk/immutable.py 55 + self._incominghome = incominghome + self._home = finalhome + self._shnum = shnum if create: # touch the file, so later callers will see that we're working on # it. Also construct the metadata. hunk ./src/allmydata/storage/backends/disk/immutable.py 61 - assert not os.path.exists(self.home) - fileutil.make_dirs(os.path.dirname(self.home)) - f = open(self.home, 'wb') + assert not finalhome.exists() + fp_make_dirs(self._incominghome.parent()) # The second field -- the four-byte share data length -- is no # longer used as of Tahoe v1.3.0, but we continue to write it in # there in case someone downgrades a storage server from >= hunk ./src/allmydata/storage/backends/disk/immutable.py 72 # the largest length that can fit into the field. That way, even # if this does happen, the old < v1.3.0 server will still allow # clients to read the first part of the share. - f.write(struct.pack(">LLL", 1, min(2**32-1, max_size), 0)) - f.close() + self._incominghome.setContent(struct.pack(">LLL", 1, min(2**32-1, max_size), 0) ) self._lease_offset = max_size + 0x0c self._num_leases = 0 else: hunk ./src/allmydata/storage/backends/disk/immutable.py 76 - f = open(self.home, 'rb') - filesize = os.path.getsize(self.home) - (version, unused, num_leases) = struct.unpack(">LLL", f.read(0xc)) - f.close() + f = self._home.open(mode='rb') + try: + (version, unused, num_leases) = struct.unpack(">LLL", f.read(0xc)) + finally: + f.close() + filesize = self._home.getsize() if version != 1: msg = "sharefile %s had version %d but we wanted 1" % \ hunk ./src/allmydata/storage/backends/disk/immutable.py 84 - (filename, version) + (self._home, version) raise UnknownImmutableContainerVersionError(msg) self._num_leases = num_leases self._lease_offset = filesize - (num_leases * self.LEASE_SIZE) hunk ./src/allmydata/storage/backends/disk/immutable.py 90 self._data_offset = 0xc + def __repr__(self): + return ("" + % (si_b2a(self._storageindex), self._shnum, quote_filepath(self._home))) + + def close(self): + fileutil.fp_make_dirs(self._home.parent()) + self._incominghome.moveTo(self._home) + try: + # self._incominghome is like storage/shares/incoming/ab/abcde/4 . + # We try to delete the parent (.../ab/abcde) to avoid leaving + # these directories lying around forever, but the delete might + # fail if we're working on another share for the same storage + # index (like ab/abcde/5). The alternative approach would be to + # use a hierarchy of objects (PrefixHolder, BucketHolder, + # ShareWriter), each of which is responsible for a single + # directory on disk, and have them use reference counting of + # their children to know when they should do the rmdir. This + # approach is simpler, but relies on os.rmdir refusing to delete + # a non-empty directory. Do *not* use fileutil.fp_remove() here! + fileutil.fp_rmdir_if_empty(self._incominghome.parent()) + # we also delete the grandparent (prefix) directory, .../ab , + # again to avoid leaving directories lying around. This might + # fail if there is another bucket open that shares a prefix (like + # ab/abfff). + fileutil.fp_rmdir_if_empty(self._incominghome.parent().parent()) + # we leave the great-grandparent (incoming/) directory in place. + except EnvironmentError: + # ignore the "can't rmdir because the directory is not empty" + # exceptions, those are normal consequences of the + # above-mentioned conditions. + pass + pass + + def get_used_space(self): + return (fileutil.get_used_space(self._home) + + fileutil.get_used_space(self._incominghome)) + + def get_storage_index(self): + return self._storageindex + + def get_shnum(self): + return self._shnum + def unlink(self): hunk ./src/allmydata/storage/backends/disk/immutable.py 134 - os.unlink(self.home) + self._home.remove() + + def get_size(self): + return self._home.getsize() + + def get_data_length(self): + return self._lease_offset - self._data_offset + + #def readv(self, read_vector): + # ... def read_share_data(self, offset, length): precondition(offset >= 0) hunk ./src/allmydata/storage/backends/disk/immutable.py 147 - # reads beyond the end of the data are truncated. Reads that start + + # Reads beyond the end of the data are truncated. Reads that start # beyond the end of the data return an empty string. seekpos = self._data_offset+offset actuallength = max(0, min(length, self._lease_offset-seekpos)) hunk ./src/allmydata/storage/backends/disk/immutable.py 154 if actuallength == 0: return "" - f = open(self.home, 'rb') - f.seek(seekpos) - return f.read(actuallength) + f = self._home.open(mode='rb') + try: + f.seek(seekpos) + sharedata = f.read(actuallength) + finally: + f.close() + return sharedata def write_share_data(self, offset, data): length = len(data) hunk ./src/allmydata/storage/backends/disk/immutable.py 167 precondition(offset >= 0, offset) if self._max_size is not None and offset+length > self._max_size: raise DataTooLargeError(self._max_size, offset, length) - f = open(self.home, 'rb+') - real_offset = self._data_offset+offset - f.seek(real_offset) - assert f.tell() == real_offset - f.write(data) - f.close() + f = self._incominghome.open(mode='rb+') + try: + real_offset = self._data_offset+offset + f.seek(real_offset) + assert f.tell() == real_offset + f.write(data) + finally: + f.close() def _write_lease_record(self, f, lease_number, lease_info): offset = self._lease_offset + lease_number * self.LEASE_SIZE hunk ./src/allmydata/storage/backends/disk/immutable.py 184 def _read_num_leases(self, f): f.seek(0x08) - (num_leases,) = struct.unpack(">L", f.read(4)) + ro = f.read(4) + (num_leases,) = struct.unpack(">L", ro) return num_leases def _write_num_leases(self, f, num_leases): hunk ./src/allmydata/storage/backends/disk/immutable.py 195 def _truncate_leases(self, f, num_leases): f.truncate(self._lease_offset + num_leases * self.LEASE_SIZE) + # These lease operations are intended for use by disk_backend.py. + # Other clients should not depend on the fact that the disk backend + # stores leases in share files. + def get_leases(self): """Yields a LeaseInfo instance for all leases.""" hunk ./src/allmydata/storage/backends/disk/immutable.py 201 - f = open(self.home, 'rb') - (version, unused, num_leases) = struct.unpack(">LLL", f.read(0xc)) - f.seek(self._lease_offset) - for i in range(num_leases): - data = f.read(self.LEASE_SIZE) - if data: - yield LeaseInfo().from_immutable_data(data) + f = self._home.open(mode='rb') + try: + (version, unused, num_leases) = struct.unpack(">LLL", f.read(0xc)) + f.seek(self._lease_offset) + for i in range(num_leases): + data = f.read(self.LEASE_SIZE) + if data: + yield LeaseInfo().from_immutable_data(data) + finally: + f.close() def add_lease(self, lease_info): hunk ./src/allmydata/storage/backends/disk/immutable.py 213 - f = open(self.home, 'rb+') - num_leases = self._read_num_leases(f) - self._write_lease_record(f, num_leases, lease_info) - self._write_num_leases(f, num_leases+1) - f.close() + f = self._incominghome.open(mode='rb') + try: + num_leases = self._read_num_leases(f) + finally: + f.close() + f = self._home.open(mode='wb+') + try: + self._write_lease_record(f, num_leases, lease_info) + self._write_num_leases(f, num_leases+1) + finally: + f.close() def renew_lease(self, renew_secret, new_expire_time): hunk ./src/allmydata/storage/backends/disk/immutable.py 226 - for i,lease in enumerate(self.get_leases()): - if constant_time_compare(lease.renew_secret, renew_secret): - # yup. See if we need to update the owner time. - if new_expire_time > lease.expiration_time: - # yes - lease.expiration_time = new_expire_time - f = open(self.home, 'rb+') - self._write_lease_record(f, i, lease) - f.close() - return + try: + for i, lease in enumerate(self.get_leases()): + if constant_time_compare(lease.renew_secret, renew_secret): + # yup. See if we need to update the owner time. + if new_expire_time > lease.expiration_time: + # yes + lease.expiration_time = new_expire_time + f = self._home.open('rb+') + try: + self._write_lease_record(f, i, lease) + finally: + f.close() + return + except IndexError, e: + raise Exception("IndexError: %s" % (e,)) raise IndexError("unable to renew non-existent lease") def add_or_renew_lease(self, lease_info): hunk ./src/allmydata/storage/backends/disk/immutable.py 249 lease_info.expiration_time) except IndexError: self.add_lease(lease_info) - - - def cancel_lease(self, cancel_secret): - """Remove a lease with the given cancel_secret. If the last lease is - cancelled, the file will be removed. Return the number of bytes that - were freed (by truncating the list of leases, and possibly by - deleting the file. Raise IndexError if there was no lease with the - given cancel_secret. - """ - - leases = list(self.get_leases()) - num_leases_removed = 0 - for i,lease in enumerate(leases): - if constant_time_compare(lease.cancel_secret, cancel_secret): - leases[i] = None - num_leases_removed += 1 - if not num_leases_removed: - raise IndexError("unable to find matching lease to cancel") - if num_leases_removed: - # pack and write out the remaining leases. We write these out in - # the same order as they were added, so that if we crash while - # doing this, we won't lose any non-cancelled leases. - leases = [l for l in leases if l] # remove the cancelled leases - f = open(self.home, 'rb+') - for i,lease in enumerate(leases): - self._write_lease_record(f, i, lease) - self._write_num_leases(f, len(leases)) - self._truncate_leases(f, len(leases)) - f.close() - space_freed = self.LEASE_SIZE * num_leases_removed - if not len(leases): - space_freed += os.stat(self.home)[stat.ST_SIZE] - self.unlink() - return space_freed - - -class BucketWriter(Referenceable): - implements(RIBucketWriter) - - def __init__(self, ss, incominghome, finalhome, max_size, lease_info, canary): - self.ss = ss - self.incominghome = incominghome - self.finalhome = finalhome - self._max_size = max_size # don't allow the client to write more than this - self._canary = canary - self._disconnect_marker = canary.notifyOnDisconnect(self._disconnected) - self.closed = False - self.throw_out_all_data = False - self._sharefile = ShareFile(incominghome, create=True, max_size=max_size) - # also, add our lease to the file now, so that other ones can be - # added by simultaneous uploaders - self._sharefile.add_lease(lease_info) - - def allocated_size(self): - return self._max_size - - def remote_write(self, offset, data): - start = time.time() - precondition(not self.closed) - if self.throw_out_all_data: - return - self._sharefile.write_share_data(offset, data) - self.ss.add_latency("write", time.time() - start) - self.ss.count("write") - - def remote_close(self): - precondition(not self.closed) - start = time.time() - - fileutil.make_dirs(os.path.dirname(self.finalhome)) - fileutil.rename(self.incominghome, self.finalhome) - try: - # self.incominghome is like storage/shares/incoming/ab/abcde/4 . - # We try to delete the parent (.../ab/abcde) to avoid leaving - # these directories lying around forever, but the delete might - # fail if we're working on another share for the same storage - # index (like ab/abcde/5). The alternative approach would be to - # use a hierarchy of objects (PrefixHolder, BucketHolder, - # ShareWriter), each of which is responsible for a single - # directory on disk, and have them use reference counting of - # their children to know when they should do the rmdir. This - # approach is simpler, but relies on os.rmdir refusing to delete - # a non-empty directory. Do *not* use fileutil.rm_dir() here! - os.rmdir(os.path.dirname(self.incominghome)) - # we also delete the grandparent (prefix) directory, .../ab , - # again to avoid leaving directories lying around. This might - # fail if there is another bucket open that shares a prefix (like - # ab/abfff). - os.rmdir(os.path.dirname(os.path.dirname(self.incominghome))) - # we leave the great-grandparent (incoming/) directory in place. - except EnvironmentError: - # ignore the "can't rmdir because the directory is not empty" - # exceptions, those are normal consequences of the - # above-mentioned conditions. - pass - self._sharefile = None - self.closed = True - self._canary.dontNotifyOnDisconnect(self._disconnect_marker) - - filelen = os.stat(self.finalhome)[stat.ST_SIZE] - self.ss.bucket_writer_closed(self, filelen) - self.ss.add_latency("close", time.time() - start) - self.ss.count("close") - - def _disconnected(self): - if not self.closed: - self._abort() - - def remote_abort(self): - log.msg("storage: aborting sharefile %s" % self.incominghome, - facility="tahoe.storage", level=log.UNUSUAL) - if not self.closed: - self._canary.dontNotifyOnDisconnect(self._disconnect_marker) - self._abort() - self.ss.count("abort") - - def _abort(self): - if self.closed: - return - - os.remove(self.incominghome) - # if we were the last share to be moved, remove the incoming/ - # directory that was our parent - parentdir = os.path.split(self.incominghome)[0] - if not os.listdir(parentdir): - os.rmdir(parentdir) - self._sharefile = None - - # We are now considered closed for further writing. We must tell - # the storage server about this so that it stops expecting us to - # use the space it allocated for us earlier. - self.closed = True - self.ss.bucket_writer_closed(self, 0) - - -class BucketReader(Referenceable): - implements(RIBucketReader) - - def __init__(self, ss, sharefname, storage_index=None, shnum=None): - self.ss = ss - self._share_file = ShareFile(sharefname) - self.storage_index = storage_index - self.shnum = shnum - - def __repr__(self): - return "<%s %s %s>" % (self.__class__.__name__, - base32.b2a_l(self.storage_index[:8], 60), - self.shnum) - - def remote_read(self, offset, length): - start = time.time() - data = self._share_file.read_share_data(offset, length) - self.ss.add_latency("read", time.time() - start) - self.ss.count("read") - return data - - def remote_advise_corrupt_share(self, reason): - return self.ss.remote_advise_corrupt_share("immutable", - self.storage_index, - self.shnum, - reason) hunk ./src/allmydata/storage/backends/disk/mutable.py 1 -import os, stat, struct hunk ./src/allmydata/storage/backends/disk/mutable.py 2 -from allmydata.interfaces import BadWriteEnablerError -from allmydata.util import idlib, log +import struct + +from zope.interface import implements + +from allmydata.interfaces import IStoredMutableShare, BadWriteEnablerError +from allmydata.util import fileutil, idlib, log from allmydata.util.assertutil import precondition from allmydata.util.hashutil import constant_time_compare hunk ./src/allmydata/storage/backends/disk/mutable.py 10 -from allmydata.storage.lease import LeaseInfo -from allmydata.storage.common import UnknownMutableContainerVersionError, \ +from allmydata.util.encodingutil import quote_filepath +from allmydata.storage.common import si_b2a, UnknownMutableContainerVersionError, \ DataTooLargeError hunk ./src/allmydata/storage/backends/disk/mutable.py 13 +from allmydata.storage.lease import LeaseInfo +from allmydata.storage.backends.base import testv_compare hunk ./src/allmydata/storage/backends/disk/mutable.py 16 -# the MutableShareFile is like the ShareFile, but used for mutable data. It -# has a different layout. See docs/mutable.txt for more details. + +# The MutableDiskShare is like the ImmutableDiskShare, but used for mutable data. +# It has a different layout. See docs/mutable.rst for more details. # # offset size name # 1 0 32 magic verstr "tahoe mutable container v1" plus binary hunk ./src/allmydata/storage/backends/disk/mutable.py 31 # 4 4 expiration timestamp # 8 32 renewal token # 40 32 cancel token -# 72 20 nodeid which accepted the tokens +# 72 20 nodeid that accepted the tokens # 7 468 (a) data # 8 ?? 4 count of extra leases # 9 ?? n*92 extra leases hunk ./src/allmydata/storage/backends/disk/mutable.py 37 -# The struct module doc says that L's are 4 bytes in size., and that Q's are +# The struct module doc says that L's are 4 bytes in size, and that Q's are # 8 bytes in size. Since compatibility depends upon this, double-check it. assert struct.calcsize(">L") == 4, struct.calcsize(">L") assert struct.calcsize(">Q") == 8, struct.calcsize(">Q") hunk ./src/allmydata/storage/backends/disk/mutable.py 42 -class MutableShareFile: + +class MutableDiskShare(object): + implements(IStoredMutableShare) sharetype = "mutable" DATA_LENGTH_OFFSET = struct.calcsize(">32s20s32s") hunk ./src/allmydata/storage/backends/disk/mutable.py 54 assert LEASE_SIZE == 92 DATA_OFFSET = HEADER_SIZE + 4*LEASE_SIZE assert DATA_OFFSET == 468, DATA_OFFSET + # our sharefiles share with a recognizable string, plus some random # binary data to reduce the chance that a regular text file will look # like a sharefile. hunk ./src/allmydata/storage/backends/disk/mutable.py 63 MAX_SIZE = 2*1000*1000*1000 # 2GB, kind of arbitrary # TODO: decide upon a policy for max share size - def __init__(self, filename, parent=None): - self.home = filename - if os.path.exists(self.home): + def __init__(self, storageindex, shnum, home, parent=None): + self._storageindex = storageindex + self._shnum = shnum + self._home = home + if self._home.exists(): # we don't cache anything, just check the magic hunk ./src/allmydata/storage/backends/disk/mutable.py 69 - f = open(self.home, 'rb') - data = f.read(self.HEADER_SIZE) - (magic, - write_enabler_nodeid, write_enabler, - data_length, extra_least_offset) = \ - struct.unpack(">32s20s32sQQ", data) - if magic != self.MAGIC: - msg = "sharefile %s had magic '%r' but we wanted '%r'" % \ - (filename, magic, self.MAGIC) - raise UnknownMutableContainerVersionError(msg) + f = self._home.open('rb') + try: + data = f.read(self.HEADER_SIZE) + (magic, + write_enabler_nodeid, write_enabler, + data_length, extra_least_offset) = \ + struct.unpack(">32s20s32sQQ", data) + if magic != self.MAGIC: + msg = "sharefile %s had magic '%r' but we wanted '%r'" % \ + (quote_filepath(self._home), magic, self.MAGIC) + raise UnknownMutableContainerVersionError(msg) + finally: + f.close() self.parent = parent # for logging def log(self, *args, **kwargs): hunk ./src/allmydata/storage/backends/disk/mutable.py 87 return self.parent.log(*args, **kwargs) - def create(self, my_nodeid, write_enabler): - assert not os.path.exists(self.home) + def create(self, serverid, write_enabler): + assert not self._home.exists() data_length = 0 extra_lease_offset = (self.HEADER_SIZE + 4 * self.LEASE_SIZE hunk ./src/allmydata/storage/backends/disk/mutable.py 95 + data_length) assert extra_lease_offset == self.DATA_OFFSET # true at creation num_extra_leases = 0 - f = open(self.home, 'wb') - header = struct.pack(">32s20s32sQQ", - self.MAGIC, my_nodeid, write_enabler, - data_length, extra_lease_offset, - ) - leases = ("\x00"*self.LEASE_SIZE) * 4 - f.write(header + leases) - # data goes here, empty after creation - f.write(struct.pack(">L", num_extra_leases)) - # extra leases go here, none at creation - f.close() + f = self._home.open('wb') + try: + header = struct.pack(">32s20s32sQQ", + self.MAGIC, serverid, write_enabler, + data_length, extra_lease_offset, + ) + leases = ("\x00"*self.LEASE_SIZE) * 4 + f.write(header + leases) + # data goes here, empty after creation + f.write(struct.pack(">L", num_extra_leases)) + # extra leases go here, none at creation + finally: + f.close() + + def __repr__(self): + return ("" + % (si_b2a(self._storageindex), self._shnum, quote_filepath(self._home))) + + def get_used_space(self): + return fileutil.get_used_space(self._home) + + def get_storage_index(self): + return self._storageindex + + def get_shnum(self): + return self._shnum def unlink(self): hunk ./src/allmydata/storage/backends/disk/mutable.py 123 - os.unlink(self.home) + self._home.remove() def _read_data_length(self, f): f.seek(self.DATA_LENGTH_OFFSET) hunk ./src/allmydata/storage/backends/disk/mutable.py 291 def get_leases(self): """Yields a LeaseInfo instance for all leases.""" - f = open(self.home, 'rb') - for i, lease in self._enumerate_leases(f): - yield lease - f.close() + f = self._home.open('rb') + try: + for i, lease in self._enumerate_leases(f): + yield lease + finally: + f.close() def _enumerate_leases(self, f): for i in range(self._get_num_lease_slots(f)): hunk ./src/allmydata/storage/backends/disk/mutable.py 303 try: data = self._read_lease_record(f, i) if data is not None: - yield i,data + yield i, data except IndexError: return hunk ./src/allmydata/storage/backends/disk/mutable.py 307 + # These lease operations are intended for use by disk_backend.py. + # Other non-test clients should not depend on the fact that the disk + # backend stores leases in share files. + def add_lease(self, lease_info): precondition(lease_info.owner_num != 0) # 0 means "no lease here" hunk ./src/allmydata/storage/backends/disk/mutable.py 313 - f = open(self.home, 'rb+') - num_lease_slots = self._get_num_lease_slots(f) - empty_slot = self._get_first_empty_lease_slot(f) - if empty_slot is not None: - self._write_lease_record(f, empty_slot, lease_info) - else: - self._write_lease_record(f, num_lease_slots, lease_info) - f.close() + f = self._home.open('rb+') + try: + num_lease_slots = self._get_num_lease_slots(f) + empty_slot = self._get_first_empty_lease_slot(f) + if empty_slot is not None: + self._write_lease_record(f, empty_slot, lease_info) + else: + self._write_lease_record(f, num_lease_slots, lease_info) + finally: + f.close() def renew_lease(self, renew_secret, new_expire_time): accepting_nodeids = set() hunk ./src/allmydata/storage/backends/disk/mutable.py 326 - f = open(self.home, 'rb+') - for (leasenum,lease) in self._enumerate_leases(f): - if constant_time_compare(lease.renew_secret, renew_secret): - # yup. See if we need to update the owner time. - if new_expire_time > lease.expiration_time: - # yes - lease.expiration_time = new_expire_time - self._write_lease_record(f, leasenum, lease) - f.close() - return - accepting_nodeids.add(lease.nodeid) - f.close() + f = self._home.open('rb+') + try: + for (leasenum, lease) in self._enumerate_leases(f): + if constant_time_compare(lease.renew_secret, renew_secret): + # yup. See if we need to update the owner time. + if new_expire_time > lease.expiration_time: + # yes + lease.expiration_time = new_expire_time + self._write_lease_record(f, leasenum, lease) + return + accepting_nodeids.add(lease.nodeid) + finally: + f.close() # Return the accepting_nodeids set, to give the client a chance to hunk ./src/allmydata/storage/backends/disk/mutable.py 340 - # update the leases on a share which has been migrated from its + # update the leases on a share that has been migrated from its # original server to a new one. msg = ("Unable to renew non-existent lease. I have leases accepted by" " nodeids: ") hunk ./src/allmydata/storage/backends/disk/mutable.py 357 except IndexError: self.add_lease(lease_info) - def cancel_lease(self, cancel_secret): - """Remove any leases with the given cancel_secret. If the last lease - is cancelled, the file will be removed. Return the number of bytes - that were freed (by truncating the list of leases, and possibly by - deleting the file. Raise IndexError if there was no lease with the - given cancel_secret.""" - - accepting_nodeids = set() - modified = 0 - remaining = 0 - blank_lease = LeaseInfo(owner_num=0, - renew_secret="\x00"*32, - cancel_secret="\x00"*32, - expiration_time=0, - nodeid="\x00"*20) - f = open(self.home, 'rb+') - for (leasenum,lease) in self._enumerate_leases(f): - accepting_nodeids.add(lease.nodeid) - if constant_time_compare(lease.cancel_secret, cancel_secret): - self._write_lease_record(f, leasenum, blank_lease) - modified += 1 - else: - remaining += 1 - if modified: - freed_space = self._pack_leases(f) - f.close() - if not remaining: - freed_space += os.stat(self.home)[stat.ST_SIZE] - self.unlink() - return freed_space - - msg = ("Unable to cancel non-existent lease. I have leases " - "accepted by nodeids: ") - msg += ",".join([("'%s'" % idlib.nodeid_b2a(anid)) - for anid in accepting_nodeids]) - msg += " ." - raise IndexError(msg) - - def _pack_leases(self, f): - # TODO: reclaim space from cancelled leases - return 0 - def _read_write_enabler_and_nodeid(self, f): f.seek(0) data = f.read(self.HEADER_SIZE) hunk ./src/allmydata/storage/backends/disk/mutable.py 369 def readv(self, readv): datav = [] - f = open(self.home, 'rb') - for (offset, length) in readv: - datav.append(self._read_share_data(f, offset, length)) - f.close() + f = self._home.open('rb') + try: + for (offset, length) in readv: + datav.append(self._read_share_data(f, offset, length)) + finally: + f.close() return datav hunk ./src/allmydata/storage/backends/disk/mutable.py 377 -# def remote_get_length(self): -# f = open(self.home, 'rb') -# data_length = self._read_data_length(f) -# f.close() -# return data_length + def get_size(self): + return self._home.getsize() + + def get_data_length(self): + f = self._home.open('rb') + try: + data_length = self._read_data_length(f) + finally: + f.close() + return data_length def check_write_enabler(self, write_enabler, si_s): hunk ./src/allmydata/storage/backends/disk/mutable.py 389 - f = open(self.home, 'rb+') - (real_write_enabler, write_enabler_nodeid) = \ - self._read_write_enabler_and_nodeid(f) - f.close() + f = self._home.open('rb+') + try: + (real_write_enabler, write_enabler_nodeid) = self._read_write_enabler_and_nodeid(f) + finally: + f.close() # avoid a timing attack #if write_enabler != real_write_enabler: if not constant_time_compare(write_enabler, real_write_enabler): hunk ./src/allmydata/storage/backends/disk/mutable.py 410 def check_testv(self, testv): test_good = True - f = open(self.home, 'rb+') - for (offset, length, operator, specimen) in testv: - data = self._read_share_data(f, offset, length) - if not testv_compare(data, operator, specimen): - test_good = False - break - f.close() + f = self._home.open('rb+') + try: + for (offset, length, operator, specimen) in testv: + data = self._read_share_data(f, offset, length) + if not testv_compare(data, operator, specimen): + test_good = False + break + finally: + f.close() return test_good def writev(self, datav, new_length): hunk ./src/allmydata/storage/backends/disk/mutable.py 422 - f = open(self.home, 'rb+') - for (offset, data) in datav: - self._write_share_data(f, offset, data) - if new_length is not None: - cur_length = self._read_data_length(f) - if new_length < cur_length: - self._write_data_length(f, new_length) - # TODO: if we're going to shrink the share file when the - # share data has shrunk, then call - # self._change_container_size() here. - f.close() - -def testv_compare(a, op, b): - assert op in ("lt", "le", "eq", "ne", "ge", "gt") - if op == "lt": - return a < b - if op == "le": - return a <= b - if op == "eq": - return a == b - if op == "ne": - return a != b - if op == "ge": - return a >= b - if op == "gt": - return a > b - # never reached + f = self._home.open('rb+') + try: + for (offset, data) in datav: + self._write_share_data(f, offset, data) + if new_length is not None: + cur_length = self._read_data_length(f) + if new_length < cur_length: + self._write_data_length(f, new_length) + # TODO: if we're going to shrink the share file when the + # share data has shrunk, then call + # self._change_container_size() here. + finally: + f.close() hunk ./src/allmydata/storage/backends/disk/mutable.py 436 -class EmptyShare: + def close(self): + pass hunk ./src/allmydata/storage/backends/disk/mutable.py 439 - def check_testv(self, testv): - test_good = True - for (offset, length, operator, specimen) in testv: - data = "" - if not testv_compare(data, operator, specimen): - test_good = False - break - return test_good hunk ./src/allmydata/storage/backends/disk/mutable.py 440 -def create_mutable_sharefile(filename, my_nodeid, write_enabler, parent): - ms = MutableShareFile(filename, parent) - ms.create(my_nodeid, write_enabler) +def create_mutable_disk_share(fp, serverid, write_enabler, parent): + ms = MutableDiskShare(fp, parent) + ms.create(serverid, write_enabler) del ms hunk ./src/allmydata/storage/backends/disk/mutable.py 444 - return MutableShareFile(filename, parent) - + return MutableDiskShare(fp, parent) addfile ./src/allmydata/storage/backends/null/__init__.py addfile ./src/allmydata/storage/backends/null/null_backend.py hunk ./src/allmydata/storage/backends/null/null_backend.py 2 +import os, struct + +from zope.interface import implements + +from allmydata.interfaces import IStorageBackend, IShareSet, IStoredShare, IStoredMutableShare +from allmydata.util.assertutil import precondition +from allmydata.util.hashutil import constant_time_compare +from allmydata.storage.backends.base import Backend, ShareSet +from allmydata.storage.bucket import BucketWriter +from allmydata.storage.common import si_b2a +from allmydata.storage.lease import LeaseInfo + + +class NullBackend(Backend): + implements(IStorageBackend) + + def __init__(self): + Backend.__init__(self) + + def get_available_space(self, reserved_space): + return None + + def get_sharesets_for_prefix(self, prefix): + pass + + def get_shareset(self, storageindex): + return NullShareSet(storageindex) + + def fill_in_space_stats(self, stats): + pass + + def set_storage_server(self, ss): + self.ss = ss + + def advise_corrupt_share(self, sharetype, storageindex, shnum, reason): + pass + + +class NullShareSet(ShareSet): + implements(IShareSet) + + def __init__(self, storageindex): + self.storageindex = storageindex + + def get_overhead(self): + return 0 + + def get_incoming_shnums(self): + return frozenset() + + def get_shares(self): + pass + + def get_share(self, shnum): + return None + + def get_storage_index(self): + return self.storageindex + + def get_storage_index_string(self): + return si_b2a(self.storageindex) + + def make_bucket_writer(self, storageserver, shnum, max_space_per_bucket, lease_info, canary): + immutableshare = ImmutableNullShare() + return BucketWriter(self.ss, immutableshare, max_space_per_bucket, lease_info, canary) + + def _create_mutable_share(self, storageserver, shnum, write_enabler): + return MutableNullShare() + + def _clean_up_after_unlink(self): + pass + + +class ImmutableNullShare: + implements(IStoredShare) + sharetype = "immutable" + + def __init__(self): + """ If max_size is not None then I won't allow more than + max_size to be written to me. If create=True then max_size + must not be None. """ + pass + + def get_shnum(self): + return self.shnum + + def unlink(self): + os.unlink(self.fname) + + def read_share_data(self, offset, length): + precondition(offset >= 0) + # Reads beyond the end of the data are truncated. Reads that start + # beyond the end of the data return an empty string. + seekpos = self._data_offset+offset + fsize = os.path.getsize(self.fname) + actuallength = max(0, min(length, fsize-seekpos)) # XXX #1528 + if actuallength == 0: + return "" + f = open(self.fname, 'rb') + f.seek(seekpos) + return f.read(actuallength) + + def write_share_data(self, offset, data): + pass + + def _write_lease_record(self, f, lease_number, lease_info): + offset = self._lease_offset + lease_number * self.LEASE_SIZE + f.seek(offset) + assert f.tell() == offset + f.write(lease_info.to_immutable_data()) + + def _read_num_leases(self, f): + f.seek(0x08) + (num_leases,) = struct.unpack(">L", f.read(4)) + return num_leases + + def _write_num_leases(self, f, num_leases): + f.seek(0x08) + f.write(struct.pack(">L", num_leases)) + + def _truncate_leases(self, f, num_leases): + f.truncate(self._lease_offset + num_leases * self.LEASE_SIZE) + + def get_leases(self): + """Yields a LeaseInfo instance for all leases.""" + f = open(self.fname, 'rb') + (version, unused, num_leases) = struct.unpack(">LLL", f.read(0xc)) + f.seek(self._lease_offset) + for i in range(num_leases): + data = f.read(self.LEASE_SIZE) + if data: + yield LeaseInfo().from_immutable_data(data) + + def add_lease(self, lease): + pass + + def renew_lease(self, renew_secret, new_expire_time): + for i,lease in enumerate(self.get_leases()): + if constant_time_compare(lease.renew_secret, renew_secret): + # yup. See if we need to update the owner time. + if new_expire_time > lease.expiration_time: + # yes + lease.expiration_time = new_expire_time + f = open(self.fname, 'rb+') + self._write_lease_record(f, i, lease) + f.close() + return + raise IndexError("unable to renew non-existent lease") + + def add_or_renew_lease(self, lease_info): + try: + self.renew_lease(lease_info.renew_secret, + lease_info.expiration_time) + except IndexError: + self.add_lease(lease_info) + + +class MutableNullShare: + implements(IStoredMutableShare) + sharetype = "mutable" + + """ XXX: TODO """ addfile ./src/allmydata/storage/bucket.py hunk ./src/allmydata/storage/bucket.py 1 + +import time + +from foolscap.api import Referenceable + +from zope.interface import implements +from allmydata.interfaces import RIBucketWriter, RIBucketReader +from allmydata.util import base32, log +from allmydata.util.assertutil import precondition + + +class BucketWriter(Referenceable): + implements(RIBucketWriter) + + def __init__(self, ss, immutableshare, max_size, lease_info, canary): + self.ss = ss + self._max_size = max_size # don't allow the client to write more than this + self._canary = canary + self._disconnect_marker = canary.notifyOnDisconnect(self._disconnected) + self.closed = False + self.throw_out_all_data = False + self._share = immutableshare + # also, add our lease to the file now, so that other ones can be + # added by simultaneous uploaders + self._share.add_lease(lease_info) + + def allocated_size(self): + return self._max_size + + def remote_write(self, offset, data): + start = time.time() + precondition(not self.closed) + if self.throw_out_all_data: + return + self._share.write_share_data(offset, data) + self.ss.add_latency("write", time.time() - start) + self.ss.count("write") + + def remote_close(self): + precondition(not self.closed) + start = time.time() + + self._share.close() + filelen = self._share.stat() + self._share = None + + self.closed = True + self._canary.dontNotifyOnDisconnect(self._disconnect_marker) + + self.ss.bucket_writer_closed(self, filelen) + self.ss.add_latency("close", time.time() - start) + self.ss.count("close") + + def _disconnected(self): + if not self.closed: + self._abort() + + def remote_abort(self): + log.msg("storage: aborting write to share %r" % self._share, + facility="tahoe.storage", level=log.UNUSUAL) + if not self.closed: + self._canary.dontNotifyOnDisconnect(self._disconnect_marker) + self._abort() + self.ss.count("abort") + + def _abort(self): + if self.closed: + return + self._share.unlink() + self._share = None + + # We are now considered closed for further writing. We must tell + # the storage server about this so that it stops expecting us to + # use the space it allocated for us earlier. + self.closed = True + self.ss.bucket_writer_closed(self, 0) + + +class BucketReader(Referenceable): + implements(RIBucketReader) + + def __init__(self, ss, share): + self.ss = ss + self._share = share + self.storageindex = share.storageindex + self.shnum = share.shnum + + def __repr__(self): + return "<%s %s %s>" % (self.__class__.__name__, + base32.b2a_l(self.storageindex[:8], 60), + self.shnum) + + def remote_read(self, offset, length): + start = time.time() + data = self._share.read_share_data(offset, length) + self.ss.add_latency("read", time.time() - start) + self.ss.count("read") + return data + + def remote_advise_corrupt_share(self, reason): + return self.ss.remote_advise_corrupt_share("immutable", + self.storageindex, + self.shnum, + reason) addfile ./src/allmydata/test/test_backends.py hunk ./src/allmydata/test/test_backends.py 1 +import os, stat +from twisted.trial import unittest +from allmydata.util.log import msg +from allmydata.test.common_util import ReallyEqualMixin +import mock + +# This is the code that we're going to be testing. +from allmydata.storage.server import StorageServer +from allmydata.storage.backends.disk.disk_backend import DiskBackend, si_si2dir +from allmydata.storage.backends.null.null_backend import NullBackend + +# The following share file content was generated with +# storage.immutable.ShareFile from Tahoe-LAFS v1.8.2 +# with share data == 'a'. The total size of this input +# is 85 bytes. +shareversionnumber = '\x00\x00\x00\x01' +sharedatalength = '\x00\x00\x00\x01' +numberofleases = '\x00\x00\x00\x01' +shareinputdata = 'a' +ownernumber = '\x00\x00\x00\x00' +renewsecret = 'x'*32 +cancelsecret = 'y'*32 +expirationtime = '\x00(\xde\x80' +nextlease = '' +containerdata = shareversionnumber + sharedatalength + numberofleases +client_data = shareinputdata + ownernumber + renewsecret + \ + cancelsecret + expirationtime + nextlease +share_data = containerdata + client_data +testnodeid = 'testnodeidxxxxxxxxxx' + + +class MockFileSystem(unittest.TestCase): + """ I simulate a filesystem that the code under test can use. I simulate + just the parts of the filesystem that the current implementation of Disk + backend needs. """ + def setUp(self): + # Make patcher, patch, and effects for disk-using functions. + msg( "%s.setUp()" % (self,)) + self.mockedfilepaths = {} + # keys are pathnames, values are MockFilePath objects. This is necessary because + # MockFilePath behavior sometimes depends on the filesystem. Where it does, + # self.mockedfilepaths has the relevant information. + self.storedir = MockFilePath('teststoredir', self.mockedfilepaths) + self.basedir = self.storedir.child('shares') + self.baseincdir = self.basedir.child('incoming') + self.sharedirfinalname = self.basedir.child('or').child('orsxg5dtorxxeylhmvpws3temv4a') + self.sharedirincomingname = self.baseincdir.child('or').child('orsxg5dtorxxeylhmvpws3temv4a') + self.shareincomingname = self.sharedirincomingname.child('0') + self.sharefinalname = self.sharedirfinalname.child('0') + + # FIXME: these patches won't work; disk_backend no longer imports FilePath, BucketCountingCrawler, + # or LeaseCheckingCrawler. + + self.FilePathFake = mock.patch('allmydata.storage.backends.disk.disk_backend.FilePath', new = MockFilePath) + self.FilePathFake.__enter__() + + self.BCountingCrawler = mock.patch('allmydata.storage.backends.disk.disk_backend.BucketCountingCrawler') + FakeBCC = self.BCountingCrawler.__enter__() + FakeBCC.side_effect = self.call_FakeBCC + + self.LeaseCheckingCrawler = mock.patch('allmydata.storage.backends.disk.disk_backend.LeaseCheckingCrawler') + FakeLCC = self.LeaseCheckingCrawler.__enter__() + FakeLCC.side_effect = self.call_FakeLCC + + self.get_available_space = mock.patch('allmydata.util.fileutil.get_available_space') + GetSpace = self.get_available_space.__enter__() + GetSpace.side_effect = self.call_get_available_space + + self.statforsize = mock.patch('allmydata.storage.backends.disk.core.filepath.stat') + getsize = self.statforsize.__enter__() + getsize.side_effect = self.call_statforsize + + def call_FakeBCC(self, StateFile): + return MockBCC() + + def call_FakeLCC(self, StateFile, HistoryFile, ExpirationPolicy): + return MockLCC() + + def call_get_available_space(self, storedir, reservedspace): + # The input vector has an input size of 85. + return 85 - reservedspace + + def call_statforsize(self, fakefpname): + return self.mockedfilepaths[fakefpname].fileobject.size() + + def tearDown(self): + msg( "%s.tearDown()" % (self,)) + self.FilePathFake.__exit__() + self.mockedfilepaths = {} + + +class MockFilePath: + def __init__(self, pathstring, ffpathsenvironment, existence=False): + # I can't just make the values MockFileObjects because they may be directories. + self.mockedfilepaths = ffpathsenvironment + self.path = pathstring + self.existence = existence + if not self.mockedfilepaths.has_key(self.path): + # The first MockFilePath object is special + self.mockedfilepaths[self.path] = self + self.fileobject = None + else: + self.fileobject = self.mockedfilepaths[self.path].fileobject + self.spawn = {} + self.antecedent = os.path.dirname(self.path) + + def setContent(self, contentstring): + # This method rewrites the data in the file that corresponds to its path + # name whether it preexisted or not. + self.fileobject = MockFileObject(contentstring) + self.existence = True + self.mockedfilepaths[self.path].fileobject = self.fileobject + self.mockedfilepaths[self.path].existence = self.existence + self.setparents() + + def create(self): + # This method chokes if there's a pre-existing file! + if self.mockedfilepaths[self.path].fileobject: + raise OSError + else: + self.existence = True + self.mockedfilepaths[self.path].fileobject = self.fileobject + self.mockedfilepaths[self.path].existence = self.existence + self.setparents() + + def open(self, mode='r'): + # XXX Makes no use of mode. + if not self.mockedfilepaths[self.path].fileobject: + # If there's no fileobject there already then make one and put it there. + self.fileobject = MockFileObject() + self.existence = True + self.mockedfilepaths[self.path].fileobject = self.fileobject + self.mockedfilepaths[self.path].existence = self.existence + else: + # Otherwise get a ref to it. + self.fileobject = self.mockedfilepaths[self.path].fileobject + self.existence = self.mockedfilepaths[self.path].existence + return self.fileobject.open(mode) + + def child(self, childstring): + arg2child = os.path.join(self.path, childstring) + child = MockFilePath(arg2child, self.mockedfilepaths) + return child + + def children(self): + childrenfromffs = [ffp for ffp in self.mockedfilepaths.values() if ffp.path.startswith(self.path)] + childrenfromffs = [ffp for ffp in childrenfromffs if not ffp.path.endswith(self.path)] + childrenfromffs = [ffp for ffp in childrenfromffs if ffp.exists()] + self.spawn = frozenset(childrenfromffs) + return self.spawn + + def parent(self): + if self.mockedfilepaths.has_key(self.antecedent): + parent = self.mockedfilepaths[self.antecedent] + else: + parent = MockFilePath(self.antecedent, self.mockedfilepaths) + return parent + + def parents(self): + antecedents = [] + def f(fps, antecedents): + newfps = os.path.split(fps)[0] + if newfps: + antecedents.append(newfps) + f(newfps, antecedents) + f(self.path, antecedents) + return antecedents + + def setparents(self): + for fps in self.parents(): + if not self.mockedfilepaths.has_key(fps): + self.mockedfilepaths[fps] = MockFilePath(fps, self.mockedfilepaths, exists=True) + + def basename(self): + return os.path.split(self.path)[1] + + def moveTo(self, newffp): + # XXX Makes no distinction between file and directory arguments, this is deviation from filepath.moveTo + if self.mockedfilepaths[newffp.path].exists(): + raise OSError + else: + self.mockedfilepaths[newffp.path] = self + self.path = newffp.path + + def getsize(self): + return self.fileobject.getsize() + + def exists(self): + return self.existence + + def isdir(self): + return True + + def makedirs(self): + # XXX These methods assume that fp_ functions in fileutil will be tested elsewhere! + pass + + def remove(self): + pass + + +class MockFileObject: + def __init__(self, contentstring=''): + self.buffer = contentstring + self.pos = 0 + def open(self, mode='r'): + return self + def write(self, instring): + begin = self.pos + padlen = begin - len(self.buffer) + if padlen > 0: + self.buffer += '\x00' * padlen + end = self.pos + len(instring) + self.buffer = self.buffer[:begin]+instring+self.buffer[end:] + self.pos = end + def close(self): + self.pos = 0 + def seek(self, pos): + self.pos = pos + def read(self, numberbytes): + return self.buffer[self.pos:self.pos+numberbytes] + def tell(self): + return self.pos + def size(self): + # XXX This method A: Is not to be found in a real file B: Is part of a wild-mung-up of filepath.stat! + # XXX Finally we shall hopefully use a getsize method soon, must consult first though. + # Hmmm... perhaps we need to sometimes stat the address when there's not a mockfileobject present? + return {stat.ST_SIZE:len(self.buffer)} + def getsize(self): + return len(self.buffer) + +class MockBCC: + def setServiceParent(self, Parent): + pass + + +class MockLCC: + def setServiceParent(self, Parent): + pass + + +class TestServerWithNullBackend(unittest.TestCase, ReallyEqualMixin): + """ NullBackend is just for testing and executable documentation, so + this test is actually a test of StorageServer in which we're using + NullBackend as helper code for the test, rather than a test of + NullBackend. """ + def setUp(self): + self.ss = StorageServer(testnodeid, NullBackend()) + + @mock.patch('os.mkdir') + @mock.patch('__builtin__.open') + @mock.patch('os.listdir') + @mock.patch('os.path.isdir') + def test_write_share(self, mockisdir, mocklistdir, mockopen, mockmkdir): + """ + Write a new share. This tests that StorageServer's remote_allocate_buckets + generates the correct return types when given test-vector arguments. That + bs is of the correct type is verified by attempting to invoke remote_write + on bs[0]. + """ + alreadygot, bs = self.ss.remote_allocate_buckets('teststorage_index', 'x'*32, 'y'*32, set((0,)), 1, mock.Mock()) + bs[0].remote_write(0, 'a') + self.failIf(mockisdir.called) + self.failIf(mocklistdir.called) + self.failIf(mockopen.called) + self.failIf(mockmkdir.called) + + +class TestServerConstruction(MockFileSystem, ReallyEqualMixin): + def test_create_server_disk_backend(self): + """ This tests whether a server instance can be constructed with a + filesystem backend. To pass the test, it mustn't use the filesystem + outside of its configured storedir. """ + StorageServer(testnodeid, DiskBackend(self.storedir)) + + +class TestServerAndDiskBackend(MockFileSystem, ReallyEqualMixin): + """ This tests both the StorageServer and the Disk backend together. """ + def setUp(self): + MockFileSystem.setUp(self) + try: + self.backend = DiskBackend(self.storedir) + self.ss = StorageServer(testnodeid, self.backend) + + self.backendwithreserve = DiskBackend(self.storedir, reserved_space = 1) + self.sswithreserve = StorageServer(testnodeid, self.backendwithreserve) + except: + MockFileSystem.tearDown(self) + raise + + @mock.patch('time.time') + @mock.patch('allmydata.util.fileutil.get_available_space') + def test_out_of_space(self, mockget_available_space, mocktime): + mocktime.return_value = 0 + + def call_get_available_space(dir, reserve): + return 0 + + mockget_available_space.side_effect = call_get_available_space + alreadygotc, bsc = self.sswithreserve.remote_allocate_buckets('teststorage_index', 'x'*32, 'y'*32, set((0,)), 1, mock.Mock()) + self.failUnlessReallyEqual(bsc, {}) + + @mock.patch('time.time') + def test_write_and_read_share(self, mocktime): + """ + Write a new share, read it, and test the server's (and disk backend's) + handling of simultaneous and successive attempts to write the same + share. + """ + mocktime.return_value = 0 + # Inspect incoming and fail unless it's empty. + incomingset = self.ss.backend.get_incoming_shnums('teststorage_index') + + self.failUnlessReallyEqual(incomingset, frozenset()) + + # Populate incoming with the sharenum: 0. + alreadygot, bs = self.ss.remote_allocate_buckets('teststorage_index', 'x'*32, 'y'*32, frozenset((0,)), 1, mock.Mock()) + + # This is a transparent-box test: Inspect incoming and fail unless the sharenum: 0 is listed there. + self.failUnlessReallyEqual(self.ss.backend.get_incoming_shnums('teststorage_index'), frozenset((0,))) + + + + # Attempt to create a second share writer with the same sharenum. + alreadygota, bsa = self.ss.remote_allocate_buckets('teststorage_index', 'x'*32, 'y'*32, frozenset((0,)), 1, mock.Mock()) + + # Show that no sharewriter results from a remote_allocate_buckets + # with the same si and sharenum, until BucketWriter.remote_close() + # has been called. + self.failIf(bsa) + + # Test allocated size. + spaceint = self.ss.allocated_size() + self.failUnlessReallyEqual(spaceint, 1) + + # Write 'a' to shnum 0. Only tested together with close and read. + bs[0].remote_write(0, 'a') + + # Preclose: Inspect final, failUnless nothing there. + self.failUnlessReallyEqual(len(list(self.backend.get_shares('teststorage_index'))), 0) + bs[0].remote_close() + + # Postclose: (Omnibus) failUnless written data is in final. + sharesinfinal = list(self.backend.get_shares('teststorage_index')) + self.failUnlessReallyEqual(len(sharesinfinal), 1) + contents = sharesinfinal[0].read_share_data(0, 73) + self.failUnlessReallyEqual(contents, client_data) + + # Exercise the case that the share we're asking to allocate is + # already (completely) uploaded. + self.ss.remote_allocate_buckets('teststorage_index', 'x'*32, 'y'*32, set((0,)), 1, mock.Mock()) + + + def test_read_old_share(self): + """ This tests whether the code correctly finds and reads + shares written out by old (Tahoe-LAFS <= v1.8.2) + servers. There is a similar test in test_download, but that one + is from the perspective of the client and exercises a deeper + stack of code. This one is for exercising just the + StorageServer object. """ + # Contruct a file with the appropriate contents in the mockfilesystem. + datalen = len(share_data) + finalhome = si_si2dir(self.basedir, 'teststorage_index').child(str(0)) + finalhome.setContent(share_data) + + # Now begin the test. + bs = self.ss.remote_get_buckets('teststorage_index') + + self.failUnlessEqual(len(bs), 1) + b = bs['0'] + # These should match by definition, the next two cases cover cases without (completely) unambiguous behaviors. + self.failUnlessReallyEqual(b.remote_read(0, datalen), client_data) + # If you try to read past the end you get the as much data as is there. + self.failUnlessReallyEqual(b.remote_read(0, datalen+20), client_data) + # If you start reading past the end of the file you get the empty string. + self.failUnlessReallyEqual(b.remote_read(datalen+1, 3), '') } [Pluggable backends -- all other changes. refs #999 david-sarah@jacaranda.org**20110919233256 Ignore-this: 1a77b6b5d178b32a9b914b699ba7e957 ] { hunk ./src/allmydata/client.py 245 sharetypes.append("immutable") if self.get_config("storage", "expire.mutable", True, boolean=True): sharetypes.append("mutable") - expiration_sharetypes = tuple(sharetypes) hunk ./src/allmydata/client.py 246 + expiration_policy = { + 'enabled': expire, + 'mode': mode, + 'override_lease_duration': o_l_d, + 'cutoff_date': cutoff_date, + 'sharetypes': tuple(sharetypes), + } ss = StorageServer(storedir, self.nodeid, reserved_space=reserved, discard_storage=discard, hunk ./src/allmydata/client.py 258 readonly_storage=readonly, stats_provider=self.stats_provider, - expiration_enabled=expire, - expiration_mode=mode, - expiration_override_lease_duration=o_l_d, - expiration_cutoff_date=cutoff_date, - expiration_sharetypes=expiration_sharetypes) + expiration_policy=expiration_policy) self.add_service(ss) d = self.when_tub_ready() hunk ./src/allmydata/immutable/offloaded.py 306 if os.path.exists(self._encoding_file): self.log("ciphertext already present, bypassing fetch", level=log.UNUSUAL) + # XXX the following comment is probably stale, since + # LocalCiphertextReader.get_plaintext_hashtree_leaves does not exist. + # # we'll still need the plaintext hashes (when # LocalCiphertextReader.get_plaintext_hashtree_leaves() is # called), and currently the easiest way to get them is to ask hunk ./src/allmydata/immutable/upload.py 765 self._status.set_progress(1, progress) return cryptdata - def get_plaintext_hashtree_leaves(self, first, last, num_segments): hunk ./src/allmydata/immutable/upload.py 766 + """OBSOLETE; Get the leaf nodes of a merkle hash tree over the + plaintext segments, i.e. get the tagged hashes of the given segments. + The segment size is expected to be generated by the + IEncryptedUploadable before any plaintext is read or ciphertext + produced, so that the segment hashes can be generated with only a + single pass. + + This returns a Deferred that fires with a sequence of hashes, using: + + tuple(segment_hashes[first:last]) + + 'num_segments' is used to assert that the number of segments that the + IEncryptedUploadable handled matches the number of segments that the + encoder was expecting. + + This method must not be called until the final byte has been read + from read_encrypted(). Once this method is called, read_encrypted() + can never be called again. + """ # this is currently unused, but will live again when we fix #453 if len(self._plaintext_segment_hashes) < num_segments: # close out the last one hunk ./src/allmydata/immutable/upload.py 803 return defer.succeed(tuple(self._plaintext_segment_hashes[first:last])) def get_plaintext_hash(self): + """OBSOLETE; Get the hash of the whole plaintext. + + This returns a Deferred that fires with a tagged SHA-256 hash of the + whole plaintext, obtained from hashutil.plaintext_hash(data). + """ + # this is currently unused, but will live again when we fix #453 h = self._plaintext_hasher.digest() return defer.succeed(h) hunk ./src/allmydata/interfaces.py 29 Number = IntegerConstraint(8) # 2**(8*8) == 16EiB ~= 18e18 ~= 18 exabytes Offset = Number ReadSize = int # the 'int' constraint is 2**31 == 2Gib -- large files are processed in not-so-large increments -WriteEnablerSecret = Hash # used to protect mutable bucket modifications -LeaseRenewSecret = Hash # used to protect bucket lease renewal requests -LeaseCancelSecret = Hash # used to protect bucket lease cancellation requests +WriteEnablerSecret = Hash # used to protect mutable share modifications +LeaseRenewSecret = Hash # used to protect lease renewal requests +LeaseCancelSecret = Hash # used to protect lease cancellation requests class RIStubClient(RemoteInterface): """Each client publishes a service announcement for a dummy object called hunk ./src/allmydata/interfaces.py 106 sharenums=SetOf(int, maxLength=MAX_BUCKETS), allocated_size=Offset, canary=Referenceable): """ - @param storage_index: the index of the bucket to be created or + @param storage_index: the index of the shareset to be created or increfed. @param sharenums: these are the share numbers (probably between 0 and 99) that the sender is proposing to store on this hunk ./src/allmydata/interfaces.py 111 server. - @param renew_secret: This is the secret used to protect bucket refresh + @param renew_secret: This is the secret used to protect lease renewal. This secret is generated by the client and stored for later comparison by the server. Each server is given a different secret. hunk ./src/allmydata/interfaces.py 115 - @param cancel_secret: Like renew_secret, but protects bucket decref. - @param canary: If the canary is lost before close(), the bucket is + @param cancel_secret: ignored + @param canary: If the canary is lost before close(), the allocation is deleted. @return: tuple of (alreadygot, allocated), where alreadygot is what we already have and allocated is what we hereby agree to accept. hunk ./src/allmydata/interfaces.py 129 renew_secret=LeaseRenewSecret, cancel_secret=LeaseCancelSecret): """ - Add a new lease on the given bucket. If the renew_secret matches an + Add a new lease on the given shareset. If the renew_secret matches an existing lease, that lease will be renewed instead. If there is no hunk ./src/allmydata/interfaces.py 131 - bucket for the given storage_index, return silently. (note that in + shareset for the given storage_index, return silently. (Note that in tahoe-1.3.0 and earlier, IndexError was raised if there was no hunk ./src/allmydata/interfaces.py 133 - bucket) + shareset.) """ return Any() # returns None now, but future versions might change hunk ./src/allmydata/interfaces.py 139 def renew_lease(storage_index=StorageIndex, renew_secret=LeaseRenewSecret): """ - Renew the lease on a given bucket, resetting the timer to 31 days. - Some networks will use this, some will not. If there is no bucket for + Renew the lease on a given shareset, resetting the timer to 31 days. + Some networks will use this, some will not. If there is no shareset for the given storage_index, IndexError will be raised. For mutable shares, if the given renew_secret does not match an hunk ./src/allmydata/interfaces.py 146 existing lease, IndexError will be raised with a note listing the server-nodeids on the existing leases, so leases on migrated shares - can be renewed or cancelled. For immutable shares, IndexError - (without the note) will be raised. + can be renewed. For immutable shares, IndexError (without the note) + will be raised. """ return Any() hunk ./src/allmydata/interfaces.py 154 def get_buckets(storage_index=StorageIndex): return DictOf(int, RIBucketReader, maxKeys=MAX_BUCKETS) - - def slot_readv(storage_index=StorageIndex, shares=ListOf(int), readv=ReadVector): """Read a vector from the numbered shares associated with the given hunk ./src/allmydata/interfaces.py 163 def slot_testv_and_readv_and_writev(storage_index=StorageIndex, secrets=TupleOf(WriteEnablerSecret, - LeaseRenewSecret, - LeaseCancelSecret), + LeaseRenewSecret), tw_vectors=TestAndWriteVectorsForShares, r_vector=ReadVector, ): hunk ./src/allmydata/interfaces.py 167 - """General-purpose test-and-set operation for mutable slots. Perform - a bunch of comparisons against the existing shares. If they all pass, - then apply a bunch of write vectors to those shares. Then use the - read vectors to extract data from all the shares and return the data. + """ + General-purpose atomic test-read-and-set operation for mutable slots. + Perform a bunch of comparisons against the existing shares. If they + all pass: use the read vectors to extract data from all the shares, + then apply a bunch of write vectors to those shares. Return the read + data, which does not include any modifications made by the writes. This method is, um, large. The goal is to allow clients to update all the shares associated with a mutable file in a single round trip. hunk ./src/allmydata/interfaces.py 177 - @param storage_index: the index of the bucket to be created or + @param storage_index: the index of the shareset to be created or increfed. @param write_enabler: a secret that is stored along with the slot. Writes are accepted from any caller who can hunk ./src/allmydata/interfaces.py 183 present the matching secret. A different secret should be used for each slot*server pair. - @param renew_secret: This is the secret used to protect bucket refresh + @param renew_secret: This is the secret used to protect lease renewal. This secret is generated by the client and stored for later comparison by the server. Each server is given a different secret. hunk ./src/allmydata/interfaces.py 187 - @param cancel_secret: Like renew_secret, but protects bucket decref. + @param cancel_secret: ignored hunk ./src/allmydata/interfaces.py 189 - The 'secrets' argument is a tuple of (write_enabler, renew_secret, - cancel_secret). The first is required to perform any write. The - latter two are used when allocating new shares. To simply acquire a - new lease on existing shares, use an empty testv and an empty writev. + The 'secrets' argument is a tuple with (write_enabler, renew_secret). + The write_enabler is required to perform any write. The renew_secret + is used when allocating new shares. Each share can have a separate test vector (i.e. a list of comparisons to perform). If all vectors for all shares pass, then all hunk ./src/allmydata/interfaces.py 280 store that on disk. """ -class IStorageBucketWriter(Interface): + +class IStorageBackend(Interface): """ hunk ./src/allmydata/interfaces.py 283 - Objects of this kind live on the client side. + Objects of this kind live on the server side and are used by the + storage server object. """ hunk ./src/allmydata/interfaces.py 286 - def put_block(segmentnum=int, data=ShareData): - """@param data: For most segments, this data will be 'blocksize' - bytes in length. The last segment might be shorter. - @return: a Deferred that fires (with None) when the operation completes + def get_available_space(): + """ + Returns available space for share storage in bytes, or + None if this information is not available or if the available + space is unlimited. + + If the backend is configured for read-only mode then this will + return 0. + """ + + def get_sharesets_for_prefix(prefix): + """ + Generates IShareSet objects for all storage indices matching the + given prefix for which this backend holds shares. + """ + + def get_shareset(storageindex): + """ + Get an IShareSet object for the given storage index. + """ + + def advise_corrupt_share(storageindex, sharetype, shnum, reason): + """ + Clients who discover hash failures in shares that they have + downloaded from me will use this method to inform me about the + failures. I will record their concern so that my operator can + manually inspect the shares in question. + + 'sharetype' is either 'mutable' or 'immutable'. 'shnum' is the integer + share number. 'reason' is a human-readable explanation of the problem, + probably including some expected hash values and the computed ones + that did not match. Corruption advisories for mutable shares should + include a hash of the public key (the same value that appears in the + mutable-file verify-cap), since the current share format does not + store that on disk. + + @param storageindex=str + @param sharetype=str + @param shnum=int + @param reason=str + """ + + +class IShareSet(Interface): + def get_storage_index(): + """ + Returns the storage index for this shareset. + """ + + def get_storage_index_string(): + """ + Returns the base32-encoded storage index for this shareset. + """ + + def get_overhead(): + """ + Returns the storage overhead, in bytes, of this shareset (exclusive + of the space used by its shares). + """ + + def get_shares(): + """ + Generates the IStoredShare objects held in this shareset. + """ + + def has_incoming(shnum): + """ + Returns True if this shareset has an incoming (partial) share with this number, otherwise False. + """ + + def make_bucket_writer(storageserver, shnum, max_space_per_bucket, lease_info, canary): + """ + Create a bucket writer that can be used to write data to a given share. + + @param storageserver=RIStorageServer + @param shnum=int: A share number in this shareset + @param max_space_per_bucket=int: The maximum space allocated for the + share, in bytes + @param lease_info=LeaseInfo: The initial lease information + @param canary=Referenceable: If the canary is lost before close(), the + bucket is deleted. + @return an IStorageBucketWriter for the given share + """ + + def make_bucket_reader(storageserver, share): + """ + Create a bucket reader that can be used to read data from a given share. + + @param storageserver=RIStorageServer + @param share=IStoredShare + @return an IStorageBucketReader for the given share + """ + + def readv(wanted_shnums, read_vector): + """ + Read a vector from the numbered shares in this shareset. An empty + wanted_shnums list means to return data from all known shares. + + @param wanted_shnums=ListOf(int) + @param read_vector=ReadVector + @return DictOf(int, ReadData): shnum -> results, with one key per share + """ + + def testv_and_readv_and_writev(storageserver, secrets, test_and_write_vectors, read_vector, expiration_time): + """ + General-purpose atomic test-read-and-set operation for mutable slots. + Perform a bunch of comparisons against the existing shares in this + shareset. If they all pass: use the read vectors to extract data from + all the shares, then apply a bunch of write vectors to those shares. + Return the read data, which does not include any modifications made by + the writes. + + See the similar method in RIStorageServer for more detail. + + @param storageserver=RIStorageServer + @param secrets=TupleOf(WriteEnablerSecret, LeaseRenewSecret[, ...]) + @param test_and_write_vectors=TestAndWriteVectorsForShares + @param read_vector=ReadVector + @param expiration_time=int + @return TupleOf(bool, DictOf(int, ReadData)) + """ + + def add_or_renew_lease(lease_info): + """ + Add a new lease on the shares in this shareset. If the renew_secret + matches an existing lease, that lease will be renewed instead. If + there are no shares in this shareset, return silently. + + @param lease_info=LeaseInfo + """ + + def renew_lease(renew_secret, new_expiration_time): + """ + Renew a lease on the shares in this shareset, resetting the timer + to 31 days. Some grids will use this, some will not. If there are no + shares in this shareset, IndexError will be raised. + + For mutable shares, if the given renew_secret does not match an + existing lease, IndexError will be raised with a note listing the + server-nodeids on the existing leases, so leases on migrated shares + can be renewed. For immutable shares, IndexError (without the note) + will be raised. + + @param renew_secret=LeaseRenewSecret + """ + + +class IStoredShare(Interface): + """ + This object contains as much as all of the share data. It is intended + for lazy evaluation, such that in many use cases substantially less than + all of the share data will be accessed. + """ + def close(): + """ + Complete writing to this share. + """ + + def get_storage_index(): + """ + Returns the storage index. + """ + + def get_shnum(): + """ + Returns the share number. + """ + + def get_data_length(): + """ + Returns the data length in bytes. + """ + + def get_size(): + """ + Returns the size of the share in bytes. + """ + + def get_used_space(): + """ + Returns the amount of backend storage including overhead, in bytes, used + by this share. + """ + + def unlink(): + """ + Signal that this share can be removed from the backend storage. This does + not guarantee that the share data will be immediately inaccessible, or + that it will be securely erased. + """ + + def readv(read_vector): + """ + XXX + """ + + +class IStoredMutableShare(IStoredShare): + def check_write_enabler(write_enabler, si_s): + """ + XXX """ hunk ./src/allmydata/interfaces.py 489 - def put_plaintext_hashes(hashes=ListOf(Hash)): + def check_testv(test_vector): + """ + XXX + """ + + def writev(datav, new_length): + """ + XXX + """ + + +class IStorageBucketWriter(Interface): + """ + Objects of this kind live on the client side. + """ + def put_block(segmentnum, data): """ hunk ./src/allmydata/interfaces.py 506 + @param segmentnum=int + @param data=ShareData: For most segments, this data will be 'blocksize' + bytes in length. The last segment might be shorter. @return: a Deferred that fires (with None) when the operation completes """ hunk ./src/allmydata/interfaces.py 512 - def put_crypttext_hashes(hashes=ListOf(Hash)): + def put_crypttext_hashes(hashes): """ hunk ./src/allmydata/interfaces.py 514 + @param hashes=ListOf(Hash) @return: a Deferred that fires (with None) when the operation completes """ hunk ./src/allmydata/interfaces.py 518 - def put_block_hashes(blockhashes=ListOf(Hash)): + def put_block_hashes(blockhashes): """ hunk ./src/allmydata/interfaces.py 520 + @param blockhashes=ListOf(Hash) @return: a Deferred that fires (with None) when the operation completes """ hunk ./src/allmydata/interfaces.py 524 - def put_share_hashes(sharehashes=ListOf(TupleOf(int, Hash))): + def put_share_hashes(sharehashes): """ hunk ./src/allmydata/interfaces.py 526 + @param sharehashes=ListOf(TupleOf(int, Hash)) @return: a Deferred that fires (with None) when the operation completes """ hunk ./src/allmydata/interfaces.py 530 - def put_uri_extension(data=URIExtensionData): + def put_uri_extension(data): """This block of data contains integrity-checking information (hashes of plaintext, crypttext, and shares), as well as encoding parameters that are necessary to recover the data. This is a serialized dict hunk ./src/allmydata/interfaces.py 535 mapping strings to other strings. The hash of this data is kept in - the URI and verified before any of the data is used. All buckets for - a given file contain identical copies of this data. + the URI and verified before any of the data is used. All share + containers for a given file contain identical copies of this data. The serialization format is specified with the following pseudocode: for k in sorted(dict.keys()): hunk ./src/allmydata/interfaces.py 543 assert re.match(r'^[a-zA-Z_\-]+$', k) write(k + ':' + netstring(dict[k])) + @param data=URIExtensionData @return: a Deferred that fires (with None) when the operation completes """ hunk ./src/allmydata/interfaces.py 558 class IStorageBucketReader(Interface): - def get_block_data(blocknum=int, blocksize=int, size=int): + def get_block_data(blocknum, blocksize, size): """Most blocks will be the same size. The last block might be shorter than the others. hunk ./src/allmydata/interfaces.py 562 + @param blocknum=int + @param blocksize=int + @param size=int @return: ShareData """ hunk ./src/allmydata/interfaces.py 573 @return: ListOf(Hash) """ - def get_block_hashes(at_least_these=SetOf(int)): + def get_block_hashes(at_least_these=()): """ hunk ./src/allmydata/interfaces.py 575 + @param at_least_these=SetOf(int) @return: ListOf(Hash) """ hunk ./src/allmydata/interfaces.py 579 - def get_share_hashes(at_least_these=SetOf(int)): + def get_share_hashes(): """ @return: ListOf(TupleOf(int, Hash)) """ hunk ./src/allmydata/interfaces.py 611 @return: unicode nickname, or None """ - # methods moved from IntroducerClient, need review - def get_all_connections(): - """Return a frozenset of (nodeid, service_name, rref) tuples, one for - each active connection we've established to a remote service. This is - mostly useful for unit tests that need to wait until a certain number - of connections have been made.""" - - def get_all_connectors(): - """Return a dict that maps from (nodeid, service_name) to a - RemoteServiceConnector instance for all services that we are actively - trying to connect to. Each RemoteServiceConnector has the following - public attributes:: - - service_name: the type of service provided, like 'storage' - announcement_time: when we first heard about this service - last_connect_time: when we last established a connection - last_loss_time: when we last lost a connection - - version: the peer's version, from the most recent connection - oldest_supported: the peer's oldest supported version, same - - rref: the RemoteReference, if connected, otherwise None - remote_host: the IAddress, if connected, otherwise None - - This method is intended for monitoring interfaces, such as a web page - that describes connecting and connected peers. - """ - - def get_all_peerids(): - """Return a frozenset of all peerids to whom we have a connection (to - one or more services) established. Mostly useful for unit tests.""" - - def get_all_connections_for(service_name): - """Return a frozenset of (nodeid, service_name, rref) tuples, one - for each active connection that provides the given SERVICE_NAME.""" - - def get_permuted_peers(service_name, key): - """Returns an ordered list of (peerid, rref) tuples, selecting from - the connections that provide SERVICE_NAME, using a hash-based - permutation keyed by KEY. This randomizes the service list in a - repeatable way, to distribute load over many peers. - """ - class IMutableSlotWriter(Interface): """ hunk ./src/allmydata/interfaces.py 616 The interface for a writer around a mutable slot on a remote server. """ - def set_checkstring(checkstring, *args): + def set_checkstring(seqnum_or_checkstring, root_hash=None, salt=None): """ Set the checkstring that I will pass to the remote server when writing. hunk ./src/allmydata/interfaces.py 640 Add a block and salt to the share. """ - def put_encprivey(encprivkey): + def put_encprivkey(encprivkey): """ Add the encrypted private key to the share. """ hunk ./src/allmydata/interfaces.py 645 - def put_blockhashes(blockhashes=list): + def put_blockhashes(blockhashes): """ hunk ./src/allmydata/interfaces.py 647 + @param blockhashes=list Add the block hash tree to the share. """ hunk ./src/allmydata/interfaces.py 651 - def put_sharehashes(sharehashes=dict): + def put_sharehashes(sharehashes): """ hunk ./src/allmydata/interfaces.py 653 + @param sharehashes=dict Add the share hash chain to the share. """ hunk ./src/allmydata/interfaces.py 739 def get_extension_params(): """Return the extension parameters in the URI""" - def set_extension_params(): + def set_extension_params(params): """Set the extension parameters that should be in the URI""" class IDirectoryURI(Interface): hunk ./src/allmydata/interfaces.py 879 writer-visible data using this writekey. """ - # TODO: Can this be overwrite instead of replace? - def replace(new_contents): - """Replace the contents of the mutable file, provided that no other + def overwrite(new_contents): + """Overwrite the contents of the mutable file, provided that no other node has published (or is attempting to publish, concurrently) a newer version of the file than this one. hunk ./src/allmydata/interfaces.py 1346 is empty, the metadata will be an empty dictionary. """ - def set_uri(name, writecap, readcap=None, metadata=None, overwrite=True): + def set_uri(name, writecap, readcap, metadata=None, overwrite=True): """I add a child (by writecap+readcap) at the specific name. I return a Deferred that fires when the operation finishes. If overwrite= is True, I will replace any existing child of the same name, otherwise hunk ./src/allmydata/interfaces.py 1745 Block Hash, and the encoding parameters, both of which must be included in the URI. - I do not choose shareholders, that is left to the IUploader. I must be - given a dict of RemoteReferences to storage buckets that are ready and - willing to receive data. + I do not choose shareholders, that is left to the IUploader. """ def set_size(size): hunk ./src/allmydata/interfaces.py 1752 """Specify the number of bytes that will be encoded. This must be peformed before get_serialized_params() can be called. """ + def set_params(params): """Override the default encoding parameters. 'params' is a tuple of (k,d,n), where 'k' is the number of required shares, 'd' is the hunk ./src/allmydata/interfaces.py 1848 download, validate, decode, and decrypt data from them, writing the results to an output file. - I do not locate the shareholders, that is left to the IDownloader. I must - be given a dict of RemoteReferences to storage buckets that are ready to - send data. + I do not locate the shareholders, that is left to the IDownloader. """ def setup(outfile): hunk ./src/allmydata/interfaces.py 1950 resuming an interrupted upload (where we need to compute the plaintext hashes, but don't need the redundant encrypted data).""" - def get_plaintext_hashtree_leaves(first, last, num_segments): - """OBSOLETE; Get the leaf nodes of a merkle hash tree over the - plaintext segments, i.e. get the tagged hashes of the given segments. - The segment size is expected to be generated by the - IEncryptedUploadable before any plaintext is read or ciphertext - produced, so that the segment hashes can be generated with only a - single pass. - - This returns a Deferred that fires with a sequence of hashes, using: - - tuple(segment_hashes[first:last]) - - 'num_segments' is used to assert that the number of segments that the - IEncryptedUploadable handled matches the number of segments that the - encoder was expecting. - - This method must not be called until the final byte has been read - from read_encrypted(). Once this method is called, read_encrypted() - can never be called again. - """ - - def get_plaintext_hash(): - """OBSOLETE; Get the hash of the whole plaintext. - - This returns a Deferred that fires with a tagged SHA-256 hash of the - whole plaintext, obtained from hashutil.plaintext_hash(data). - """ - def close(): """Just like IUploadable.close().""" hunk ./src/allmydata/interfaces.py 2144 returns a Deferred that fires with an IUploadResults instance, from which the URI of the file can be obtained as results.uri .""" - def upload_ssk(write_capability, new_version, uploadable): - """TODO: how should this work?""" - class ICheckable(Interface): def check(monitor, verify=False, add_lease=False): """Check up on my health, optionally repairing any problems. hunk ./src/allmydata/interfaces.py 2505 class IRepairResults(Interface): """I contain the results of a repair operation.""" - def get_successful(self): + def get_successful(): """Returns a boolean: True if the repair made the file healthy, False if not. Repair failure generally indicates a file that has been damaged beyond repair.""" hunk ./src/allmydata/interfaces.py 2577 Tahoe process will typically have a single NodeMaker, but unit tests may create simplified/mocked forms for testing purposes. """ - def create_from_cap(writecap, readcap=None, **kwargs): + def create_from_cap(writecap, readcap=None, deep_immutable=False, name=u""): """I create an IFilesystemNode from the given writecap/readcap. I can only provide nodes for existing file/directory objects: use my other methods to create new objects. I return synchronously.""" hunk ./src/allmydata/monitor.py 30 # the following methods are provided for the operation code - def is_cancelled(self): + def is_cancelled(): """Returns True if the operation has been cancelled. If True, operation code should stop creating new work, and attempt to stop any work already in progress.""" hunk ./src/allmydata/monitor.py 35 - def raise_if_cancelled(self): + def raise_if_cancelled(): """Raise OperationCancelledError if the operation has been cancelled. Operation code that has a robust error-handling path can simply call this periodically.""" hunk ./src/allmydata/monitor.py 40 - def set_status(self, status): + def set_status(status): """Sets the Monitor's 'status' object to an arbitrary value. Different operations will store different sorts of status information here. Operation code should use get+modify+set sequences to update hunk ./src/allmydata/monitor.py 46 this.""" - def get_status(self): + def get_status(): """Return the status object. If the operation failed, this will be a Failure instance.""" hunk ./src/allmydata/monitor.py 50 - def finish(self, status): + def finish(status): """Call this when the operation is done, successful or not. The Monitor's lifetime is influenced by the completion of the operation it is monitoring. The Monitor's 'status' value will be set with the hunk ./src/allmydata/monitor.py 63 # the following methods are provided for the initiator of the operation - def is_finished(self): + def is_finished(): """Return a boolean, True if the operation is done (whether successful or failed), False if it is still running.""" hunk ./src/allmydata/monitor.py 67 - def when_done(self): + def when_done(): """Return a Deferred that fires when the operation is complete. It will fire with the operation status, the same value as returned by get_status().""" hunk ./src/allmydata/monitor.py 72 - def cancel(self): + def cancel(): """Cancel the operation as soon as possible. is_cancelled() will start returning True after this is called.""" hunk ./src/allmydata/mutable/filenode.py 753 self._writekey = writekey self._serializer = defer.succeed(None) - def get_sequence_number(self): """ Get the sequence number of the mutable version that I represent. hunk ./src/allmydata/mutable/filenode.py 759 """ return self._version[0] # verinfo[0] == the sequence number + def get_servermap(self): + return self._servermap hunk ./src/allmydata/mutable/filenode.py 762 - # TODO: Terminology? def get_writekey(self): """ I return a writekey or None if I don't have a writekey. hunk ./src/allmydata/mutable/filenode.py 768 """ return self._writekey - def set_downloader_hints(self, hints): """ I set the downloader hints. hunk ./src/allmydata/mutable/filenode.py 776 self._downloader_hints = hints - def get_downloader_hints(self): """ I return the downloader hints. hunk ./src/allmydata/mutable/filenode.py 782 """ return self._downloader_hints - def overwrite(self, new_contents): """ I overwrite the contents of this mutable file version with the hunk ./src/allmydata/mutable/filenode.py 791 return self._do_serialized(self._overwrite, new_contents) - def _overwrite(self, new_contents): assert IMutableUploadable.providedBy(new_contents) assert self._servermap.last_update_mode == MODE_WRITE hunk ./src/allmydata/mutable/filenode.py 797 return self._upload(new_contents) - def modify(self, modifier, backoffer=None): """I use a modifier callback to apply a change to the mutable file. I implement the following pseudocode:: hunk ./src/allmydata/mutable/filenode.py 841 return self._do_serialized(self._modify, modifier, backoffer) - def _modify(self, modifier, backoffer): if backoffer is None: backoffer = BackoffAgent().delay hunk ./src/allmydata/mutable/filenode.py 846 return self._modify_and_retry(modifier, backoffer, True) - def _modify_and_retry(self, modifier, backoffer, first_time): """ I try to apply modifier to the contents of this version of the hunk ./src/allmydata/mutable/filenode.py 878 d.addErrback(_retry) return d - def _modify_once(self, modifier, first_time): """ I attempt to apply a modifier to the contents of the mutable hunk ./src/allmydata/mutable/filenode.py 913 d.addCallback(_apply) return d - def is_readonly(self): """ I return True if this MutableFileVersion provides no write hunk ./src/allmydata/mutable/filenode.py 921 """ return self._writekey is None - def is_mutable(self): """ I return True, since mutable files are always mutable by hunk ./src/allmydata/mutable/filenode.py 928 """ return True - def get_storage_index(self): """ I return the storage index of the reference that I encapsulate. hunk ./src/allmydata/mutable/filenode.py 934 """ return self._storage_index - def get_size(self): """ I return the length, in bytes, of this readable object. hunk ./src/allmydata/mutable/filenode.py 940 """ return self._servermap.size_of_version(self._version) - def download_to_data(self, fetch_privkey=False): """ I return a Deferred that fires with the contents of this hunk ./src/allmydata/mutable/filenode.py 951 d.addCallback(lambda mc: "".join(mc.chunks)) return d - def _try_to_download_data(self): """ I am an unserialized cousin of download_to_data; I am called hunk ./src/allmydata/mutable/filenode.py 963 d.addCallback(lambda mc: "".join(mc.chunks)) return d - def read(self, consumer, offset=0, size=None, fetch_privkey=False): """ I read a portion (possibly all) of the mutable file that I hunk ./src/allmydata/mutable/filenode.py 971 return self._do_serialized(self._read, consumer, offset, size, fetch_privkey) - def _read(self, consumer, offset=0, size=None, fetch_privkey=False): """ I am the serialized companion of read. hunk ./src/allmydata/mutable/filenode.py 981 d = r.download(consumer, offset, size) return d - def _do_serialized(self, cb, *args, **kwargs): # note: to avoid deadlock, this callable is *not* allowed to invoke # other serialized methods within this (or any other) hunk ./src/allmydata/mutable/filenode.py 999 self._serializer.addErrback(log.err) return d - def _upload(self, new_contents): #assert self._pubkey, "update_servermap must be called before publish" p = Publish(self._node, self._storage_broker, self._servermap) hunk ./src/allmydata/mutable/filenode.py 1009 d.addCallback(self._did_upload, new_contents.get_size()) return d - def _did_upload(self, res, size): self._most_recent_size = size return res hunk ./src/allmydata/mutable/filenode.py 1029 """ return self._do_serialized(self._update, data, offset) - def _update(self, data, offset): """ I update the mutable file version represented by this particular hunk ./src/allmydata/mutable/filenode.py 1058 d.addCallback(self._build_uploadable_and_finish, data, offset) return d - def _do_modify_update(self, data, offset): """ I perform a file update by modifying the contents of the file hunk ./src/allmydata/mutable/filenode.py 1073 return new return self._modify(m, None) - def _do_update_update(self, data, offset): """ I start the Servermap update that gets us the data we need to hunk ./src/allmydata/mutable/filenode.py 1108 return self._update_servermap(update_range=(start_segment, end_segment)) - def _decode_and_decrypt_segments(self, ignored, data, offset): """ After the servermap update, I take the encrypted and encoded hunk ./src/allmydata/mutable/filenode.py 1148 d3 = defer.succeed(blockhashes) return deferredutil.gatherResults([d1, d2, d3]) - def _build_uploadable_and_finish(self, segments_and_bht, data, offset): """ After the process has the plaintext segments, I build the hunk ./src/allmydata/mutable/filenode.py 1163 p = Publish(self._node, self._storage_broker, self._servermap) return p.update(u, offset, segments_and_bht[2], self._version) - def _update_servermap(self, mode=MODE_WRITE, update_range=None): """ I update the servermap. I return a Deferred that fires when the hunk ./src/allmydata/storage/common.py 1 - -import os.path from allmydata.util import base32 class DataTooLargeError(Exception): hunk ./src/allmydata/storage/common.py 5 pass + class UnknownMutableContainerVersionError(Exception): pass hunk ./src/allmydata/storage/common.py 8 + class UnknownImmutableContainerVersionError(Exception): pass hunk ./src/allmydata/storage/common.py 18 def si_a2b(ascii_storageindex): return base32.a2b(ascii_storageindex) - -def storage_index_to_dir(storageindex): - sia = si_b2a(storageindex) - return os.path.join(sia[:2], sia) hunk ./src/allmydata/storage/crawler.py 2 -import os, time, struct +import time, struct import cPickle as pickle from twisted.internet import reactor from twisted.application import service hunk ./src/allmydata/storage/crawler.py 6 + +from allmydata.util.assertutil import precondition +from allmydata.interfaces import IStorageBackend from allmydata.storage.common import si_b2a hunk ./src/allmydata/storage/crawler.py 10 -from allmydata.util import fileutil + class TimeSliceExceeded(Exception): pass hunk ./src/allmydata/storage/crawler.py 15 + class ShareCrawler(service.MultiService): hunk ./src/allmydata/storage/crawler.py 17 - """A ShareCrawler subclass is attached to a StorageServer, and - periodically walks all of its shares, processing each one in some - fashion. This crawl is rate-limited, to reduce the IO burden on the host, - since large servers can easily have a terabyte of shares, in several - million files, which can take hours or days to read. + """ + An instance of a subclass of ShareCrawler is attached to a storage + backend, and periodically walks the backend's shares, processing them + in some fashion. This crawl is rate-limited to reduce the I/O burden on + the host, since large servers can easily have a terabyte of shares in + several million files, which can take hours or days to read. Once the crawler starts a cycle, it will proceed at a rate limited by the allowed_cpu_percentage= and cpu_slice= parameters: yielding the reactor hunk ./src/allmydata/storage/crawler.py 33 long enough to ensure that 'minimum_cycle_time' elapses between the start of two consecutive cycles. - We assume that the normal upload/download/get_buckets traffic of a tahoe + We assume that the normal upload/download/DYHB traffic of a Tahoe-LAFS grid will cause the prefixdir contents to be mostly cached in the kernel, hunk ./src/allmydata/storage/crawler.py 35 - or that the number of buckets in each prefixdir will be small enough to - load quickly. A 1TB allmydata.com server was measured to have 2.56M - buckets, spread into the 1024 prefixdirs, with about 2500 buckets per + or that the number of sharesets in each prefixdir will be small enough to + load quickly. A 1TB allmydata.com server was measured to have 2.56 million + sharesets, spread into the 1024 prefixdirs, with about 2500 sharesets per prefix. On this server, each prefixdir took 130ms-200ms to list the first time, and 17ms to list the second time. hunk ./src/allmydata/storage/crawler.py 41 - To use a crawler, create a subclass which implements the process_bucket() - method. It will be called with a prefixdir and a base32 storage index - string. process_bucket() must run synchronously. Any keys added to - self.state will be preserved. Override add_initial_state() to set up - initial state keys. Override finished_cycle() to perform additional - processing when the cycle is complete. Any status that the crawler - produces should be put in the self.state dictionary. Status renderers - (like a web page which describes the accomplishments of your crawler) - will use crawler.get_state() to retrieve this dictionary; they can - present the contents as they see fit. + To implement a crawler, create a subclass that implements the + process_shareset() method. It will be called with a prefixdir and an + object providing the IShareSet interface. process_shareset() must run + synchronously. Any keys added to self.state will be preserved. Override + add_initial_state() to set up initial state keys. Override + finished_cycle() to perform additional processing when the cycle is + complete. Any status that the crawler produces should be put in the + self.state dictionary. Status renderers (like a web page describing the + accomplishments of your crawler) will use crawler.get_state() to retrieve + this dictionary; they can present the contents as they see fit. hunk ./src/allmydata/storage/crawler.py 52 - Then create an instance, with a reference to a StorageServer and a - filename where it can store persistent state. The statefile is used to - keep track of how far around the ring the process has travelled, as well - as timing history to allow the pace to be predicted and controlled. The - statefile will be updated and written to disk after each time slice (just - before the crawler yields to the reactor), and also after each cycle is - finished, and also when stopService() is called. Note that this means - that a crawler which is interrupted with SIGKILL while it is in the - middle of a time slice will lose progress: the next time the node is - started, the crawler will repeat some unknown amount of work. + Then create an instance, with a reference to a backend object providing + the IStorageBackend interface, and a filename where it can store + persistent state. The statefile is used to keep track of how far around + the ring the process has travelled, as well as timing history to allow + the pace to be predicted and controlled. The statefile will be updated + and written to disk after each time slice (just before the crawler yields + to the reactor), and also after each cycle is finished, and also when + stopService() is called. Note that this means that a crawler that is + interrupted with SIGKILL while it is in the middle of a time slice will + lose progress: the next time the node is started, the crawler will repeat + some unknown amount of work. The crawler instance must be started with startService() before it will hunk ./src/allmydata/storage/crawler.py 65 - do any work. To make it stop doing work, call stopService(). + do any work. To make it stop doing work, call stopService(). A crawler + is usually a child service of a StorageServer, although it should not + depend on that. + + For historical reasons, some dictionary key names use the term "bucket" + for what is now preferably called a "shareset" (the set of shares that a + server holds under a given storage index). """ slow_start = 300 # don't start crawling for 5 minutes after startup hunk ./src/allmydata/storage/crawler.py 80 cpu_slice = 1.0 # use up to 1.0 seconds before yielding minimum_cycle_time = 300 # don't run a cycle faster than this - def __init__(self, server, statefile, allowed_cpu_percentage=None): + def __init__(self, backend, statefp, allowed_cpu_percentage=None): + precondition(IStorageBackend.providedBy(backend), backend) service.MultiService.__init__(self) hunk ./src/allmydata/storage/crawler.py 83 + self.backend = backend + self.statefp = statefp if allowed_cpu_percentage is not None: self.allowed_cpu_percentage = allowed_cpu_percentage hunk ./src/allmydata/storage/crawler.py 87 - self.server = server - self.sharedir = server.sharedir - self.statefile = statefile self.prefixes = [si_b2a(struct.pack(">H", i << (16-10)))[:2] for i in range(2**10)] self.prefixes.sort() hunk ./src/allmydata/storage/crawler.py 91 self.timer = None - self.bucket_cache = (None, []) + self.shareset_cache = (None, []) self.current_sleep_time = None self.next_wake_time = None self.last_prefix_finished_time = None hunk ./src/allmydata/storage/crawler.py 154 left = len(self.prefixes) - self.last_complete_prefix_index remaining = left * self.last_prefix_elapsed_time # TODO: remainder of this prefix: we need to estimate the - # per-bucket time, probably by measuring the time spent on - # this prefix so far, divided by the number of buckets we've + # per-shareset time, probably by measuring the time spent on + # this prefix so far, divided by the number of sharesets we've # processed. d["estimated-cycle-complete-time-left"] = remaining # it's possible to call get_progress() from inside a crawler's hunk ./src/allmydata/storage/crawler.py 175 state dictionary. If we are not currently sleeping (i.e. get_state() was called from - inside the process_prefixdir, process_bucket, or finished_cycle() + inside the process_prefixdir, process_shareset, or finished_cycle() methods, or if startService has not yet been called on this crawler), these two keys will be None. hunk ./src/allmydata/storage/crawler.py 188 def load_state(self): # we use this to store state for both the crawler's internals and # anything the subclass-specific code needs. The state is stored - # after each bucket is processed, after each prefixdir is processed, + # after each shareset is processed, after each prefixdir is processed, # and after a cycle is complete. The internal keys we use are: # ["version"]: int, always 1 # ["last-cycle-finished"]: int, or None if we have not yet finished hunk ./src/allmydata/storage/crawler.py 202 # are sleeping between cycles, or if we # have not yet finished any prefixdir since # a cycle was started - # ["last-complete-bucket"]: str, base32 storage index bucket name - # of the last bucket to be processed, or - # None if we are sleeping between cycles + # ["last-complete-bucket"]: str, base32 storage index of the last + # shareset to be processed, or None if we + # are sleeping between cycles try: hunk ./src/allmydata/storage/crawler.py 206 - f = open(self.statefile, "rb") - state = pickle.load(f) - f.close() + state = pickle.loads(self.statefp.getContent()) except EnvironmentError: state = {"version": 1, "last-cycle-finished": None, hunk ./src/allmydata/storage/crawler.py 242 else: last_complete_prefix = self.prefixes[lcpi] self.state["last-complete-prefix"] = last_complete_prefix - tmpfile = self.statefile + ".tmp" - f = open(tmpfile, "wb") - pickle.dump(self.state, f) - f.close() - fileutil.move_into_place(tmpfile, self.statefile) + self.statefp.setContent(pickle.dumps(self.state)) def startService(self): # arrange things to look like we were just sleeping, so hunk ./src/allmydata/storage/crawler.py 284 sleep_time = (this_slice / self.allowed_cpu_percentage) - this_slice # if the math gets weird, or a timequake happens, don't sleep # forever. Note that this means that, while a cycle is running, we - # will process at least one bucket every 5 minutes, no matter how - # long that bucket takes. + # will process at least one shareset every 5 minutes, no matter how + # long that shareset takes. sleep_time = max(0.0, min(sleep_time, 299)) if finished_cycle: # how long should we sleep between cycles? Don't run faster than hunk ./src/allmydata/storage/crawler.py 315 for i in range(self.last_complete_prefix_index+1, len(self.prefixes)): # if we want to yield earlier, just raise TimeSliceExceeded() prefix = self.prefixes[i] - prefixdir = os.path.join(self.sharedir, prefix) - if i == self.bucket_cache[0]: - buckets = self.bucket_cache[1] + if i == self.shareset_cache[0]: + sharesets = self.shareset_cache[1] else: hunk ./src/allmydata/storage/crawler.py 318 - try: - buckets = os.listdir(prefixdir) - buckets.sort() - except EnvironmentError: - buckets = [] - self.bucket_cache = (i, buckets) - self.process_prefixdir(cycle, prefix, prefixdir, - buckets, start_slice) + sharesets = self.backend.get_sharesets_for_prefix(prefix) + self.shareset_cache = (i, sharesets) + self.process_prefixdir(cycle, prefix, sharesets, start_slice) self.last_complete_prefix_index = i now = time.time() hunk ./src/allmydata/storage/crawler.py 345 self.finished_cycle(cycle) self.save_state() - def process_prefixdir(self, cycle, prefix, prefixdir, buckets, start_slice): - """This gets a list of bucket names (i.e. storage index strings, + def process_prefixdir(self, cycle, prefix, sharesets, start_slice): + """ + This gets a list of shareset names (i.e. storage index strings, base32-encoded) in sorted order. You can override this if your crawler doesn't care about the actual hunk ./src/allmydata/storage/crawler.py 352 shares, for example a crawler which merely keeps track of how many - buckets are being managed by this server. + sharesets are being managed by this server. hunk ./src/allmydata/storage/crawler.py 354 - Subclasses which *do* care about actual bucket should leave this - method along, and implement process_bucket() instead. + Subclasses which *do* care about actual shareset should leave this + method alone, and implement process_shareset() instead. """ hunk ./src/allmydata/storage/crawler.py 358 - for bucket in buckets: - if bucket <= self.state["last-complete-bucket"]: + for shareset in sharesets: + base32si = shareset.get_storage_index_string() + if base32si <= self.state["last-complete-bucket"]: continue hunk ./src/allmydata/storage/crawler.py 362 - self.process_bucket(cycle, prefix, prefixdir, bucket) - self.state["last-complete-bucket"] = bucket + self.process_shareset(cycle, prefix, shareset) + self.state["last-complete-bucket"] = base32si if time.time() >= start_slice + self.cpu_slice: raise TimeSliceExceeded() hunk ./src/allmydata/storage/crawler.py 370 # the remaining methods are explictly for subclasses to implement. def started_cycle(self, cycle): - """Notify a subclass that the crawler is about to start a cycle. + """ + Notify a subclass that the crawler is about to start a cycle. This method is for subclasses to override. No upcall is necessary. """ hunk ./src/allmydata/storage/crawler.py 377 pass - def process_bucket(self, cycle, prefix, prefixdir, storage_index_b32): - """Examine a single bucket. Subclasses should do whatever they want + def process_shareset(self, cycle, prefix, shareset): + """ + Examine a single shareset. Subclasses should do whatever they want to do to the shares therein, then update self.state as necessary. If the crawler is never interrupted by SIGKILL, this method will be hunk ./src/allmydata/storage/crawler.py 383 - called exactly once per share (per cycle). If it *is* interrupted, + called exactly once per shareset (per cycle). If it *is* interrupted, then the next time the node is started, some amount of work will be duplicated, according to when self.save_state() was last called. By default, save_state() is called at the end of each timeslice, and hunk ./src/allmydata/storage/crawler.py 391 To reduce the chance of duplicate work (i.e. to avoid adding multiple records to a database), you can call save_state() at the end of your - process_bucket() method. This will reduce the maximum duplicated work - to one bucket per SIGKILL. It will also add overhead, probably 1-20ms - per bucket (and some disk writes), which will count against your - allowed_cpu_percentage, and which may be considerable if - process_bucket() runs quickly. + process_shareset() method. This will reduce the maximum duplicated + work to one shareset per SIGKILL. It will also add overhead, probably + 1-20ms per shareset (and some disk writes), which will count against + your allowed_cpu_percentage, and which may be considerable if + process_shareset() runs quickly. This method is for subclasses to override. No upcall is necessary. """ hunk ./src/allmydata/storage/crawler.py 402 pass def finished_prefix(self, cycle, prefix): - """Notify a subclass that the crawler has just finished processing a - prefix directory (all buckets with the same two-character/10bit + """ + Notify a subclass that the crawler has just finished processing a + prefix directory (all sharesets with the same two-character/10-bit prefix). To impose a limit on how much work might be duplicated by a SIGKILL that occurs during a timeslice, you can call self.save_state() here, but be aware that it may represent a hunk ./src/allmydata/storage/crawler.py 415 pass def finished_cycle(self, cycle): - """Notify subclass that a cycle (one complete traversal of all + """ + Notify subclass that a cycle (one complete traversal of all prefixdirs) has just finished. 'cycle' is the number of the cycle that just finished. This method should perform summary work and update self.state to publish information to status displays. hunk ./src/allmydata/storage/crawler.py 433 pass def yielding(self, sleep_time): - """The crawler is about to sleep for 'sleep_time' seconds. This + """ + The crawler is about to sleep for 'sleep_time' seconds. This method is mostly for the convenience of unit tests. This method is for subclasses to override. No upcall is necessary. hunk ./src/allmydata/storage/crawler.py 443 class BucketCountingCrawler(ShareCrawler): - """I keep track of how many buckets are being managed by this server. - This is equivalent to the number of distributed files and directories for - which I am providing storage. The actual number of files+directories in - the full grid is probably higher (especially when there are more servers - than 'N', the number of generated shares), because some files+directories - will have shares on other servers instead of me. Also note that the - number of buckets will differ from the number of shares in small grids, - when more than one share is placed on a single server. + """ + I keep track of how many sharesets, each corresponding to a storage index, + are being managed by this server. This is equivalent to the number of + distributed files and directories for which I am providing storage. The + actual number of files and directories in the full grid is probably higher + (especially when there are more servers than 'N', the number of generated + shares), because some files and directories will have shares on other + servers instead of me. Also note that the number of sharesets will differ + from the number of shares in small grids, when more than one share is + placed on a single server. """ minimum_cycle_time = 60*60 # we don't need this more than once an hour hunk ./src/allmydata/storage/crawler.py 457 - def __init__(self, server, statefile, num_sample_prefixes=1): - ShareCrawler.__init__(self, server, statefile) + def __init__(self, backend, statefp, num_sample_prefixes=1): + ShareCrawler.__init__(self, backend, statefp) self.num_sample_prefixes = num_sample_prefixes def add_initial_state(self): hunk ./src/allmydata/storage/crawler.py 471 self.state.setdefault("last-complete-bucket-count", None) self.state.setdefault("storage-index-samples", {}) - def process_prefixdir(self, cycle, prefix, prefixdir, buckets, start_slice): + def process_prefixdir(self, cycle, prefix, sharesets, start_slice): # we override process_prefixdir() because we don't want to look at hunk ./src/allmydata/storage/crawler.py 473 - # the individual buckets. We'll save state after each one. On my + # the individual sharesets. We'll save state after each one. On my # laptop, a mostly-empty storage server can process about 70 # prefixdirs in a 1.0s slice. if cycle not in self.state["bucket-counts"]: hunk ./src/allmydata/storage/crawler.py 478 self.state["bucket-counts"][cycle] = {} - self.state["bucket-counts"][cycle][prefix] = len(buckets) + self.state["bucket-counts"][cycle][prefix] = len(sharesets) if prefix in self.prefixes[:self.num_sample_prefixes]: hunk ./src/allmydata/storage/crawler.py 480 - self.state["storage-index-samples"][prefix] = (cycle, buckets) + self.state["storage-index-samples"][prefix] = (cycle, sharesets) def finished_cycle(self, cycle): last_counts = self.state["bucket-counts"].get(cycle, []) hunk ./src/allmydata/storage/crawler.py 486 if len(last_counts) == len(self.prefixes): # great, we have a whole cycle. - num_buckets = sum(last_counts.values()) - self.state["last-complete-bucket-count"] = num_buckets + num_sharesets = sum(last_counts.values()) + self.state["last-complete-bucket-count"] = num_sharesets # get rid of old counts for old_cycle in list(self.state["bucket-counts"].keys()): if old_cycle != cycle: hunk ./src/allmydata/storage/crawler.py 494 del self.state["bucket-counts"][old_cycle] # get rid of old samples too for prefix in list(self.state["storage-index-samples"].keys()): - old_cycle,buckets = self.state["storage-index-samples"][prefix] + old_cycle, storage_indices = self.state["storage-index-samples"][prefix] if old_cycle != cycle: del self.state["storage-index-samples"][prefix] hunk ./src/allmydata/storage/crawler.py 497 - hunk ./src/allmydata/storage/expirer.py 1 -import time, os, pickle, struct + +import time, pickle, struct +from twisted.python import log as twlog + from allmydata.storage.crawler import ShareCrawler hunk ./src/allmydata/storage/expirer.py 6 -from allmydata.storage.shares import get_share_file -from allmydata.storage.common import UnknownMutableContainerVersionError, \ +from allmydata.storage.common import si_b2a, UnknownMutableContainerVersionError, \ UnknownImmutableContainerVersionError hunk ./src/allmydata/storage/expirer.py 8 -from twisted.python import log as twlog + class LeaseCheckingCrawler(ShareCrawler): """I examine the leases on all shares, determining which are still valid hunk ./src/allmydata/storage/expirer.py 17 removed. I collect statistics on the leases and make these available to a web - status page, including:: + status page, including: Space recovered during this cycle-so-far: actual (only if expiration_enabled=True): hunk ./src/allmydata/storage/expirer.py 21 - num-buckets, num-shares, sum of share sizes, real disk usage + num-storage-indices, num-shares, sum of share sizes, real disk usage ('real disk usage' means we use stat(fn).st_blocks*512 and include any space used by the directory) what it would have been with the original lease expiration time hunk ./src/allmydata/storage/expirer.py 32 Space recovered during the last 10 cycles <-- saved in separate pickle - Shares/buckets examined: + Shares/storage-indices examined: this cycle-so-far prediction of rest of cycle during last 10 cycles <-- separate pickle hunk ./src/allmydata/storage/expirer.py 42 Histogram of leases-per-share: this-cycle-to-date last 10 cycles <-- separate pickle - Histogram of lease ages, buckets = 1day + Histogram of lease ages, storage-indices over 1 day cycle-to-date last 10 cycles <-- separate pickle hunk ./src/allmydata/storage/expirer.py 53 slow_start = 360 # wait 6 minutes after startup minimum_cycle_time = 12*60*60 # not more than twice per day - def __init__(self, server, statefile, historyfile, - expiration_enabled, mode, - override_lease_duration, # used if expiration_mode=="age" - cutoff_date, # used if expiration_mode=="cutoff-date" - sharetypes): - self.historyfile = historyfile - self.expiration_enabled = expiration_enabled - self.mode = mode + def __init__(self, backend, statefp, historyfp, expiration_policy): + # ShareCrawler.__init__ will call add_initial_state, so self.historyfp has to be set first. + self.historyfp = historyfp + ShareCrawler.__init__(self, backend, statefp) + + self.expiration_enabled = expiration_policy['enabled'] + self.mode = expiration_policy['mode'] self.override_lease_duration = None self.cutoff_date = None if self.mode == "age": hunk ./src/allmydata/storage/expirer.py 63 - assert isinstance(override_lease_duration, (int, type(None))) - self.override_lease_duration = override_lease_duration # seconds + assert isinstance(expiration_policy['override_lease_duration'], (int, type(None))) + self.override_lease_duration = expiration_policy['override_lease_duration'] # seconds elif self.mode == "cutoff-date": hunk ./src/allmydata/storage/expirer.py 66 - assert isinstance(cutoff_date, int) # seconds-since-epoch - assert cutoff_date is not None - self.cutoff_date = cutoff_date + assert isinstance(expiration_policy['cutoff_date'], int) # seconds-since-epoch + self.cutoff_date = expiration_policy['cutoff_date'] else: hunk ./src/allmydata/storage/expirer.py 69 - raise ValueError("GC mode '%s' must be 'age' or 'cutoff-date'" % mode) - self.sharetypes_to_expire = sharetypes - ShareCrawler.__init__(self, server, statefile) + raise ValueError("GC mode '%s' must be 'age' or 'cutoff-date'" % expiration_policy['mode']) + self.sharetypes_to_expire = expiration_policy['sharetypes'] def add_initial_state(self): # we fill ["cycle-to-date"] here (even though they will be reset in hunk ./src/allmydata/storage/expirer.py 84 self.state["cycle-to-date"].setdefault(k, so_far[k]) # initialize history - if not os.path.exists(self.historyfile): + if not self.historyfp.exists(): history = {} # cyclenum -> dict hunk ./src/allmydata/storage/expirer.py 86 - f = open(self.historyfile, "wb") - pickle.dump(history, f) - f.close() + self.historyfp.setContent(pickle.dumps(history)) def create_empty_cycle_dict(self): recovered = self.create_empty_recovered_dict() hunk ./src/allmydata/storage/expirer.py 99 def create_empty_recovered_dict(self): recovered = {} + # "buckets" is ambiguous; here it means the number of sharesets (one per storage index per server) for a in ("actual", "original", "configured", "examined"): for b in ("buckets", "shares", "sharebytes", "diskbytes"): recovered[a+"-"+b] = 0 hunk ./src/allmydata/storage/expirer.py 110 def started_cycle(self, cycle): self.state["cycle-to-date"] = self.create_empty_cycle_dict() - def stat(self, fn): - return os.stat(fn) - - def process_bucket(self, cycle, prefix, prefixdir, storage_index_b32): - bucketdir = os.path.join(prefixdir, storage_index_b32) - s = self.stat(bucketdir) + def process_storage_index(self, cycle, prefix, container): would_keep_shares = [] wks = None hunk ./src/allmydata/storage/expirer.py 113 + sharetype = None hunk ./src/allmydata/storage/expirer.py 115 - for fn in os.listdir(bucketdir): - try: - shnum = int(fn) - except ValueError: - continue # non-numeric means not a sharefile - sharefile = os.path.join(bucketdir, fn) + for share in container.get_shares(): + sharetype = share.sharetype try: hunk ./src/allmydata/storage/expirer.py 118 - wks = self.process_share(sharefile) + wks = self.process_share(share) except (UnknownMutableContainerVersionError, UnknownImmutableContainerVersionError, struct.error): hunk ./src/allmydata/storage/expirer.py 122 - twlog.msg("lease-checker error processing %s" % sharefile) + twlog.msg("lease-checker error processing %r" % (share,)) twlog.err() hunk ./src/allmydata/storage/expirer.py 124 - which = (storage_index_b32, shnum) + which = (si_b2a(share.storageindex), share.get_shnum()) self.state["cycle-to-date"]["corrupt-shares"].append(which) wks = (1, 1, 1, "unknown") would_keep_shares.append(wks) hunk ./src/allmydata/storage/expirer.py 129 - sharetype = None + container_type = None if wks: hunk ./src/allmydata/storage/expirer.py 131 - # use the last share's sharetype as the buckettype - sharetype = wks[3] + # use the last share's sharetype as the container type + container_type = wks[3] rec = self.state["cycle-to-date"]["space-recovered"] self.increment(rec, "examined-buckets", 1) if sharetype: hunk ./src/allmydata/storage/expirer.py 136 - self.increment(rec, "examined-buckets-"+sharetype, 1) + self.increment(rec, "examined-buckets-"+container_type, 1) + + container_diskbytes = container.get_overhead() hunk ./src/allmydata/storage/expirer.py 140 - try: - bucket_diskbytes = s.st_blocks * 512 - except AttributeError: - bucket_diskbytes = 0 # no stat().st_blocks on windows if sum([wks[0] for wks in would_keep_shares]) == 0: hunk ./src/allmydata/storage/expirer.py 141 - self.increment_bucketspace("original", bucket_diskbytes, sharetype) + self.increment_container_space("original", container_diskbytes, sharetype) if sum([wks[1] for wks in would_keep_shares]) == 0: hunk ./src/allmydata/storage/expirer.py 143 - self.increment_bucketspace("configured", bucket_diskbytes, sharetype) + self.increment_container_space("configured", container_diskbytes, sharetype) if sum([wks[2] for wks in would_keep_shares]) == 0: hunk ./src/allmydata/storage/expirer.py 145 - self.increment_bucketspace("actual", bucket_diskbytes, sharetype) + self.increment_container_space("actual", container_diskbytes, sharetype) hunk ./src/allmydata/storage/expirer.py 147 - def process_share(self, sharefilename): - # first, find out what kind of a share it is - sf = get_share_file(sharefilename) - sharetype = sf.sharetype + def process_share(self, share): + sharetype = share.sharetype now = time.time() hunk ./src/allmydata/storage/expirer.py 150 - s = self.stat(sharefilename) + sharebytes = share.get_size() + diskbytes = share.get_used_space() num_leases = 0 num_valid_leases_original = 0 hunk ./src/allmydata/storage/expirer.py 158 num_valid_leases_configured = 0 expired_leases_configured = [] - for li in sf.get_leases(): + for li in share.get_leases(): num_leases += 1 original_expiration_time = li.get_expiration_time() grant_renew_time = li.get_grant_renew_time_time() hunk ./src/allmydata/storage/expirer.py 171 # expired-or-not according to our configured age limit expired = False - if self.mode == "age": - age_limit = original_expiration_time - if self.override_lease_duration is not None: - age_limit = self.override_lease_duration - if age > age_limit: - expired = True - else: - assert self.mode == "cutoff-date" - if grant_renew_time < self.cutoff_date: - expired = True - if sharetype not in self.sharetypes_to_expire: - expired = False + if sharetype in self.sharetypes_to_expire: + if self.mode == "age": + age_limit = original_expiration_time + if self.override_lease_duration is not None: + age_limit = self.override_lease_duration + if age > age_limit: + expired = True + else: + assert self.mode == "cutoff-date" + if grant_renew_time < self.cutoff_date: + expired = True if expired: expired_leases_configured.append(li) hunk ./src/allmydata/storage/expirer.py 190 so_far = self.state["cycle-to-date"] self.increment(so_far["leases-per-share-histogram"], num_leases, 1) - self.increment_space("examined", s, sharetype) + self.increment_space("examined", diskbytes, sharetype) would_keep_share = [1, 1, 1, sharetype] hunk ./src/allmydata/storage/expirer.py 196 if self.expiration_enabled: for li in expired_leases_configured: - sf.cancel_lease(li.cancel_secret) + share.cancel_lease(li.cancel_secret) if num_valid_leases_original == 0: would_keep_share[0] = 0 hunk ./src/allmydata/storage/expirer.py 200 - self.increment_space("original", s, sharetype) + self.increment_space("original", sharebytes, diskbytes, sharetype) if num_valid_leases_configured == 0: would_keep_share[1] = 0 hunk ./src/allmydata/storage/expirer.py 204 - self.increment_space("configured", s, sharetype) + self.increment_space("configured", sharebytes, diskbytes, sharetype) if self.expiration_enabled: would_keep_share[2] = 0 hunk ./src/allmydata/storage/expirer.py 207 - self.increment_space("actual", s, sharetype) + self.increment_space("actual", sharebytes, diskbytes, sharetype) return would_keep_share hunk ./src/allmydata/storage/expirer.py 211 - def increment_space(self, a, s, sharetype): - sharebytes = s.st_size - try: - # note that stat(2) says that st_blocks is 512 bytes, and that - # st_blksize is "optimal file sys I/O ops blocksize", which is - # independent of the block-size that st_blocks uses. - diskbytes = s.st_blocks * 512 - except AttributeError: - # the docs say that st_blocks is only on linux. I also see it on - # MacOS. But it isn't available on windows. - diskbytes = sharebytes + def increment_space(self, a, sharebytes, diskbytes, sharetype): so_far_sr = self.state["cycle-to-date"]["space-recovered"] self.increment(so_far_sr, a+"-shares", 1) self.increment(so_far_sr, a+"-sharebytes", sharebytes) hunk ./src/allmydata/storage/expirer.py 221 self.increment(so_far_sr, a+"-sharebytes-"+sharetype, sharebytes) self.increment(so_far_sr, a+"-diskbytes-"+sharetype, diskbytes) - def increment_bucketspace(self, a, bucket_diskbytes, sharetype): + def increment_container_space(self, a, container_diskbytes, container_type): rec = self.state["cycle-to-date"]["space-recovered"] hunk ./src/allmydata/storage/expirer.py 223 - self.increment(rec, a+"-diskbytes", bucket_diskbytes) + self.increment(rec, a+"-diskbytes", container_diskbytes) self.increment(rec, a+"-buckets", 1) hunk ./src/allmydata/storage/expirer.py 225 - if sharetype: - self.increment(rec, a+"-diskbytes-"+sharetype, bucket_diskbytes) - self.increment(rec, a+"-buckets-"+sharetype, 1) + if container_type: + self.increment(rec, a+"-diskbytes-"+container_type, container_diskbytes) + self.increment(rec, a+"-buckets-"+container_type, 1) def increment(self, d, k, delta=1): if k not in d: hunk ./src/allmydata/storage/expirer.py 281 # copy() needs to become a deepcopy h["space-recovered"] = s["space-recovered"].copy() - history = pickle.load(open(self.historyfile, "rb")) + history = pickle.load(self.historyfp.getContent()) history[cycle] = h while len(history) > 10: oldcycles = sorted(history.keys()) hunk ./src/allmydata/storage/expirer.py 286 del history[oldcycles[0]] - f = open(self.historyfile, "wb") - pickle.dump(history, f) - f.close() + self.historyfp.setContent(pickle.dumps(history)) def get_state(self): """In addition to the crawler state described in hunk ./src/allmydata/storage/expirer.py 355 progress = self.get_progress() state = ShareCrawler.get_state(self) # does a shallow copy - history = pickle.load(open(self.historyfile, "rb")) + history = pickle.load(self.historyfp.getContent()) state["history"] = history if not progress["cycle-in-progress"]: hunk ./src/allmydata/storage/lease.py 3 import struct, time + +class NonExistentLeaseError(Exception): + pass + class LeaseInfo: def __init__(self, owner_num=None, renew_secret=None, cancel_secret=None, expiration_time=None, nodeid=None): hunk ./src/allmydata/storage/lease.py 21 def get_expiration_time(self): return self.expiration_time + def get_grant_renew_time_time(self): # hack, based upon fixed 31day expiration period return self.expiration_time - 31*24*60*60 hunk ./src/allmydata/storage/lease.py 25 + def get_age(self): return time.time() - self.get_grant_renew_time_time() hunk ./src/allmydata/storage/lease.py 36 self.expiration_time) = struct.unpack(">L32s32sL", data) self.nodeid = None return self + def to_immutable_data(self): return struct.pack(">L32s32sL", self.owner_num, hunk ./src/allmydata/storage/lease.py 49 int(self.expiration_time), self.renew_secret, self.cancel_secret, self.nodeid) + def from_mutable_data(self, data): (self.owner_num, self.expiration_time, hunk ./src/allmydata/storage/server.py 1 -import os, re, weakref, struct, time +import weakref, time from foolscap.api import Referenceable from twisted.application import service hunk ./src/allmydata/storage/server.py 7 from zope.interface import implements -from allmydata.interfaces import RIStorageServer, IStatsProducer -from allmydata.util import fileutil, idlib, log, time_format +from allmydata.interfaces import RIStorageServer, IStatsProducer, IStorageBackend +from allmydata.util.assertutil import precondition +from allmydata.util import idlib, log import allmydata # for __full_version__ hunk ./src/allmydata/storage/server.py 12 -from allmydata.storage.common import si_b2a, si_a2b, storage_index_to_dir -_pyflakes_hush = [si_b2a, si_a2b, storage_index_to_dir] # re-exported +from allmydata.storage.common import si_a2b, si_b2a +[si_a2b] # hush pyflakes from allmydata.storage.lease import LeaseInfo hunk ./src/allmydata/storage/server.py 15 -from allmydata.storage.mutable import MutableShareFile, EmptyShare, \ - create_mutable_sharefile -from allmydata.storage.immutable import ShareFile, BucketWriter, BucketReader -from allmydata.storage.crawler import BucketCountingCrawler from allmydata.storage.expirer import LeaseCheckingCrawler hunk ./src/allmydata/storage/server.py 16 - -# storage/ -# storage/shares/incoming -# incoming/ holds temp dirs named $START/$STORAGEINDEX/$SHARENUM which will -# be moved to storage/shares/$START/$STORAGEINDEX/$SHARENUM upon success -# storage/shares/$START/$STORAGEINDEX -# storage/shares/$START/$STORAGEINDEX/$SHARENUM - -# Where "$START" denotes the first 10 bits worth of $STORAGEINDEX (that's 2 -# base-32 chars). - -# $SHARENUM matches this regex: -NUM_RE=re.compile("^[0-9]+$") - +from allmydata.storage.crawler import BucketCountingCrawler class StorageServer(service.MultiService, Referenceable): hunk ./src/allmydata/storage/server.py 21 implements(RIStorageServer, IStatsProducer) + name = 'storage' LeaseCheckerClass = LeaseCheckingCrawler hunk ./src/allmydata/storage/server.py 24 + DEFAULT_EXPIRATION_POLICY = { + 'enabled': False, + 'mode': 'age', + 'override_lease_duration': None, + 'cutoff_date': None, + 'sharetypes': ('mutable', 'immutable'), + } hunk ./src/allmydata/storage/server.py 32 - def __init__(self, storedir, nodeid, reserved_space=0, - discard_storage=False, readonly_storage=False, + def __init__(self, serverid, backend, statedir, stats_provider=None, hunk ./src/allmydata/storage/server.py 34 - expiration_enabled=False, - expiration_mode="age", - expiration_override_lease_duration=None, - expiration_cutoff_date=None, - expiration_sharetypes=("mutable", "immutable")): + expiration_policy=None): service.MultiService.__init__(self) hunk ./src/allmydata/storage/server.py 36 - assert isinstance(nodeid, str) - assert len(nodeid) == 20 - self.my_nodeid = nodeid - self.storedir = storedir - sharedir = os.path.join(storedir, "shares") - fileutil.make_dirs(sharedir) - self.sharedir = sharedir - # we don't actually create the corruption-advisory dir until necessary - self.corruption_advisory_dir = os.path.join(storedir, - "corruption-advisories") - self.reserved_space = int(reserved_space) - self.no_storage = discard_storage - self.readonly_storage = readonly_storage + precondition(IStorageBackend.providedBy(backend), backend) + precondition(isinstance(serverid, str), serverid) + precondition(len(serverid) == 20, serverid) + + self._serverid = serverid self.stats_provider = stats_provider if self.stats_provider: self.stats_provider.register_producer(self) hunk ./src/allmydata/storage/server.py 44 - self.incomingdir = os.path.join(sharedir, 'incoming') - self._clean_incomplete() - fileutil.make_dirs(self.incomingdir) self._active_writers = weakref.WeakKeyDictionary() hunk ./src/allmydata/storage/server.py 45 + self.backend = backend + self.backend.setServiceParent(self) + self._statedir = statedir log.msg("StorageServer created", facility="tahoe.storage") hunk ./src/allmydata/storage/server.py 50 - if reserved_space: - if self.get_available_space() is None: - log.msg("warning: [storage]reserved_space= is set, but this platform does not support an API to get disk statistics (statvfs(2) or GetDiskFreeSpaceEx), so this reservation cannot be honored", - umin="0wZ27w", level=log.UNUSUAL) - self.latencies = {"allocate": [], # immutable "write": [], "close": [], hunk ./src/allmydata/storage/server.py 61 "renew": [], "cancel": [], } - self.add_bucket_counter() - - statefile = os.path.join(self.storedir, "lease_checker.state") - historyfile = os.path.join(self.storedir, "lease_checker.history") - klass = self.LeaseCheckerClass - self.lease_checker = klass(self, statefile, historyfile, - expiration_enabled, expiration_mode, - expiration_override_lease_duration, - expiration_cutoff_date, - expiration_sharetypes) - self.lease_checker.setServiceParent(self) + self._setup_bucket_counter() + self._setup_lease_checker(expiration_policy or self.DEFAULT_EXPIRATION_POLICY) def __repr__(self): hunk ./src/allmydata/storage/server.py 65 - return "" % (idlib.shortnodeid_b2a(self.my_nodeid),) + return "" % (idlib.shortnodeid_b2a(self._serverid),) hunk ./src/allmydata/storage/server.py 67 - def add_bucket_counter(self): - statefile = os.path.join(self.storedir, "bucket_counter.state") - self.bucket_counter = BucketCountingCrawler(self, statefile) + def _setup_bucket_counter(self): + statefp = self._statedir.child("bucket_counter.state") + self.bucket_counter = BucketCountingCrawler(self.backend, statefp) self.bucket_counter.setServiceParent(self) hunk ./src/allmydata/storage/server.py 72 + def _setup_lease_checker(self, expiration_policy): + statefp = self._statedir.child("lease_checker.state") + historyfp = self._statedir.child("lease_checker.history") + self.lease_checker = self.LeaseCheckerClass(self.backend, statefp, historyfp, expiration_policy) + self.lease_checker.setServiceParent(self) + def count(self, name, delta=1): if self.stats_provider: self.stats_provider.count("storage_server." + name, delta) hunk ./src/allmydata/storage/server.py 92 """Return a dict, indexed by category, that contains a dict of latency numbers for each category. If there are sufficient samples for unambiguous interpretation, each dict will contain the - following keys: mean, 01_0_percentile, 10_0_percentile, + following keys: samplesize, mean, 01_0_percentile, 10_0_percentile, 50_0_percentile (median), 90_0_percentile, 95_0_percentile, 99_0_percentile, 99_9_percentile. If there are insufficient samples for a given percentile to be interpreted unambiguously hunk ./src/allmydata/storage/server.py 114 else: stats["mean"] = None - orderstatlist = [(0.01, "01_0_percentile", 100), (0.1, "10_0_percentile", 10),\ - (0.50, "50_0_percentile", 10), (0.90, "90_0_percentile", 10),\ - (0.95, "95_0_percentile", 20), (0.99, "99_0_percentile", 100),\ + orderstatlist = [(0.1, "10_0_percentile", 10), (0.5, "50_0_percentile", 10), \ + (0.9, "90_0_percentile", 10), (0.95, "95_0_percentile", 20), \ + (0.01, "01_0_percentile", 100), (0.99, "99_0_percentile", 100),\ (0.999, "99_9_percentile", 1000)] for percentile, percentilestring, minnumtoobserve in orderstatlist: hunk ./src/allmydata/storage/server.py 133 kwargs["facility"] = "tahoe.storage" return log.msg(*args, **kwargs) - def _clean_incomplete(self): - fileutil.rm_dir(self.incomingdir) + def get_serverid(self): + return self._serverid def get_stats(self): # remember: RIStatsProvider requires that our return dict hunk ./src/allmydata/storage/server.py 138 - # contains numeric values. + # contains numeric, or None values. stats = { 'storage_server.allocated': self.allocated_size(), } hunk ./src/allmydata/storage/server.py 140 - stats['storage_server.reserved_space'] = self.reserved_space for category,ld in self.get_latencies().items(): for name,v in ld.items(): stats['storage_server.latencies.%s.%s' % (category, name)] = v hunk ./src/allmydata/storage/server.py 144 - try: - disk = fileutil.get_disk_stats(self.sharedir, self.reserved_space) - writeable = disk['avail'] > 0 - - # spacetime predictors should use disk_avail / (d(disk_used)/dt) - stats['storage_server.disk_total'] = disk['total'] - stats['storage_server.disk_used'] = disk['used'] - stats['storage_server.disk_free_for_root'] = disk['free_for_root'] - stats['storage_server.disk_free_for_nonroot'] = disk['free_for_nonroot'] - stats['storage_server.disk_avail'] = disk['avail'] - except AttributeError: - writeable = True - except EnvironmentError: - log.msg("OS call to get disk statistics failed", level=log.UNUSUAL) - writeable = False - - if self.readonly_storage: - stats['storage_server.disk_avail'] = 0 - writeable = False + self.backend.fill_in_space_stats(stats) hunk ./src/allmydata/storage/server.py 146 - stats['storage_server.accepting_immutable_shares'] = int(writeable) s = self.bucket_counter.get_state() bucket_count = s.get("last-complete-bucket-count") if bucket_count: hunk ./src/allmydata/storage/server.py 153 return stats def get_available_space(self): - """Returns available space for share storage in bytes, or None if no - API to get this information is available.""" - - if self.readonly_storage: - return 0 - return fileutil.get_available_space(self.sharedir, self.reserved_space) + return self.backend.get_available_space() def allocated_size(self): space = 0 hunk ./src/allmydata/storage/server.py 162 return space def remote_get_version(self): - remaining_space = self.get_available_space() + remaining_space = self.backend.get_available_space() if remaining_space is None: # We're on a platform that has no API to get disk stats. remaining_space = 2**64 hunk ./src/allmydata/storage/server.py 178 } return version - def remote_allocate_buckets(self, storage_index, + def remote_allocate_buckets(self, storageindex, renew_secret, cancel_secret, sharenums, allocated_size, canary, owner_num=0): hunk ./src/allmydata/storage/server.py 182 + # cancel_secret is no longer used. # owner_num is not for clients to set, but rather it should be hunk ./src/allmydata/storage/server.py 184 - # curried into the PersonalStorageServer instance that is dedicated - # to a particular owner. + # curried into a StorageServer instance dedicated to a particular + # owner. start = time.time() self.count("allocate") hunk ./src/allmydata/storage/server.py 188 - alreadygot = set() bucketwriters = {} # k: shnum, v: BucketWriter hunk ./src/allmydata/storage/server.py 189 - si_dir = storage_index_to_dir(storage_index) - si_s = si_b2a(storage_index) hunk ./src/allmydata/storage/server.py 190 + si_s = si_b2a(storageindex) log.msg("storage: allocate_buckets %s" % si_s) hunk ./src/allmydata/storage/server.py 193 - # in this implementation, the lease information (including secrets) - # goes into the share files themselves. It could also be put into a - # separate database. Note that the lease should not be added until - # the BucketWriter has been closed. + # Note that the lease should not be added until the BucketWriter + # has been closed. expire_time = time.time() + 31*24*60*60 hunk ./src/allmydata/storage/server.py 196 - lease_info = LeaseInfo(owner_num, - renew_secret, cancel_secret, - expire_time, self.my_nodeid) + lease_info = LeaseInfo(owner_num, renew_secret, + expire_time, self._serverid) max_space_per_bucket = allocated_size hunk ./src/allmydata/storage/server.py 201 - remaining_space = self.get_available_space() + remaining_space = self.backend.get_available_space() limited = remaining_space is not None if limited: hunk ./src/allmydata/storage/server.py 204 - # this is a bit conservative, since some of this allocated_size() - # has already been written to disk, where it will show up in + # This is a bit conservative, since some of this allocated_size() + # has already been written to the backend, where it will show up in # get_available_space. remaining_space -= self.allocated_size() hunk ./src/allmydata/storage/server.py 208 - # self.readonly_storage causes remaining_space <= 0 + # If the backend is read-only, remaining_space will be <= 0. + + shareset = self.backend.get_shareset(storageindex) hunk ./src/allmydata/storage/server.py 212 - # fill alreadygot with all shares that we have, not just the ones + # Fill alreadygot with all shares that we have, not just the ones # they asked about: this will save them a lot of work. Add or update # leases for all of them: if they want us to hold shares for this hunk ./src/allmydata/storage/server.py 215 - # file, they'll want us to hold leases for this file. - for (shnum, fn) in self._get_bucket_shares(storage_index): - alreadygot.add(shnum) - sf = ShareFile(fn) - sf.add_or_renew_lease(lease_info) + # file, they'll want us to hold leases for all the shares of it. + # + # XXX should we be making the assumption here that lease info is + # duplicated in all shares? + alreadygot = set() + for share in shareset.get_shares(): + share.add_or_renew_lease(lease_info) + alreadygot.add(share.shnum) hunk ./src/allmydata/storage/server.py 224 - for shnum in sharenums: - incominghome = os.path.join(self.incomingdir, si_dir, "%d" % shnum) - finalhome = os.path.join(self.sharedir, si_dir, "%d" % shnum) - if os.path.exists(finalhome): - # great! we already have it. easy. - pass - elif os.path.exists(incominghome): + for shnum in sharenums - alreadygot: + if shareset.has_incoming(shnum): # Note that we don't create BucketWriters for shnums that # have a partial share (in incoming/), so if a second upload # occurs while the first is still in progress, the second hunk ./src/allmydata/storage/server.py 232 # uploader will use different storage servers. pass elif (not limited) or (remaining_space >= max_space_per_bucket): - # ok! we need to create the new share file. - bw = BucketWriter(self, incominghome, finalhome, - max_space_per_bucket, lease_info, canary) - if self.no_storage: - bw.throw_out_all_data = True + bw = shareset.make_bucket_writer(self, shnum, max_space_per_bucket, + lease_info, canary) bucketwriters[shnum] = bw self._active_writers[bw] = 1 if limited: hunk ./src/allmydata/storage/server.py 239 remaining_space -= max_space_per_bucket else: - # bummer! not enough space to accept this bucket + # Bummer not enough space to accept this share. pass hunk ./src/allmydata/storage/server.py 242 - if bucketwriters: - fileutil.make_dirs(os.path.join(self.sharedir, si_dir)) - self.add_latency("allocate", time.time() - start) return alreadygot, bucketwriters hunk ./src/allmydata/storage/server.py 245 - def _iter_share_files(self, storage_index): - for shnum, filename in self._get_bucket_shares(storage_index): - f = open(filename, 'rb') - header = f.read(32) - f.close() - if header[:32] == MutableShareFile.MAGIC: - sf = MutableShareFile(filename, self) - # note: if the share has been migrated, the renew_lease() - # call will throw an exception, with information to help the - # client update the lease. - elif header[:4] == struct.pack(">L", 1): - sf = ShareFile(filename) - else: - continue # non-sharefile - yield sf - - def remote_add_lease(self, storage_index, renew_secret, cancel_secret, + def remote_add_lease(self, storageindex, renew_secret, cancel_secret, owner_num=1): hunk ./src/allmydata/storage/server.py 247 + # cancel_secret is no longer used. start = time.time() self.count("add-lease") new_expire_time = time.time() + 31*24*60*60 hunk ./src/allmydata/storage/server.py 251 - lease_info = LeaseInfo(owner_num, - renew_secret, cancel_secret, - new_expire_time, self.my_nodeid) - for sf in self._iter_share_files(storage_index): - sf.add_or_renew_lease(lease_info) - self.add_latency("add-lease", time.time() - start) - return None + lease_info = LeaseInfo(owner_num, renew_secret, + new_expire_time, self._serverid) hunk ./src/allmydata/storage/server.py 254 - def remote_renew_lease(self, storage_index, renew_secret): + try: + self.backend.add_or_renew_lease(lease_info) + finally: + self.add_latency("add-lease", time.time() - start) + + def remote_renew_lease(self, storageindex, renew_secret): start = time.time() self.count("renew") hunk ./src/allmydata/storage/server.py 262 - new_expire_time = time.time() + 31*24*60*60 - found_buckets = False - for sf in self._iter_share_files(storage_index): - found_buckets = True - sf.renew_lease(renew_secret, new_expire_time) - self.add_latency("renew", time.time() - start) - if not found_buckets: - raise IndexError("no such lease to renew") + + try: + shareset = self.backend.get_shareset(storageindex) + new_expiration_time = start + 31*24*60*60 # one month from now + shareset.renew_lease(renew_secret, new_expiration_time) + finally: + self.add_latency("renew", time.time() - start) def bucket_writer_closed(self, bw, consumed_size): if self.stats_provider: hunk ./src/allmydata/storage/server.py 275 self.stats_provider.count('storage_server.bytes_added', consumed_size) del self._active_writers[bw] - def _get_bucket_shares(self, storage_index): - """Return a list of (shnum, pathname) tuples for files that hold - shares for this storage_index. In each tuple, 'shnum' will always be - the integer form of the last component of 'pathname'.""" - storagedir = os.path.join(self.sharedir, storage_index_to_dir(storage_index)) - try: - for f in os.listdir(storagedir): - if NUM_RE.match(f): - filename = os.path.join(storagedir, f) - yield (int(f), filename) - except OSError: - # Commonly caused by there being no buckets at all. - pass - - def remote_get_buckets(self, storage_index): + def remote_get_buckets(self, storageindex): start = time.time() self.count("get") hunk ./src/allmydata/storage/server.py 278 - si_s = si_b2a(storage_index) + si_s = si_b2a(storageindex) log.msg("storage: get_buckets %s" % si_s) bucketreaders = {} # k: sharenum, v: BucketReader hunk ./src/allmydata/storage/server.py 281 - for shnum, filename in self._get_bucket_shares(storage_index): - bucketreaders[shnum] = BucketReader(self, filename, - storage_index, shnum) - self.add_latency("get", time.time() - start) - return bucketreaders hunk ./src/allmydata/storage/server.py 282 - def get_leases(self, storage_index): - """Provide an iterator that yields all of the leases attached to this - bucket. Each lease is returned as a LeaseInfo instance. + try: + shareset = self.backend.get_shareset(storageindex) + for share in shareset.get_shares(): + bucketreaders[share.get_shnum()] = shareset.make_bucket_reader(self, share) + return bucketreaders + finally: + self.add_latency("get", time.time() - start) hunk ./src/allmydata/storage/server.py 290 - This method is not for client use. + def get_leases(self, storageindex): """ hunk ./src/allmydata/storage/server.py 292 + Provide an iterator that yields all of the leases attached to this + bucket. Each lease is returned as a LeaseInfo instance. hunk ./src/allmydata/storage/server.py 295 - # since all shares get the same lease data, we just grab the leases - # from the first share - try: - shnum, filename = self._get_bucket_shares(storage_index).next() - sf = ShareFile(filename) - return sf.get_leases() - except StopIteration: - return iter([]) + This method is not for client use. XXX do we need it at all? + """ + return self.backend.get_shareset(storageindex).get_leases() hunk ./src/allmydata/storage/server.py 299 - def remote_slot_testv_and_readv_and_writev(self, storage_index, + def remote_slot_testv_and_readv_and_writev(self, storageindex, secrets, test_and_write_vectors, read_vector): hunk ./src/allmydata/storage/server.py 305 start = time.time() self.count("writev") - si_s = si_b2a(storage_index) + si_s = si_b2a(storageindex) log.msg("storage: slot_writev %s" % si_s) hunk ./src/allmydata/storage/server.py 307 - si_dir = storage_index_to_dir(storage_index) - (write_enabler, renew_secret, cancel_secret) = secrets - # shares exist if there is a file for them - bucketdir = os.path.join(self.sharedir, si_dir) - shares = {} - if os.path.isdir(bucketdir): - for sharenum_s in os.listdir(bucketdir): - try: - sharenum = int(sharenum_s) - except ValueError: - continue - filename = os.path.join(bucketdir, sharenum_s) - msf = MutableShareFile(filename, self) - msf.check_write_enabler(write_enabler, si_s) - shares[sharenum] = msf - # write_enabler is good for all existing shares. - - # Now evaluate test vectors. - testv_is_good = True - for sharenum in test_and_write_vectors: - (testv, datav, new_length) = test_and_write_vectors[sharenum] - if sharenum in shares: - if not shares[sharenum].check_testv(testv): - self.log("testv failed: [%d]: %r" % (sharenum, testv)) - testv_is_good = False - break - else: - # compare the vectors against an empty share, in which all - # reads return empty strings. - if not EmptyShare().check_testv(testv): - self.log("testv failed (empty): [%d] %r" % (sharenum, - testv)) - testv_is_good = False - break - - # now gather the read vectors, before we do any writes - read_data = {} - for sharenum, share in shares.items(): - read_data[sharenum] = share.readv(read_vector) - - ownerid = 1 # TODO - expire_time = time.time() + 31*24*60*60 # one month - lease_info = LeaseInfo(ownerid, - renew_secret, cancel_secret, - expire_time, self.my_nodeid) - - if testv_is_good: - # now apply the write vectors - for sharenum in test_and_write_vectors: - (testv, datav, new_length) = test_and_write_vectors[sharenum] - if new_length == 0: - if sharenum in shares: - shares[sharenum].unlink() - else: - if sharenum not in shares: - # allocate a new share - allocated_size = 2000 # arbitrary, really - share = self._allocate_slot_share(bucketdir, secrets, - sharenum, - allocated_size, - owner_num=0) - shares[sharenum] = share - shares[sharenum].writev(datav, new_length) - # and update the lease - shares[sharenum].add_or_renew_lease(lease_info) - - if new_length == 0: - # delete empty bucket directories - if not os.listdir(bucketdir): - os.rmdir(bucketdir) hunk ./src/allmydata/storage/server.py 308 + try: + shareset = self.backend.get_shareset(storageindex) + expiration_time = start + 31*24*60*60 # one month from now + return shareset.testv_and_readv_and_writev(self, secrets, test_and_write_vectors, + read_vector, expiration_time) + finally: + self.add_latency("writev", time.time() - start) hunk ./src/allmydata/storage/server.py 316 - # all done - self.add_latency("writev", time.time() - start) - return (testv_is_good, read_data) - - def _allocate_slot_share(self, bucketdir, secrets, sharenum, - allocated_size, owner_num=0): - (write_enabler, renew_secret, cancel_secret) = secrets - my_nodeid = self.my_nodeid - fileutil.make_dirs(bucketdir) - filename = os.path.join(bucketdir, "%d" % sharenum) - share = create_mutable_sharefile(filename, my_nodeid, write_enabler, - self) - return share - - def remote_slot_readv(self, storage_index, shares, readv): + def remote_slot_readv(self, storageindex, shares, readv): start = time.time() self.count("readv") hunk ./src/allmydata/storage/server.py 319 - si_s = si_b2a(storage_index) - lp = log.msg("storage: slot_readv %s %s" % (si_s, shares), - facility="tahoe.storage", level=log.OPERATIONAL) - si_dir = storage_index_to_dir(storage_index) - # shares exist if there is a file for them - bucketdir = os.path.join(self.sharedir, si_dir) - if not os.path.isdir(bucketdir): + si_s = si_b2a(storageindex) + log.msg("storage: slot_readv %s %s" % (si_s, shares), + facility="tahoe.storage", level=log.OPERATIONAL) + + try: + shareset = self.backend.get_shareset(storageindex) + return shareset.readv(self, shares, readv) + finally: self.add_latency("readv", time.time() - start) hunk ./src/allmydata/storage/server.py 328 - return {} - datavs = {} - for sharenum_s in os.listdir(bucketdir): - try: - sharenum = int(sharenum_s) - except ValueError: - continue - if sharenum in shares or not shares: - filename = os.path.join(bucketdir, sharenum_s) - msf = MutableShareFile(filename, self) - datavs[sharenum] = msf.readv(readv) - log.msg("returning shares %s" % (datavs.keys(),), - facility="tahoe.storage", level=log.NOISY, parent=lp) - self.add_latency("readv", time.time() - start) - return datavs hunk ./src/allmydata/storage/server.py 329 - def remote_advise_corrupt_share(self, share_type, storage_index, shnum, - reason): - fileutil.make_dirs(self.corruption_advisory_dir) - now = time_format.iso_utc(sep="T") - si_s = si_b2a(storage_index) - # windows can't handle colons in the filename - fn = os.path.join(self.corruption_advisory_dir, - "%s--%s-%d" % (now, si_s, shnum)).replace(":","") - f = open(fn, "w") - f.write("report: Share Corruption\n") - f.write("type: %s\n" % share_type) - f.write("storage_index: %s\n" % si_s) - f.write("share_number: %d\n" % shnum) - f.write("\n") - f.write(reason) - f.write("\n") - f.close() - log.msg(format=("client claims corruption in (%(share_type)s) " + - "%(si)s-%(shnum)d: %(reason)s"), - share_type=share_type, si=si_s, shnum=shnum, reason=reason, - level=log.SCARY, umid="SGx2fA") - return None + def remote_advise_corrupt_share(self, share_type, storage_index, shnum, reason): + self.backend.advise_corrupt_share(share_type, storage_index, shnum, reason) hunk ./src/allmydata/test/common.py 20 from allmydata.mutable.common import CorruptShareError from allmydata.mutable.layout import unpack_header from allmydata.mutable.publish import MutableData -from allmydata.storage.mutable import MutableShareFile +from allmydata.storage.backends.disk.mutable import MutableDiskShare from allmydata.util import hashutil, log, fileutil, pollmixin from allmydata.util.assertutil import precondition from allmydata.util.consumer import download_to_data hunk ./src/allmydata/test/common.py 1297 def _corrupt_mutable_share_data(data, debug=False): prefix = data[:32] - assert prefix == MutableShareFile.MAGIC, "This function is designed to corrupt mutable shares of v1, and the magic number doesn't look right: %r vs %r" % (prefix, MutableShareFile.MAGIC) - data_offset = MutableShareFile.DATA_OFFSET + assert prefix == MutableDiskShare.MAGIC, "This function is designed to corrupt mutable shares of v1, and the magic number doesn't look right: %r vs %r" % (prefix, MutableDiskShare.MAGIC) + data_offset = MutableDiskShare.DATA_OFFSET sharetype = data[data_offset:data_offset+1] assert sharetype == "\x00", "non-SDMF mutable shares not supported" (version, ig_seqnum, ig_roothash, ig_IV, ig_k, ig_N, ig_segsize, hunk ./src/allmydata/test/no_network.py 21 from twisted.application import service from twisted.internet import defer, reactor from twisted.python.failure import Failure +from twisted.python.filepath import FilePath from foolscap.api import Referenceable, fireEventually, RemoteException from base64 import b32encode hunk ./src/allmydata/test/no_network.py 24 + from allmydata import uri as tahoe_uri from allmydata.client import Client hunk ./src/allmydata/test/no_network.py 27 -from allmydata.storage.server import StorageServer, storage_index_to_dir +from allmydata.storage.server import StorageServer +from allmydata.storage.backends.disk.disk_backend import DiskBackend from allmydata.util import fileutil, idlib, hashutil from allmydata.util.hashutil import sha1 from allmydata.test.common_web import HTTPClientGETFactory hunk ./src/allmydata/test/no_network.py 155 seed = server.get_permutation_seed() return sha1(peer_selection_index + seed).digest() return sorted(self.get_connected_servers(), key=_permuted) + def get_connected_servers(self): return self.client._servers hunk ./src/allmydata/test/no_network.py 158 + def get_nickname_for_serverid(self, serverid): return None hunk ./src/allmydata/test/no_network.py 162 + def get_known_servers(self): + return self.get_connected_servers() + + def get_all_serverids(self): + return self.client.get_all_serverids() + + class NoNetworkClient(Client): def create_tub(self): pass hunk ./src/allmydata/test/no_network.py 262 def make_server(self, i, readonly=False): serverid = hashutil.tagged_hash("serverid", str(i))[:20] - serverdir = os.path.join(self.basedir, "servers", - idlib.shortnodeid_b2a(serverid), "storage") - fileutil.make_dirs(serverdir) - ss = StorageServer(serverdir, serverid, stats_provider=SimpleStats(), - readonly_storage=readonly) + storagedir = FilePath(self.basedir).child("servers").child(idlib.shortnodeid_b2a(serverid)).child("storage") + + # The backend will make the storage directory and any necessary parents. + backend = DiskBackend(storagedir, readonly=readonly) + ss = StorageServer(serverid, backend, storagedir, stats_provider=SimpleStats()) ss._no_network_server_number = i return ss hunk ./src/allmydata/test/no_network.py 276 middleman = service.MultiService() middleman.setServiceParent(self) ss.setServiceParent(middleman) - serverid = ss.my_nodeid + serverid = ss.get_serverid() self.servers_by_number[i] = ss wrapper = wrap_storage_server(ss) self.wrappers_by_id[serverid] = wrapper hunk ./src/allmydata/test/no_network.py 295 # it's enough to remove the server from c._servers (we don't actually # have to detach and stopService it) for i,ss in self.servers_by_number.items(): - if ss.my_nodeid == serverid: + if ss.get_serverid() == serverid: del self.servers_by_number[i] break del self.wrappers_by_id[serverid] hunk ./src/allmydata/test/no_network.py 345 def get_clientdir(self, i=0): return self.g.clients[i].basedir + def get_server(self, i): + return self.g.servers_by_number[i] + def get_serverdir(self, i): hunk ./src/allmydata/test/no_network.py 349 - return self.g.servers_by_number[i].storedir + return self.g.servers_by_number[i].backend.storedir + + def remove_server(self, i): + self.g.remove_server(self.g.servers_by_number[i].get_serverid()) def iterate_servers(self): for i in sorted(self.g.servers_by_number.keys()): hunk ./src/allmydata/test/no_network.py 357 ss = self.g.servers_by_number[i] - yield (i, ss, ss.storedir) + yield (i, ss, ss.backend.storedir) def find_uri_shares(self, uri): si = tahoe_uri.from_string(uri).get_storage_index() hunk ./src/allmydata/test/no_network.py 361 - prefixdir = storage_index_to_dir(si) shares = [] for i,ss in self.g.servers_by_number.items(): hunk ./src/allmydata/test/no_network.py 363 - serverid = ss.my_nodeid - basedir = os.path.join(ss.sharedir, prefixdir) - if not os.path.exists(basedir): - continue - for f in os.listdir(basedir): - try: - shnum = int(f) - shares.append((shnum, serverid, os.path.join(basedir, f))) - except ValueError: - pass + for share in ss.backend.get_shareset(si).get_shares(): + shares.append((share.get_shnum(), ss.get_serverid(), share._home)) return sorted(shares) hunk ./src/allmydata/test/no_network.py 367 + def count_leases(self, uri): + """Return (filename, leasecount) pairs in arbitrary order.""" + si = tahoe_uri.from_string(uri).get_storage_index() + lease_counts = [] + for i,ss in self.g.servers_by_number.items(): + for share in ss.backend.get_shareset(si).get_shares(): + num_leases = len(list(share.get_leases())) + lease_counts.append( (share._home.path, num_leases) ) + return lease_counts + def copy_shares(self, uri): shares = {} hunk ./src/allmydata/test/no_network.py 379 - for (shnum, serverid, sharefile) in self.find_uri_shares(uri): - shares[sharefile] = open(sharefile, "rb").read() + for (shnum, serverid, sharefp) in self.find_uri_shares(uri): + shares[sharefp.path] = sharefp.getContent() return shares hunk ./src/allmydata/test/no_network.py 383 + def copy_share(self, from_share, uri, to_server): + si = uri.from_string(self.uri).get_storage_index() + (i_shnum, i_serverid, i_sharefp) = from_share + shares_dir = to_server.backend.get_shareset(si)._sharehomedir + i_sharefp.copyTo(shares_dir.child(str(i_shnum))) + def restore_all_shares(self, shares): hunk ./src/allmydata/test/no_network.py 390 - for sharefile, data in shares.items(): - open(sharefile, "wb").write(data) + for share, data in shares.items(): + share.home.setContent(data) hunk ./src/allmydata/test/no_network.py 393 - def delete_share(self, (shnum, serverid, sharefile)): - os.unlink(sharefile) + def delete_share(self, (shnum, serverid, sharefp)): + sharefp.remove() def delete_shares_numbered(self, uri, shnums): hunk ./src/allmydata/test/no_network.py 397 - for (i_shnum, i_serverid, i_sharefile) in self.find_uri_shares(uri): + for (i_shnum, i_serverid, i_sharefp) in self.find_uri_shares(uri): if i_shnum in shnums: hunk ./src/allmydata/test/no_network.py 399 - os.unlink(i_sharefile) + i_sharefp.remove() hunk ./src/allmydata/test/no_network.py 401 - def corrupt_share(self, (shnum, serverid, sharefile), corruptor_function): - sharedata = open(sharefile, "rb").read() - corruptdata = corruptor_function(sharedata) - open(sharefile, "wb").write(corruptdata) + def corrupt_share(self, (shnum, serverid, sharefp), corruptor_function, debug=False): + sharedata = sharefp.getContent() + corruptdata = corruptor_function(sharedata, debug=debug) + sharefp.setContent(corruptdata) def corrupt_shares_numbered(self, uri, shnums, corruptor, debug=False): hunk ./src/allmydata/test/no_network.py 407 - for (i_shnum, i_serverid, i_sharefile) in self.find_uri_shares(uri): + for (i_shnum, i_serverid, i_sharefp) in self.find_uri_shares(uri): if i_shnum in shnums: hunk ./src/allmydata/test/no_network.py 409 - sharedata = open(i_sharefile, "rb").read() - corruptdata = corruptor(sharedata, debug=debug) - open(i_sharefile, "wb").write(corruptdata) + self.corrupt_share((i_shnum, i_serverid, i_sharefp), corruptor, debug=debug) def corrupt_all_shares(self, uri, corruptor, debug=False): hunk ./src/allmydata/test/no_network.py 412 - for (i_shnum, i_serverid, i_sharefile) in self.find_uri_shares(uri): - sharedata = open(i_sharefile, "rb").read() - corruptdata = corruptor(sharedata, debug=debug) - open(i_sharefile, "wb").write(corruptdata) + for (i_shnum, i_serverid, i_sharefp) in self.find_uri_shares(uri): + self.corrupt_share((i_shnum, i_serverid, i_sharefp), corruptor, debug=debug) def GET(self, urlpath, followRedirect=False, return_response=False, method="GET", clientnum=0, **kwargs): hunk ./src/allmydata/test/test_download.py 6 # a previous run. This asserts that the current code is capable of decoding # shares from a previous version. -import os from twisted.trial import unittest from twisted.internet import defer, reactor from allmydata import uri hunk ./src/allmydata/test/test_download.py 9 -from allmydata.storage.server import storage_index_to_dir from allmydata.util import base32, fileutil, spans, log, hashutil from allmydata.util.consumer import download_to_data, MemoryConsumer from allmydata.immutable import upload, layout hunk ./src/allmydata/test/test_download.py 85 u = upload.Data(plaintext, None) d = self.c0.upload(u) f = open("stored_shares.py", "w") - def _created_immutable(ur): - # write the generated shares and URI to a file, which can then be - # incorporated into this one next time. - f.write('immutable_uri = "%s"\n' % ur.uri) - f.write('immutable_shares = {\n') - si = uri.from_string(ur.uri).get_storage_index() - si_dir = storage_index_to_dir(si) + + def _write_py(uri): + si = uri.from_string(uri).get_storage_index() for (i,ss,ssdir) in self.iterate_servers(): hunk ./src/allmydata/test/test_download.py 89 - sharedir = os.path.join(ssdir, "shares", si_dir) shares = {} hunk ./src/allmydata/test/test_download.py 90 - for fn in os.listdir(sharedir): - shnum = int(fn) - sharedata = open(os.path.join(sharedir, fn), "rb").read() - shares[shnum] = sharedata - fileutil.rm_dir(sharedir) + shareset = ss.backend.get_shareset(si) + for share in shareset.get_shares(): + sharedata = share._home.getContent() + shares[share.get_shnum()] = sharedata + + fileutil.fp_remove(shareset._sharehomedir) if shares: f.write(' %d: { # client[%d]\n' % (i, i)) for shnum in sorted(shares.keys()): hunk ./src/allmydata/test/test_download.py 103 (shnum, base32.b2a(shares[shnum]))) f.write(' },\n') f.write('}\n') - f.write('\n') hunk ./src/allmydata/test/test_download.py 104 + def _created_immutable(ur): + # write the generated shares and URI to a file, which can then be + # incorporated into this one next time. + f.write('immutable_uri = "%s"\n' % ur.uri) + f.write('immutable_shares = {\n') + _write_py(ur.uri) + f.write('\n') d.addCallback(_created_immutable) d.addCallback(lambda ignored: hunk ./src/allmydata/test/test_download.py 118 def _created_mutable(n): f.write('mutable_uri = "%s"\n' % n.get_uri()) f.write('mutable_shares = {\n') - si = uri.from_string(n.get_uri()).get_storage_index() - si_dir = storage_index_to_dir(si) - for (i,ss,ssdir) in self.iterate_servers(): - sharedir = os.path.join(ssdir, "shares", si_dir) - shares = {} - for fn in os.listdir(sharedir): - shnum = int(fn) - sharedata = open(os.path.join(sharedir, fn), "rb").read() - shares[shnum] = sharedata - fileutil.rm_dir(sharedir) - if shares: - f.write(' %d: { # client[%d]\n' % (i, i)) - for shnum in sorted(shares.keys()): - f.write(' %d: base32.a2b("%s"),\n' % - (shnum, base32.b2a(shares[shnum]))) - f.write(' },\n') - f.write('}\n') - - f.close() + _write_py(n.get_uri()) d.addCallback(_created_mutable) def _done(ignored): hunk ./src/allmydata/test/test_download.py 123 f.close() - d.addCallback(_done) + d.addBoth(_done) return d hunk ./src/allmydata/test/test_download.py 127 + def _write_shares(self, uri, shares): + si = uri.from_string(uri).get_storage_index() + for i in shares: + shares_for_server = shares[i] + for shnum in shares_for_server: + share_dir = self.get_server(i).backend.get_shareset(si)._sharehomedir + fileutil.fp_make_dirs(share_dir) + share_dir.child(str(shnum)).setContent(shares[shnum]) + def load_shares(self, ignored=None): # this uses the data generated by create_shares() to populate the # storage servers with pre-generated shares hunk ./src/allmydata/test/test_download.py 139 - si = uri.from_string(immutable_uri).get_storage_index() - si_dir = storage_index_to_dir(si) - for i in immutable_shares: - shares = immutable_shares[i] - for shnum in shares: - dn = os.path.join(self.get_serverdir(i), "shares", si_dir) - fileutil.make_dirs(dn) - fn = os.path.join(dn, str(shnum)) - f = open(fn, "wb") - f.write(shares[shnum]) - f.close() - - si = uri.from_string(mutable_uri).get_storage_index() - si_dir = storage_index_to_dir(si) - for i in mutable_shares: - shares = mutable_shares[i] - for shnum in shares: - dn = os.path.join(self.get_serverdir(i), "shares", si_dir) - fileutil.make_dirs(dn) - fn = os.path.join(dn, str(shnum)) - f = open(fn, "wb") - f.write(shares[shnum]) - f.close() + self._write_shares(immutable_uri, immutable_shares) + self._write_shares(mutable_uri, mutable_shares) def download_immutable(self, ignored=None): n = self.c0.create_node_from_uri(immutable_uri) hunk ./src/allmydata/test/test_download.py 183 self.load_shares() si = uri.from_string(immutable_uri).get_storage_index() - si_dir = storage_index_to_dir(si) n = self.c0.create_node_from_uri(immutable_uri) d = download_to_data(n) hunk ./src/allmydata/test/test_download.py 198 for clientnum in immutable_shares: for shnum in immutable_shares[clientnum]: if s._shnum == shnum: - fn = os.path.join(self.get_serverdir(clientnum), - "shares", si_dir, str(shnum)) - os.unlink(fn) + share_dir = self.get_server(clientnum).backend.get_shareset(si)._sharehomedir + share_dir.child(str(shnum)).remove() d.addCallback(_clobber_some_shares) d.addCallback(lambda ign: download_to_data(n)) d.addCallback(_got_data) hunk ./src/allmydata/test/test_download.py 212 for shnum in immutable_shares[clientnum]: if shnum == save_me: continue - fn = os.path.join(self.get_serverdir(clientnum), - "shares", si_dir, str(shnum)) - if os.path.exists(fn): - os.unlink(fn) + share_dir = self.get_server(clientnum).backend.get_shareset(si)._sharehomedir + fileutil.fp_remove(share_dir.child(str(shnum))) # now the download should fail with NotEnoughSharesError return self.shouldFail(NotEnoughSharesError, "1shares", None, download_to_data, n) hunk ./src/allmydata/test/test_download.py 223 # delete the last remaining share for clientnum in immutable_shares: for shnum in immutable_shares[clientnum]: - fn = os.path.join(self.get_serverdir(clientnum), - "shares", si_dir, str(shnum)) - if os.path.exists(fn): - os.unlink(fn) + share_dir = self.get_server(clientnum).backend.get_shareset(si)._sharehomedir + share_dir.child(str(shnum)).remove() # now a new download should fail with NoSharesError. We want a # new ImmutableFileNode so it will forget about the old shares. # If we merely called create_node_from_uri() without first hunk ./src/allmydata/test/test_download.py 801 # will report two shares, and the ShareFinder will handle the # duplicate by attaching both to the same CommonShare instance. si = uri.from_string(immutable_uri).get_storage_index() - si_dir = storage_index_to_dir(si) - sh0_file = [sharefile - for (shnum, serverid, sharefile) - in self.find_uri_shares(immutable_uri) - if shnum == 0][0] - sh0_data = open(sh0_file, "rb").read() + sh0_fp = [sharefp for (shnum, serverid, sharefp) + in self.find_uri_shares(immutable_uri) + if shnum == 0][0] + sh0_data = sh0_fp.getContent() for clientnum in immutable_shares: if 0 in immutable_shares[clientnum]: continue hunk ./src/allmydata/test/test_download.py 808 - cdir = self.get_serverdir(clientnum) - target = os.path.join(cdir, "shares", si_dir, "0") - outf = open(target, "wb") - outf.write(sh0_data) - outf.close() + cdir = self.get_server(clientnum).backend.get_shareset(si)._sharehomedir + fileutil.fp_make_dirs(cdir) + cdir.child(str(shnum)).setContent(sh0_data) d = self.download_immutable() return d hunk ./src/allmydata/test/test_encode.py 134 d.addCallback(_try) return d - def get_share_hashes(self, at_least_these=()): + def get_share_hashes(self): d = self._start() def _try(unused=None): if self.mode == "bad sharehash": hunk ./src/allmydata/test/test_hung_server.py 3 # -*- coding: utf-8 -*- -import os, shutil from twisted.trial import unittest from twisted.internet import defer hunk ./src/allmydata/test/test_hung_server.py 5 -from allmydata import uri + from allmydata.util.consumer import download_to_data from allmydata.immutable import upload from allmydata.mutable.common import UnrecoverableFileError hunk ./src/allmydata/test/test_hung_server.py 10 from allmydata.mutable.publish import MutableData -from allmydata.storage.common import storage_index_to_dir from allmydata.test.no_network import GridTestMixin from allmydata.test.common import ShouldFailMixin from allmydata.util.pollmixin import PollMixin hunk ./src/allmydata/test/test_hung_server.py 18 immutable_plaintext = "data" * 10000 mutable_plaintext = "muta" * 10000 + class HungServerDownloadTest(GridTestMixin, ShouldFailMixin, PollMixin, unittest.TestCase): # Many of these tests take around 60 seconds on François's ARM buildslave: hunk ./src/allmydata/test/test_hung_server.py 31 timeout = 240 def _break(self, servers): - for (id, ss) in servers: - self.g.break_server(id) + for ss in servers: + self.g.break_server(ss.get_serverid()) def _hang(self, servers, **kwargs): hunk ./src/allmydata/test/test_hung_server.py 35 - for (id, ss) in servers: - self.g.hang_server(id, **kwargs) + for ss in servers: + self.g.hang_server(ss.get_serverid(), **kwargs) def _unhang(self, servers, **kwargs): hunk ./src/allmydata/test/test_hung_server.py 39 - for (id, ss) in servers: - self.g.unhang_server(id, **kwargs) + for ss in servers: + self.g.unhang_server(ss.get_serverid(), **kwargs) def _hang_shares(self, shnums, **kwargs): # hang all servers who are holding the given shares hunk ./src/allmydata/test/test_hung_server.py 52 hung_serverids.add(i_serverid) def _delete_all_shares_from(self, servers): - serverids = [id for (id, ss) in servers] - for (i_shnum, i_serverid, i_sharefile) in self.shares: + serverids = [ss.get_serverid() for ss in servers] + for (i_shnum, i_serverid, i_sharefp) in self.shares: if i_serverid in serverids: hunk ./src/allmydata/test/test_hung_server.py 55 - os.unlink(i_sharefile) + i_sharefp.remove() def _corrupt_all_shares_in(self, servers, corruptor_func): hunk ./src/allmydata/test/test_hung_server.py 58 - serverids = [id for (id, ss) in servers] - for (i_shnum, i_serverid, i_sharefile) in self.shares: + serverids = [ss.get_serverid() for ss in servers] + for (i_shnum, i_serverid, i_sharefp) in self.shares: if i_serverid in serverids: hunk ./src/allmydata/test/test_hung_server.py 61 - self._corrupt_share((i_shnum, i_sharefile), corruptor_func) + self.corrupt_share((i_shnum, i_serverid, i_sharefp), corruptor_func) def _copy_all_shares_from(self, from_servers, to_server): hunk ./src/allmydata/test/test_hung_server.py 64 - serverids = [id for (id, ss) in from_servers] - for (i_shnum, i_serverid, i_sharefile) in self.shares: + serverids = [ss.get_serverid() for ss in from_servers] + for (i_shnum, i_serverid, i_sharefp) in self.shares: if i_serverid in serverids: hunk ./src/allmydata/test/test_hung_server.py 67 - self._copy_share((i_shnum, i_sharefile), to_server) + self.copy_share((i_shnum, i_serverid, i_sharefp), self.uri, to_server) hunk ./src/allmydata/test/test_hung_server.py 69 - def _copy_share(self, share, to_server): - (sharenum, sharefile) = share - (id, ss) = to_server - shares_dir = os.path.join(ss.original.storedir, "shares") - si = uri.from_string(self.uri).get_storage_index() - si_dir = os.path.join(shares_dir, storage_index_to_dir(si)) - if not os.path.exists(si_dir): - os.makedirs(si_dir) - new_sharefile = os.path.join(si_dir, str(sharenum)) - shutil.copy(sharefile, new_sharefile) self.shares = self.find_uri_shares(self.uri) hunk ./src/allmydata/test/test_hung_server.py 70 - # Make sure that the storage server has the share. - self.failUnless((sharenum, ss.original.my_nodeid, new_sharefile) - in self.shares) - - def _corrupt_share(self, share, corruptor_func): - (sharenum, sharefile) = share - data = open(sharefile, "rb").read() - newdata = corruptor_func(data) - os.unlink(sharefile) - wf = open(sharefile, "wb") - wf.write(newdata) - wf.close() def _set_up(self, mutable, testdir, num_clients=1, num_servers=10): self.mutable = mutable hunk ./src/allmydata/test/test_hung_server.py 82 self.c0 = self.g.clients[0] nm = self.c0.nodemaker - self.servers = sorted([(s.get_serverid(), s.get_rref()) - for s in nm.storage_broker.get_connected_servers()]) + unsorted = [(s.get_serverid(), s.get_rref()) for s in nm.storage_broker.get_connected_servers()] + self.servers = [ss for (id, ss) in sorted(unsorted)] self.servers = self.servers[5:] + self.servers[:5] if mutable: hunk ./src/allmydata/test/test_hung_server.py 244 # stuck-but-not-overdue, and 4 live requests. All 4 live requests # will retire before the download is complete and the ShareFinder # is shut off. That will leave 4 OVERDUE and 1 - # stuck-but-not-overdue, for a total of 5 requests in in + # stuck-but-not-overdue, for a total of 5 requests in # _sf.pending_requests for t in self._sf.overdue_timers.values()[:4]: t.reset(-1.0) hunk ./src/allmydata/test/test_mutable.py 21 from foolscap.api import eventually, fireEventually from foolscap.logging import log from allmydata.storage_client import StorageFarmBroker -from allmydata.storage.common import storage_index_to_dir from allmydata.scripts import debug from allmydata.mutable.filenode import MutableFileNode, BackoffAgent hunk ./src/allmydata/test/test_mutable.py 3669 # Now execute each assignment by writing the storage. for (share, servernum) in assignments: sharedata = base64.b64decode(self.sdmf_old_shares[share]) - storedir = self.get_serverdir(servernum) - storage_path = os.path.join(storedir, "shares", - storage_index_to_dir(si)) - fileutil.make_dirs(storage_path) - fileutil.write(os.path.join(storage_path, "%d" % share), - sharedata) + storage_dir = self.get_server(servernum).backend.get_shareset(si).sharehomedir + fileutil.fp_make_dirs(storage_dir) + storage_dir.child("%d" % share).setContent(sharedata) # ...and verify that the shares are there. shares = self.find_uri_shares(self.sdmf_old_cap) assert len(shares) == 10 hunk ./src/allmydata/test/test_provisioning.py 13 from nevow import inevow from zope.interface import implements -class MyRequest: +class MockRequest: implements(inevow.IRequest) pass hunk ./src/allmydata/test/test_provisioning.py 26 def test_load(self): pt = provisioning.ProvisioningTool() self.fields = {} - #r = MyRequest() + #r = MockRequest() #r.fields = self.fields #ctx = RequestContext() #unfilled = pt.renderSynchronously(ctx) hunk ./src/allmydata/test/test_repairer.py 537 # happiness setting. def _delete_some_servers(ignored): for i in xrange(7): - self.g.remove_server(self.g.servers_by_number[i].my_nodeid) + self.remove_server(i) assert len(self.g.servers_by_number) == 3 hunk ./src/allmydata/test/test_storage.py 14 from allmydata import interfaces from allmydata.util import fileutil, hashutil, base32, pollmixin, time_format from allmydata.storage.server import StorageServer -from allmydata.storage.mutable import MutableShareFile -from allmydata.storage.immutable import BucketWriter, BucketReader -from allmydata.storage.common import DataTooLargeError, storage_index_to_dir, \ +from allmydata.storage.backends.disk.mutable import MutableDiskShare +from allmydata.storage.bucket import BucketWriter, BucketReader +from allmydata.storage.common import DataTooLargeError, \ UnknownMutableContainerVersionError, UnknownImmutableContainerVersionError from allmydata.storage.lease import LeaseInfo from allmydata.storage.crawler import BucketCountingCrawler hunk ./src/allmydata/test/test_storage.py 474 w[0].remote_write(0, "\xff"*10) w[0].remote_close() - fn = os.path.join(ss.sharedir, storage_index_to_dir("si1"), "0") - f = open(fn, "rb+") + fp = ss.backend.get_shareset("si1").sharehomedir.child("0") + f = fp.open("rb+") f.seek(0) f.write(struct.pack(">L", 0)) # this is invalid: minimum used is v1 f.close() hunk ./src/allmydata/test/test_storage.py 814 def test_bad_magic(self): ss = self.create("test_bad_magic") self.allocate(ss, "si1", "we1", self._lease_secret.next(), set([0]), 10) - fn = os.path.join(ss.sharedir, storage_index_to_dir("si1"), "0") - f = open(fn, "rb+") + fp = ss.backend.get_shareset("si1").sharehomedir.child("0") + f = fp.open("rb+") f.seek(0) f.write("BAD MAGIC") f.close() hunk ./src/allmydata/test/test_storage.py 842 # Trying to make the container too large (by sending a write vector # whose offset is too high) will raise an exception. - TOOBIG = MutableShareFile.MAX_SIZE + 10 + TOOBIG = MutableDiskShare.MAX_SIZE + 10 self.failUnlessRaises(DataTooLargeError, rstaraw, "si1", secrets, {0: ([], [(TOOBIG,data)], None)}, hunk ./src/allmydata/test/test_storage.py 1229 # create a random non-numeric file in the bucket directory, to # exercise the code that's supposed to ignore those. - bucket_dir = os.path.join(self.workdir("test_leases"), - "shares", storage_index_to_dir("si1")) - f = open(os.path.join(bucket_dir, "ignore_me.txt"), "w") - f.write("you ought to be ignoring me\n") - f.close() + bucket_dir = ss.backend.get_shareset("si1").sharehomedir + bucket_dir.child("ignore_me.txt").setContent("you ought to be ignoring me\n") hunk ./src/allmydata/test/test_storage.py 1232 - s0 = MutableShareFile(os.path.join(bucket_dir, "0")) + s0 = MutableDiskShare(os.path.join(bucket_dir, "0")) self.failUnlessEqual(len(list(s0.get_leases())), 1) # add-lease on a missing storage index is silently ignored hunk ./src/allmydata/test/test_storage.py 3118 [immutable_si_0, immutable_si_1, mutable_si_2, mutable_si_3] = self.sis # add a non-sharefile to exercise another code path - fn = os.path.join(ss.sharedir, - storage_index_to_dir(immutable_si_0), - "not-a-share") - f = open(fn, "wb") - f.write("I am not a share.\n") - f.close() + fp = ss.backend.get_shareset(immutable_si_0).sharehomedir.child("not-a-share") + fp.setContent("I am not a share.\n") # this is before the crawl has started, so we're not in a cycle yet initial_state = lc.get_state() hunk ./src/allmydata/test/test_storage.py 3282 def test_expire_age(self): basedir = "storage/LeaseCrawler/expire_age" fileutil.make_dirs(basedir) - # setting expiration_time to 2000 means that any lease which is more - # than 2000s old will be expired. - ss = InstrumentedStorageServer(basedir, "\x00" * 20, - expiration_enabled=True, - expiration_mode="age", - expiration_override_lease_duration=2000) + # setting 'override_lease_duration' to 2000 means that any lease that + # is more than 2000 seconds old will be expired. + expiration_policy = { + 'enabled': True, + 'mode': 'age', + 'override_lease_duration': 2000, + 'sharetypes': ('mutable', 'immutable'), + } + ss = InstrumentedStorageServer(basedir, "\x00" * 20, expiration_policy) # make it start sooner than usual. lc = ss.lease_checker lc.slow_start = 0 hunk ./src/allmydata/test/test_storage.py 3423 def test_expire_cutoff_date(self): basedir = "storage/LeaseCrawler/expire_cutoff_date" fileutil.make_dirs(basedir) - # setting cutoff-date to 2000 seconds ago means that any lease which - # is more than 2000s old will be expired. + # setting 'cutoff_date' to 2000 seconds ago means that any lease that + # is more than 2000 seconds old will be expired. now = time.time() then = int(now - 2000) hunk ./src/allmydata/test/test_storage.py 3427 - ss = InstrumentedStorageServer(basedir, "\x00" * 20, - expiration_enabled=True, - expiration_mode="cutoff-date", - expiration_cutoff_date=then) + expiration_policy = { + 'enabled': True, + 'mode': 'cutoff-date', + 'cutoff_date': then, + 'sharetypes': ('mutable', 'immutable'), + } + ss = InstrumentedStorageServer(basedir, "\x00" * 20, expiration_policy) # make it start sooner than usual. lc = ss.lease_checker lc.slow_start = 0 hunk ./src/allmydata/test/test_storage.py 3575 def test_only_immutable(self): basedir = "storage/LeaseCrawler/only_immutable" fileutil.make_dirs(basedir) + # setting 'cutoff_date' to 2000 seconds ago means that any lease that + # is more than 2000 seconds old will be expired. now = time.time() then = int(now - 2000) hunk ./src/allmydata/test/test_storage.py 3579 - ss = StorageServer(basedir, "\x00" * 20, - expiration_enabled=True, - expiration_mode="cutoff-date", - expiration_cutoff_date=then, - expiration_sharetypes=("immutable",)) + expiration_policy = { + 'enabled': True, + 'mode': 'cutoff-date', + 'cutoff_date': then, + 'sharetypes': ('immutable',), + } + ss = StorageServer(basedir, "\x00" * 20, expiration_policy) lc = ss.lease_checker lc.slow_start = 0 webstatus = StorageStatus(ss) hunk ./src/allmydata/test/test_storage.py 3636 def test_only_mutable(self): basedir = "storage/LeaseCrawler/only_mutable" fileutil.make_dirs(basedir) + # setting 'cutoff_date' to 2000 seconds ago means that any lease that + # is more than 2000 seconds old will be expired. now = time.time() then = int(now - 2000) hunk ./src/allmydata/test/test_storage.py 3640 - ss = StorageServer(basedir, "\x00" * 20, - expiration_enabled=True, - expiration_mode="cutoff-date", - expiration_cutoff_date=then, - expiration_sharetypes=("mutable",)) + expiration_policy = { + 'enabled': True, + 'mode': 'cutoff-date', + 'cutoff_date': then, + 'sharetypes': ('mutable',), + } + ss = StorageServer(basedir, "\x00" * 20, expiration_policy) lc = ss.lease_checker lc.slow_start = 0 webstatus = StorageStatus(ss) hunk ./src/allmydata/test/test_storage.py 3819 def test_no_st_blocks(self): basedir = "storage/LeaseCrawler/no_st_blocks" fileutil.make_dirs(basedir) - ss = No_ST_BLOCKS_StorageServer(basedir, "\x00" * 20, - expiration_mode="age", - expiration_override_lease_duration=-1000) - # a negative expiration_time= means the "configured-" + # A negative 'override_lease_duration' means that the "configured-" # space-recovered counts will be non-zero, since all shares will have hunk ./src/allmydata/test/test_storage.py 3821 - # expired by then + # expired by then. + expiration_policy = { + 'enabled': True, + 'mode': 'age', + 'override_lease_duration': -1000, + 'sharetypes': ('mutable', 'immutable'), + } + ss = No_ST_BLOCKS_StorageServer(basedir, "\x00" * 20, expiration_policy) # make it start sooner than usual. lc = ss.lease_checker hunk ./src/allmydata/test/test_storage.py 3877 [immutable_si_0, immutable_si_1, mutable_si_2, mutable_si_3] = self.sis first = min(self.sis) first_b32 = base32.b2a(first) - fn = os.path.join(ss.sharedir, storage_index_to_dir(first), "0") - f = open(fn, "rb+") + fp = ss.backend.get_shareset(first).sharehomedir.child("0") + f = fp.open("rb+") f.seek(0) f.write("BAD MAGIC") f.close() hunk ./src/allmydata/test/test_storage.py 3890 # also create an empty bucket empty_si = base32.b2a("\x04"*16) - empty_bucket_dir = os.path.join(ss.sharedir, - storage_index_to_dir(empty_si)) - fileutil.make_dirs(empty_bucket_dir) + empty_bucket_dir = ss.backend.get_shareset(empty_si).sharehomedir + fileutil.fp_make_dirs(empty_bucket_dir) ss.setServiceParent(self.s) hunk ./src/allmydata/test/test_system.py 10 import allmydata from allmydata import uri -from allmydata.storage.mutable import MutableShareFile +from allmydata.storage.backends.disk.mutable import MutableDiskShare from allmydata.storage.server import si_a2b from allmydata.immutable import offloaded, upload from allmydata.immutable.literal import LiteralFileNode hunk ./src/allmydata/test/test_system.py 421 return shares def _corrupt_mutable_share(self, filename, which): - msf = MutableShareFile(filename) + msf = MutableDiskShare(filename) datav = msf.readv([ (0, 1000000) ]) final_share = datav[0] assert len(final_share) < 1000000 # ought to be truncated hunk ./src/allmydata/test/test_upload.py 22 from allmydata.util.happinessutil import servers_of_happiness, \ shares_by_server, merge_servers from allmydata.storage_client import StorageFarmBroker -from allmydata.storage.server import storage_index_to_dir MiB = 1024*1024 hunk ./src/allmydata/test/test_upload.py 821 def _copy_share_to_server(self, share_number, server_number): ss = self.g.servers_by_number[server_number] - # Copy share i from the directory associated with the first - # storage server to the directory associated with this one. - assert self.g, "I tried to find a grid at self.g, but failed" - assert self.shares, "I tried to find shares at self.shares, but failed" - old_share_location = self.shares[share_number][2] - new_share_location = os.path.join(ss.storedir, "shares") - si = uri.from_string(self.uri).get_storage_index() - new_share_location = os.path.join(new_share_location, - storage_index_to_dir(si)) - if not os.path.exists(new_share_location): - os.makedirs(new_share_location) - new_share_location = os.path.join(new_share_location, - str(share_number)) - if old_share_location != new_share_location: - shutil.copy(old_share_location, new_share_location) - shares = self.find_uri_shares(self.uri) - # Make sure that the storage server has the share. - self.failUnless((share_number, ss.my_nodeid, new_share_location) - in shares) + self.copy_share(self.shares[share_number], ss) def _setup_grid(self): """ hunk ./src/allmydata/test/test_upload.py 1103 self._copy_share_to_server(i, 2) d.addCallback(_copy_shares) # Remove the first server, and add a placeholder with share 0 - d.addCallback(lambda ign: - self.g.remove_server(self.g.servers_by_number[0].my_nodeid)) + d.addCallback(lambda ign: self.remove_server(0)) d.addCallback(lambda ign: self._add_server_with_share(server_number=4, share_number=0)) # Now try uploading. hunk ./src/allmydata/test/test_upload.py 1134 d.addCallback(lambda ign: self._add_server(server_number=4)) d.addCallback(_copy_shares) - d.addCallback(lambda ign: - self.g.remove_server(self.g.servers_by_number[0].my_nodeid)) + d.addCallback(lambda ign: self.remove_server(0)) d.addCallback(_reset_encoding_parameters) d.addCallback(lambda client: client.upload(upload.Data("data" * 10000, convergence=""))) hunk ./src/allmydata/test/test_upload.py 1196 self._copy_share_to_server(i, 2) d.addCallback(_copy_shares) # Remove server 0, and add another in its place - d.addCallback(lambda ign: - self.g.remove_server(self.g.servers_by_number[0].my_nodeid)) + d.addCallback(lambda ign: self.remove_server(0)) d.addCallback(lambda ign: self._add_server_with_share(server_number=4, share_number=0, readonly=True)) hunk ./src/allmydata/test/test_upload.py 1237 for i in xrange(1, 10): self._copy_share_to_server(i, 2) d.addCallback(_copy_shares) - d.addCallback(lambda ign: - self.g.remove_server(self.g.servers_by_number[0].my_nodeid)) + d.addCallback(lambda ign: self.remove_server(0)) def _reset_encoding_parameters(ign, happy=4): client = self.g.clients[0] client.DEFAULT_ENCODING_PARAMETERS['happy'] = happy hunk ./src/allmydata/test/test_upload.py 1273 # remove the original server # (necessary to ensure that the Tahoe2ServerSelector will distribute # all the shares) - def _remove_server(ign): - server = self.g.servers_by_number[0] - self.g.remove_server(server.my_nodeid) - d.addCallback(_remove_server) + d.addCallback(lambda ign: self.remove_server(0)) # This should succeed; we still have 4 servers, and the # happiness of the upload is 4. d.addCallback(lambda ign: hunk ./src/allmydata/test/test_upload.py 1285 d.addCallback(lambda ign: self._setup_and_upload()) d.addCallback(_do_server_setup) - d.addCallback(_remove_server) + d.addCallback(lambda ign: self.remove_server(0)) d.addCallback(lambda ign: self.shouldFail(UploadUnhappinessError, "test_dropped_servers_in_encoder", hunk ./src/allmydata/test/test_upload.py 1307 self._add_server_with_share(4, 7, readonly=True) self._add_server_with_share(5, 8, readonly=True) d.addCallback(_do_server_setup_2) - d.addCallback(_remove_server) + d.addCallback(lambda ign: self.remove_server(0)) d.addCallback(lambda ign: self._do_upload_with_broken_servers(1)) d.addCallback(_set_basedir) hunk ./src/allmydata/test/test_upload.py 1314 d.addCallback(lambda ign: self._setup_and_upload()) d.addCallback(_do_server_setup_2) - d.addCallback(_remove_server) + d.addCallback(lambda ign: self.remove_server(0)) d.addCallback(lambda ign: self.shouldFail(UploadUnhappinessError, "test_dropped_servers_in_encoder", hunk ./src/allmydata/test/test_upload.py 1528 for i in xrange(1, 10): self._copy_share_to_server(i, 1) d.addCallback(_copy_shares) - d.addCallback(lambda ign: - self.g.remove_server(self.g.servers_by_number[0].my_nodeid)) + d.addCallback(lambda ign: self.remove_server(0)) def _prepare_client(ign): client = self.g.clients[0] client.DEFAULT_ENCODING_PARAMETERS['happy'] = 4 hunk ./src/allmydata/test/test_upload.py 1550 def _setup(ign): for i in xrange(1, 11): self._add_server(server_number=i) - self.g.remove_server(self.g.servers_by_number[0].my_nodeid) + self.remove_server(0) c = self.g.clients[0] # We set happy to an unsatisfiable value so that we can check the # counting in the exception message. The same progress message hunk ./src/allmydata/test/test_upload.py 1577 self._add_server(server_number=i) self._add_server(server_number=11, readonly=True) self._add_server(server_number=12, readonly=True) - self.g.remove_server(self.g.servers_by_number[0].my_nodeid) + self.remove_server(0) c = self.g.clients[0] c.DEFAULT_ENCODING_PARAMETERS['happy'] = 45 return c hunk ./src/allmydata/test/test_upload.py 1605 # the first one that the selector sees. for i in xrange(10): self._copy_share_to_server(i, 9) - # Remove server 0, and its contents - self.g.remove_server(self.g.servers_by_number[0].my_nodeid) + self.remove_server(0) # Make happiness unsatisfiable c = self.g.clients[0] c.DEFAULT_ENCODING_PARAMETERS['happy'] = 45 hunk ./src/allmydata/test/test_upload.py 1625 def _then(ign): for i in xrange(1, 11): self._add_server(server_number=i, readonly=True) - self.g.remove_server(self.g.servers_by_number[0].my_nodeid) + self.remove_server(0) c = self.g.clients[0] c.DEFAULT_ENCODING_PARAMETERS['k'] = 2 c.DEFAULT_ENCODING_PARAMETERS['happy'] = 4 hunk ./src/allmydata/test/test_upload.py 1661 self._add_server(server_number=4, readonly=True)) d.addCallback(lambda ign: self._add_server(server_number=5, readonly=True)) - d.addCallback(lambda ign: - self.g.remove_server(self.g.servers_by_number[0].my_nodeid)) + d.addCallback(lambda ign: self.remove_server(0)) def _reset_encoding_parameters(ign, happy=4): client = self.g.clients[0] client.DEFAULT_ENCODING_PARAMETERS['happy'] = happy hunk ./src/allmydata/test/test_upload.py 1696 d.addCallback(lambda ign: self._add_server(server_number=2)) def _break_server_2(ign): - serverid = self.g.servers_by_number[2].my_nodeid + serverid = self.get_server(2).get_serverid() self.g.break_server(serverid) d.addCallback(_break_server_2) d.addCallback(lambda ign: hunk ./src/allmydata/test/test_upload.py 1705 self._add_server(server_number=4, readonly=True)) d.addCallback(lambda ign: self._add_server(server_number=5, readonly=True)) - d.addCallback(lambda ign: - self.g.remove_server(self.g.servers_by_number[0].my_nodeid)) + d.addCallback(lambda ign: self.remove_server(0)) d.addCallback(_reset_encoding_parameters) d.addCallback(lambda client: self.shouldFail(UploadUnhappinessError, "test_selection_exceptions", hunk ./src/allmydata/test/test_upload.py 1816 # Copy shares self._copy_share_to_server(1, 1) self._copy_share_to_server(2, 1) - # Remove server 0 - self.g.remove_server(self.g.servers_by_number[0].my_nodeid) + self.remove_server(0) client = self.g.clients[0] client.DEFAULT_ENCODING_PARAMETERS['happy'] = 3 return client hunk ./src/allmydata/test/test_upload.py 1930 readonly=True) self._add_server_with_share(server_number=4, share_number=3, readonly=True) - # Remove server 0. - self.g.remove_server(self.g.servers_by_number[0].my_nodeid) + self.remove_server(0) # Set the client appropriately c = self.g.clients[0] c.DEFAULT_ENCODING_PARAMETERS['happy'] = 4 hunk ./src/allmydata/test/test_util.py 9 from twisted.trial import unittest from twisted.internet import defer, reactor from twisted.python.failure import Failure +from twisted.python.filepath import FilePath from twisted.python import log from pycryptopp.hash.sha256 import SHA256 as _hash hunk ./src/allmydata/test/test_util.py 508 os.chdir(saved_cwd) def test_disk_stats(self): - avail = fileutil.get_available_space('.', 2**14) + avail = fileutil.get_available_space(FilePath('.'), 2**14) if avail == 0: raise unittest.SkipTest("This test will spuriously fail there is no disk space left.") hunk ./src/allmydata/test/test_util.py 512 - disk = fileutil.get_disk_stats('.', 2**13) + disk = fileutil.get_disk_stats(FilePath('.'), 2**13) self.failUnless(disk['total'] > 0, disk['total']) self.failUnless(disk['used'] > 0, disk['used']) self.failUnless(disk['free_for_root'] > 0, disk['free_for_root']) hunk ./src/allmydata/test/test_util.py 521 def test_disk_stats_avail_nonnegative(self): # This test will spuriously fail if you have more than 2^128 - # bytes of available space on your filesystem. - disk = fileutil.get_disk_stats('.', 2**128) + # bytes of available space on your filesystem (lucky you). + disk = fileutil.get_disk_stats(FilePath('.'), 2**128) self.failUnlessEqual(disk['avail'], 0) class PollMixinTests(unittest.TestCase): hunk ./src/allmydata/test/test_web.py 12 from twisted.python import failure, log from nevow import rend from allmydata import interfaces, uri, webish, dirnode -from allmydata.storage.shares import get_share_file from allmydata.storage_client import StorageFarmBroker from allmydata.immutable import upload from allmydata.immutable.downloader.status import DownloadStatus hunk ./src/allmydata/test/test_web.py 4111 good_shares = self.find_uri_shares(self.uris["good"]) self.failUnlessReallyEqual(len(good_shares), 10) sick_shares = self.find_uri_shares(self.uris["sick"]) - os.unlink(sick_shares[0][2]) + sick_shares[0][2].remove() dead_shares = self.find_uri_shares(self.uris["dead"]) for i in range(1, 10): hunk ./src/allmydata/test/test_web.py 4114 - os.unlink(dead_shares[i][2]) + dead_shares[i][2].remove() c_shares = self.find_uri_shares(self.uris["corrupt"]) cso = CorruptShareOptions() cso.stdout = StringIO() hunk ./src/allmydata/test/test_web.py 4118 - cso.parseOptions([c_shares[0][2]]) + cso.parseOptions([c_shares[0][2].path]) corrupt_share(cso) d.addCallback(_clobber_shares) hunk ./src/allmydata/test/test_web.py 4253 good_shares = self.find_uri_shares(self.uris["good"]) self.failUnlessReallyEqual(len(good_shares), 10) sick_shares = self.find_uri_shares(self.uris["sick"]) - os.unlink(sick_shares[0][2]) + sick_shares[0][2].remove() dead_shares = self.find_uri_shares(self.uris["dead"]) for i in range(1, 10): hunk ./src/allmydata/test/test_web.py 4256 - os.unlink(dead_shares[i][2]) + dead_shares[i][2].remove() c_shares = self.find_uri_shares(self.uris["corrupt"]) cso = CorruptShareOptions() cso.stdout = StringIO() hunk ./src/allmydata/test/test_web.py 4260 - cso.parseOptions([c_shares[0][2]]) + cso.parseOptions([c_shares[0][2].path]) corrupt_share(cso) d.addCallback(_clobber_shares) hunk ./src/allmydata/test/test_web.py 4319 def _clobber_shares(ignored): sick_shares = self.find_uri_shares(self.uris["sick"]) - os.unlink(sick_shares[0][2]) + sick_shares[0][2].remove() d.addCallback(_clobber_shares) d.addCallback(self.CHECK, "sick", "t=check&repair=true&output=json") hunk ./src/allmydata/test/test_web.py 4811 good_shares = self.find_uri_shares(self.uris["good"]) self.failUnlessReallyEqual(len(good_shares), 10) sick_shares = self.find_uri_shares(self.uris["sick"]) - os.unlink(sick_shares[0][2]) + sick_shares[0][2].remove() #dead_shares = self.find_uri_shares(self.uris["dead"]) #for i in range(1, 10): hunk ./src/allmydata/test/test_web.py 4814 - # os.unlink(dead_shares[i][2]) + # dead_shares[i][2].remove() #c_shares = self.find_uri_shares(self.uris["corrupt"]) #cso = CorruptShareOptions() hunk ./src/allmydata/test/test_web.py 4819 #cso.stdout = StringIO() - #cso.parseOptions([c_shares[0][2]]) + #cso.parseOptions([c_shares[0][2].path]) #corrupt_share(cso) d.addCallback(_clobber_shares) hunk ./src/allmydata/test/test_web.py 4870 d.addErrback(self.explain_web_error) return d - def _count_leases(self, ignored, which): - u = self.uris[which] - shares = self.find_uri_shares(u) - lease_counts = [] - for shnum, serverid, fn in shares: - sf = get_share_file(fn) - num_leases = len(list(sf.get_leases())) - lease_counts.append( (fn, num_leases) ) - return lease_counts - - def _assert_leasecount(self, lease_counts, expected): + def _assert_leasecount(self, ignored, which, expected): + lease_counts = self.count_leases(self.uris[which]) for (fn, num_leases) in lease_counts: if num_leases != expected: self.fail("expected %d leases, have %d, on %s" % hunk ./src/allmydata/test/test_web.py 4903 self.fileurls[which] = "uri/" + urllib.quote(self.uris[which]) d.addCallback(_compute_fileurls) - d.addCallback(self._count_leases, "one") - d.addCallback(self._assert_leasecount, 1) - d.addCallback(self._count_leases, "two") - d.addCallback(self._assert_leasecount, 1) - d.addCallback(self._count_leases, "mutable") - d.addCallback(self._assert_leasecount, 1) + d.addCallback(self._assert_leasecount, "one", 1) + d.addCallback(self._assert_leasecount, "two", 1) + d.addCallback(self._assert_leasecount, "mutable", 1) d.addCallback(self.CHECK, "one", "t=check") # no add-lease def _got_html_good(res): hunk ./src/allmydata/test/test_web.py 4913 self.failIf("Not Healthy" in res, res) d.addCallback(_got_html_good) - d.addCallback(self._count_leases, "one") - d.addCallback(self._assert_leasecount, 1) - d.addCallback(self._count_leases, "two") - d.addCallback(self._assert_leasecount, 1) - d.addCallback(self._count_leases, "mutable") - d.addCallback(self._assert_leasecount, 1) + d.addCallback(self._assert_leasecount, "one", 1) + d.addCallback(self._assert_leasecount, "two", 1) + d.addCallback(self._assert_leasecount, "mutable", 1) # this CHECK uses the original client, which uses the same # lease-secrets, so it will just renew the original lease hunk ./src/allmydata/test/test_web.py 4922 d.addCallback(self.CHECK, "one", "t=check&add-lease=true") d.addCallback(_got_html_good) - d.addCallback(self._count_leases, "one") - d.addCallback(self._assert_leasecount, 1) - d.addCallback(self._count_leases, "two") - d.addCallback(self._assert_leasecount, 1) - d.addCallback(self._count_leases, "mutable") - d.addCallback(self._assert_leasecount, 1) + d.addCallback(self._assert_leasecount, "one", 1) + d.addCallback(self._assert_leasecount, "two", 1) + d.addCallback(self._assert_leasecount, "mutable", 1) # this CHECK uses an alternate client, which adds a second lease d.addCallback(self.CHECK, "one", "t=check&add-lease=true", clientnum=1) hunk ./src/allmydata/test/test_web.py 4930 d.addCallback(_got_html_good) - d.addCallback(self._count_leases, "one") - d.addCallback(self._assert_leasecount, 2) - d.addCallback(self._count_leases, "two") - d.addCallback(self._assert_leasecount, 1) - d.addCallback(self._count_leases, "mutable") - d.addCallback(self._assert_leasecount, 1) + d.addCallback(self._assert_leasecount, "one", 2) + d.addCallback(self._assert_leasecount, "two", 1) + d.addCallback(self._assert_leasecount, "mutable", 1) d.addCallback(self.CHECK, "mutable", "t=check&add-lease=true") d.addCallback(_got_html_good) hunk ./src/allmydata/test/test_web.py 4937 - d.addCallback(self._count_leases, "one") - d.addCallback(self._assert_leasecount, 2) - d.addCallback(self._count_leases, "two") - d.addCallback(self._assert_leasecount, 1) - d.addCallback(self._count_leases, "mutable") - d.addCallback(self._assert_leasecount, 1) + d.addCallback(self._assert_leasecount, "one", 2) + d.addCallback(self._assert_leasecount, "two", 1) + d.addCallback(self._assert_leasecount, "mutable", 1) d.addCallback(self.CHECK, "mutable", "t=check&add-lease=true", clientnum=1) hunk ./src/allmydata/test/test_web.py 4945 d.addCallback(_got_html_good) - d.addCallback(self._count_leases, "one") - d.addCallback(self._assert_leasecount, 2) - d.addCallback(self._count_leases, "two") - d.addCallback(self._assert_leasecount, 1) - d.addCallback(self._count_leases, "mutable") - d.addCallback(self._assert_leasecount, 2) + d.addCallback(self._assert_leasecount, "one", 2) + d.addCallback(self._assert_leasecount, "two", 1) + d.addCallback(self._assert_leasecount, "mutable", 2) d.addErrback(self.explain_web_error) return d hunk ./src/allmydata/test/test_web.py 4989 self.failUnlessReallyEqual(len(units), 4+1) d.addCallback(_done) - d.addCallback(self._count_leases, "root") - d.addCallback(self._assert_leasecount, 1) - d.addCallback(self._count_leases, "one") - d.addCallback(self._assert_leasecount, 1) - d.addCallback(self._count_leases, "mutable") - d.addCallback(self._assert_leasecount, 1) + d.addCallback(self._assert_leasecount, "root", 1) + d.addCallback(self._assert_leasecount, "one", 1) + d.addCallback(self._assert_leasecount, "mutable", 1) d.addCallback(self.CHECK, "root", "t=stream-deep-check&add-lease=true") d.addCallback(_done) hunk ./src/allmydata/test/test_web.py 4996 - d.addCallback(self._count_leases, "root") - d.addCallback(self._assert_leasecount, 1) - d.addCallback(self._count_leases, "one") - d.addCallback(self._assert_leasecount, 1) - d.addCallback(self._count_leases, "mutable") - d.addCallback(self._assert_leasecount, 1) + d.addCallback(self._assert_leasecount, "root", 1) + d.addCallback(self._assert_leasecount, "one", 1) + d.addCallback(self._assert_leasecount, "mutable", 1) d.addCallback(self.CHECK, "root", "t=stream-deep-check&add-lease=true", clientnum=1) hunk ./src/allmydata/test/test_web.py 5004 d.addCallback(_done) - d.addCallback(self._count_leases, "root") - d.addCallback(self._assert_leasecount, 2) - d.addCallback(self._count_leases, "one") - d.addCallback(self._assert_leasecount, 2) - d.addCallback(self._count_leases, "mutable") - d.addCallback(self._assert_leasecount, 2) + d.addCallback(self._assert_leasecount, "root", 2) + d.addCallback(self._assert_leasecount, "one", 2) + d.addCallback(self._assert_leasecount, "mutable", 2) d.addErrback(self.explain_web_error) return d merger 0.0 ( hunk ./src/allmydata/uri.py 829 + def is_readonly(self): + return True + + def get_readonly(self): + return self + + hunk ./src/allmydata/uri.py 829 + def is_readonly(self): + return True + + def get_readonly(self): + return self + + ) merger 0.0 ( hunk ./src/allmydata/uri.py 848 + def is_readonly(self): + return True + + def get_readonly(self): + return self + hunk ./src/allmydata/uri.py 848 + def is_readonly(self): + return True + + def get_readonly(self): + return self + ) hunk ./src/allmydata/util/encodingutil.py 221 def quote_path(path, quotemarks=True): return quote_output("/".join(map(to_str, path)), quotemarks=quotemarks) +def quote_filepath(fp, quotemarks=True, encoding=None): + path = fp.path + if isinstance(path, str): + try: + path = path.decode(filesystem_encoding) + except UnicodeDecodeError: + return 'b"%s"' % (ESCAPABLE_8BIT.sub(_str_escape, path),) + + return quote_output(path, quotemarks=quotemarks, encoding=encoding) + def unicode_platform(): """ hunk ./src/allmydata/util/fileutil.py 5 Futz with files like a pro. """ -import sys, exceptions, os, stat, tempfile, time, binascii +import errno, sys, exceptions, os, stat, tempfile, time, binascii + +from allmydata.util.assertutil import precondition from twisted.python import log hunk ./src/allmydata/util/fileutil.py 10 +from twisted.python.filepath import FilePath, UnlistableError from pycryptopp.cipher.aes import AES hunk ./src/allmydata/util/fileutil.py 189 raise tx raise exceptions.IOError, "unknown error prevented creation of directory, or deleted the directory immediately after creation: %s" % dirname # careful not to construct an IOError with a 2-tuple, as that has a special meaning... -def rm_dir(dirname): +def fp_make_dirs(dirfp): + """ + An idempotent version of FilePath.makedirs(). If the dir already + exists, do nothing and return without raising an exception. If this + call creates the dir, return without raising an exception. If there is + an error that prevents creation or if the directory gets deleted after + fp_make_dirs() creates it and before fp_make_dirs() checks that it + exists, raise an exception. + """ + log.msg( "xxx 0 %s" % (dirfp,)) + tx = None + try: + dirfp.makedirs() + except OSError, x: + tx = x + + if not dirfp.isdir(): + if tx: + raise tx + raise exceptions.IOError, "unknown error prevented creation of directory, or deleted the directory immediately after creation: %s" % dirfp # careful not to construct an IOError with a 2-tuple, as that has a special meaning... + +def fp_rmdir_if_empty(dirfp): + """ Remove the directory if it is empty. """ + try: + os.rmdir(dirfp.path) + except OSError, e: + if e.errno != errno.ENOTEMPTY: + raise + else: + dirfp.changed() + +def rmtree(dirname): """ A threadsafe and idempotent version of shutil.rmtree(). If the dir is already gone, do nothing and return without raising an exception. If this hunk ./src/allmydata/util/fileutil.py 239 else: remove(fullname) os.rmdir(dirname) - except Exception, le: - # Ignore "No such file or directory" - if (not isinstance(le, OSError)) or le.args[0] != 2: + except EnvironmentError, le: + # Ignore "No such file or directory", collect any other exception. + if (le.args[0] != 2 and le.args[0] != 3) or (le.args[0] != errno.ENOENT): excs.append(le) hunk ./src/allmydata/util/fileutil.py 243 + except Exception, le: + excs.append(le) # Okay, now we've recursively removed everything, ignoring any "No # such file or directory" errors, and collecting any other errors. hunk ./src/allmydata/util/fileutil.py 256 raise OSError, "Failed to remove dir for unknown reason." raise OSError, excs +def fp_remove(fp): + """ + An idempotent version of shutil.rmtree(). If the file/dir is already + gone, do nothing and return without raising an exception. If this call + removes the file/dir, return without raising an exception. If there is + an error that prevents removal, or if a file or directory at the same + path gets created again by someone else after this deletes it and before + this checks that it is gone, raise an exception. + """ + try: + fp.remove() + except UnlistableError, e: + if e.originalException.errno != errno.ENOENT: + raise + except OSError, e: + if e.errno != errno.ENOENT: + raise + +def rm_dir(dirname): + # Renamed to be like shutil.rmtree and unlike rmdir. + return rmtree(dirname) def remove_if_possible(f): try: hunk ./src/allmydata/util/fileutil.py 387 import traceback traceback.print_exc() -def get_disk_stats(whichdir, reserved_space=0): +def get_disk_stats(whichdirfp, reserved_space=0): """Return disk statistics for the storage disk, in the form of a dict with the following fields. total: total bytes on disk hunk ./src/allmydata/util/fileutil.py 408 you can pass how many bytes you would like to leave unused on this filesystem as reserved_space. """ + precondition(isinstance(whichdirfp, FilePath), whichdirfp) if have_GetDiskFreeSpaceExW: # If this is a Windows system and GetDiskFreeSpaceExW is available, use it. hunk ./src/allmydata/util/fileutil.py 419 n_free_for_nonroot = c_ulonglong(0) n_total = c_ulonglong(0) n_free_for_root = c_ulonglong(0) - retval = GetDiskFreeSpaceExW(whichdir, byref(n_free_for_nonroot), - byref(n_total), - byref(n_free_for_root)) + retval = GetDiskFreeSpaceExW(whichdirfp.path, byref(n_free_for_nonroot), + byref(n_total), + byref(n_free_for_root)) if retval == 0: raise OSError("Windows error %d attempting to get disk statistics for %r" hunk ./src/allmydata/util/fileutil.py 424 - % (GetLastError(), whichdir)) + % (GetLastError(), whichdirfp.path)) free_for_nonroot = n_free_for_nonroot.value total = n_total.value free_for_root = n_free_for_root.value hunk ./src/allmydata/util/fileutil.py 433 # # # - s = os.statvfs(whichdir) + s = os.statvfs(whichdirfp.path) # on my mac laptop: # statvfs(2) is a wrapper around statfs(2). hunk ./src/allmydata/util/fileutil.py 460 'avail': avail, } -def get_available_space(whichdir, reserved_space): +def get_available_space(whichdirfp, reserved_space): """Returns available space for share storage in bytes, or None if no API to get this information is available. hunk ./src/allmydata/util/fileutil.py 472 you can pass how many bytes you would like to leave unused on this filesystem as reserved_space. """ + precondition(isinstance(whichdirfp, FilePath), whichdirfp) try: hunk ./src/allmydata/util/fileutil.py 474 - return get_disk_stats(whichdir, reserved_space)['avail'] + return get_disk_stats(whichdirfp, reserved_space)['avail'] except AttributeError: return None hunk ./src/allmydata/util/fileutil.py 477 - except EnvironmentError: - log.msg("OS call to get disk statistics failed") + + +def get_used_space(fp): + if fp is None: return 0 hunk ./src/allmydata/util/fileutil.py 482 + try: + s = os.stat(fp.path) + except EnvironmentError: + if not fp.exists(): + return 0 + raise + else: + # POSIX defines st_blocks (originally a BSDism): + # + # but does not require stat() to give it a "meaningful value" + # + # and says: + # "The unit for the st_blocks member of the stat structure is not defined + # within IEEE Std 1003.1-2001. In some implementations it is 512 bytes. + # It may differ on a file system basis. There is no correlation between + # values of the st_blocks and st_blksize, and the f_bsize (from ) + # structure members." + # + # The Linux docs define it as "the number of blocks allocated to the file, + # [in] 512-byte units." It is also defined that way on MacOS X. Python does + # not set the attribute on Windows. + # + # We consider platforms that define st_blocks but give it a wrong value, or + # measure it in a unit other than 512 bytes, to be broken. See also + # . + + if hasattr(s, 'st_blocks'): + return s.st_blocks * 512 + else: + return s.st_size } [Work-in-progress, includes fix to bug involving BucketWriter. refs #999 david-sarah@jacaranda.org**20110920033803 Ignore-this: 64e9e019421454e4d08141d10b6e4eed ] { hunk ./src/allmydata/client.py 9 from twisted.internet import reactor, defer from twisted.application import service from twisted.application.internet import TimerService +from twisted.python.filepath import FilePath from foolscap.api import Referenceable from pycryptopp.publickey import rsa hunk ./src/allmydata/client.py 15 import allmydata from allmydata.storage.server import StorageServer +from allmydata.storage.backends.disk.disk_backend import DiskBackend from allmydata import storage_client from allmydata.immutable.upload import Uploader from allmydata.immutable.offloaded import Helper hunk ./src/allmydata/client.py 213 return readonly = self.get_config("storage", "readonly", False, boolean=True) - storedir = os.path.join(self.basedir, self.STOREDIR) + storedir = FilePath(self.basedir).child(self.STOREDIR) data = self.get_config("storage", "reserved_space", None) reserved = None hunk ./src/allmydata/client.py 255 'cutoff_date': cutoff_date, 'sharetypes': tuple(sharetypes), } - ss = StorageServer(storedir, self.nodeid, - reserved_space=reserved, - discard_storage=discard, - readonly_storage=readonly, + + backend = DiskBackend(storedir, readonly=readonly, reserved_space=reserved, + discard_storage=discard) + ss = StorageServer(nodeid, backend, storedir, stats_provider=self.stats_provider, expiration_policy=expiration_policy) self.add_service(ss) hunk ./src/allmydata/interfaces.py 348 def get_shares(): """ - Generates the IStoredShare objects held in this shareset. + Generates IStoredShare objects for all completed shares in this shareset. """ def has_incoming(shnum): hunk ./src/allmydata/storage/backends/base.py 69 # def _create_mutable_share(self, storageserver, shnum, write_enabler): # """create a mutable share with the given shnum and write_enabler""" - # secrets might be a triple with cancel_secret in secrets[2], but if - # so we ignore the cancel_secret. write_enabler = secrets[0] renew_secret = secrets[1] hunk ./src/allmydata/storage/backends/base.py 71 + cancel_secret = '\x00'*32 + if len(secrets) > 2: + cancel_secret = secrets[2] si_s = self.get_storage_index_string() shares = {} hunk ./src/allmydata/storage/backends/base.py 110 read_data[shnum] = share.readv(read_vector) ownerid = 1 # TODO - lease_info = LeaseInfo(ownerid, renew_secret, + lease_info = LeaseInfo(ownerid, renew_secret, cancel_secret, expiration_time, storageserver.get_serverid()) if testv_is_good: hunk ./src/allmydata/storage/backends/disk/disk_backend.py 34 return newfp.child(sia) -def get_share(fp): +def get_share(storageindex, shnum, fp): f = fp.open('rb') try: prefix = f.read(32) hunk ./src/allmydata/storage/backends/disk/disk_backend.py 42 f.close() if prefix == MutableDiskShare.MAGIC: - return MutableDiskShare(fp) + return MutableDiskShare(storageindex, shnum, fp) else: # assume it's immutable hunk ./src/allmydata/storage/backends/disk/disk_backend.py 45 - return ImmutableDiskShare(fp) + return ImmutableDiskShare(storageindex, shnum, fp) class DiskBackend(Backend): hunk ./src/allmydata/storage/backends/disk/disk_backend.py 174 if not NUM_RE.match(shnumstr): continue sharehome = self._sharehomedir.child(shnumstr) - yield self.get_share(sharehome) + yield get_share(self.get_storage_index(), int(shnumstr), sharehome) except UnlistableError: # There is no shares directory at all. pass hunk ./src/allmydata/storage/backends/disk/disk_backend.py 185 return self._incominghomedir.child(str(shnum)).exists() def make_bucket_writer(self, storageserver, shnum, max_space_per_bucket, lease_info, canary): - sharehome = self._sharehomedir.child(str(shnum)) + finalhome = self._sharehomedir.child(str(shnum)) incominghome = self._incominghomedir.child(str(shnum)) hunk ./src/allmydata/storage/backends/disk/disk_backend.py 187 - immsh = ImmutableDiskShare(self.get_storage_index(), shnum, sharehome, incominghome, - max_size=max_space_per_bucket, create=True) + immsh = ImmutableDiskShare(self.get_storage_index(), shnum, incominghome, finalhome, + max_size=max_space_per_bucket) bw = BucketWriter(storageserver, immsh, max_space_per_bucket, lease_info, canary) if self._discard_storage: bw.throw_out_all_data = True hunk ./src/allmydata/storage/backends/disk/disk_backend.py 198 fileutil.fp_make_dirs(self._sharehomedir) sharehome = self._sharehomedir.child(str(shnum)) serverid = storageserver.get_serverid() - return create_mutable_disk_share(sharehome, serverid, write_enabler, storageserver) + return create_mutable_disk_share(self.get_storage_index(), shnum, sharehome, serverid, write_enabler, storageserver) def _clean_up_after_unlink(self): fileutil.fp_rmdir_if_empty(self._sharehomedir) hunk ./src/allmydata/storage/backends/disk/immutable.py 48 LEASE_SIZE = struct.calcsize(">L32s32sL") - def __init__(self, storageindex, shnum, finalhome=None, incominghome=None, max_size=None, create=False): - """ If max_size is not None then I won't allow more than - max_size to be written to me. If create=True then max_size - must not be None. """ - precondition((max_size is not None) or (not create), max_size, create) + def __init__(self, storageindex, shnum, home, finalhome=None, max_size=None): + """ + If max_size is not None then I won't allow more than max_size to be written to me. + If finalhome is not None (meaning that we are creating the share) then max_size + must not be None. + """ + precondition((max_size is not None) or (finalhome is None), max_size, finalhome) self._storageindex = storageindex self._max_size = max_size hunk ./src/allmydata/storage/backends/disk/immutable.py 57 - self._incominghome = incominghome - self._home = finalhome + + # If we are creating the share, _finalhome refers to the final path and + # _home to the incoming path. Otherwise, _finalhome is None. + self._finalhome = finalhome + self._home = home self._shnum = shnum hunk ./src/allmydata/storage/backends/disk/immutable.py 63 - if create: - # touch the file, so later callers will see that we're working on + + if self._finalhome is not None: + # Touch the file, so later callers will see that we're working on # it. Also construct the metadata. hunk ./src/allmydata/storage/backends/disk/immutable.py 67 - assert not finalhome.exists() - fp_make_dirs(self._incominghome.parent()) + assert not self._finalhome.exists() + fp_make_dirs(self._home.parent()) # The second field -- the four-byte share data length -- is no # longer used as of Tahoe v1.3.0, but we continue to write it in # there in case someone downgrades a storage server from >= hunk ./src/allmydata/storage/backends/disk/immutable.py 78 # the largest length that can fit into the field. That way, even # if this does happen, the old < v1.3.0 server will still allow # clients to read the first part of the share. - self._incominghome.setContent(struct.pack(">LLL", 1, min(2**32-1, max_size), 0) ) + self._home.setContent(struct.pack(">LLL", 1, min(2**32-1, max_size), 0) ) self._lease_offset = max_size + 0x0c self._num_leases = 0 else: hunk ./src/allmydata/storage/backends/disk/immutable.py 101 % (si_b2a(self._storageindex), self._shnum, quote_filepath(self._home))) def close(self): - fileutil.fp_make_dirs(self._home.parent()) - self._incominghome.moveTo(self._home) - try: - # self._incominghome is like storage/shares/incoming/ab/abcde/4 . - # We try to delete the parent (.../ab/abcde) to avoid leaving - # these directories lying around forever, but the delete might - # fail if we're working on another share for the same storage - # index (like ab/abcde/5). The alternative approach would be to - # use a hierarchy of objects (PrefixHolder, BucketHolder, - # ShareWriter), each of which is responsible for a single - # directory on disk, and have them use reference counting of - # their children to know when they should do the rmdir. This - # approach is simpler, but relies on os.rmdir refusing to delete - # a non-empty directory. Do *not* use fileutil.fp_remove() here! - fileutil.fp_rmdir_if_empty(self._incominghome.parent()) - # we also delete the grandparent (prefix) directory, .../ab , - # again to avoid leaving directories lying around. This might - # fail if there is another bucket open that shares a prefix (like - # ab/abfff). - fileutil.fp_rmdir_if_empty(self._incominghome.parent().parent()) - # we leave the great-grandparent (incoming/) directory in place. - except EnvironmentError: - # ignore the "can't rmdir because the directory is not empty" - # exceptions, those are normal consequences of the - # above-mentioned conditions. - pass - pass + fileutil.fp_make_dirs(self._finalhome.parent()) + self._home.moveTo(self._finalhome) + + # self._home is like storage/shares/incoming/ab/abcde/4 . + # We try to delete the parent (.../ab/abcde) to avoid leaving + # these directories lying around forever, but the delete might + # fail if we're working on another share for the same storage + # index (like ab/abcde/5). The alternative approach would be to + # use a hierarchy of objects (PrefixHolder, BucketHolder, + # ShareWriter), each of which is responsible for a single + # directory on disk, and have them use reference counting of + # their children to know when they should do the rmdir. This + # approach is simpler, but relies on os.rmdir (used by + # fp_rmdir_if_empty) refusing to delete a non-empty directory. + # Do *not* use fileutil.fp_remove() here! + parent = self._home.parent() + fileutil.fp_rmdir_if_empty(parent) + + # we also delete the grandparent (prefix) directory, .../ab , + # again to avoid leaving directories lying around. This might + # fail if there is another bucket open that shares a prefix (like + # ab/abfff). + fileutil.fp_rmdir_if_empty(parent.parent()) + + # we leave the great-grandparent (incoming/) directory in place. + + # allow lease changes after closing. + self._home = self._finalhome + self._finalhome = None def get_used_space(self): hunk ./src/allmydata/storage/backends/disk/immutable.py 132 - return (fileutil.get_used_space(self._home) + - fileutil.get_used_space(self._incominghome)) + return (fileutil.get_used_space(self._finalhome) + + fileutil.get_used_space(self._home)) def get_storage_index(self): return self._storageindex hunk ./src/allmydata/storage/backends/disk/immutable.py 175 precondition(offset >= 0, offset) if self._max_size is not None and offset+length > self._max_size: raise DataTooLargeError(self._max_size, offset, length) - f = self._incominghome.open(mode='rb+') + f = self._home.open(mode='rb+') try: real_offset = self._data_offset+offset f.seek(real_offset) hunk ./src/allmydata/storage/backends/disk/immutable.py 205 # These lease operations are intended for use by disk_backend.py. # Other clients should not depend on the fact that the disk backend - # stores leases in share files. + # stores leases in share files. XXX bucket.py also relies on this. def get_leases(self): """Yields a LeaseInfo instance for all leases.""" hunk ./src/allmydata/storage/backends/disk/immutable.py 221 f.close() def add_lease(self, lease_info): - f = self._incominghome.open(mode='rb') + f = self._home.open(mode='rb+') try: num_leases = self._read_num_leases(f) hunk ./src/allmydata/storage/backends/disk/immutable.py 224 - finally: - f.close() - f = self._home.open(mode='wb+') - try: self._write_lease_record(f, num_leases, lease_info) self._write_num_leases(f, num_leases+1) finally: hunk ./src/allmydata/storage/backends/disk/mutable.py 440 pass -def create_mutable_disk_share(fp, serverid, write_enabler, parent): - ms = MutableDiskShare(fp, parent) +def create_mutable_disk_share(storageindex, shnum, fp, serverid, write_enabler, parent): + ms = MutableDiskShare(storageindex, shnum, fp, parent) ms.create(serverid, write_enabler) del ms hunk ./src/allmydata/storage/backends/disk/mutable.py 444 - return MutableDiskShare(fp, parent) + return MutableDiskShare(storageindex, shnum, fp, parent) hunk ./src/allmydata/storage/bucket.py 44 start = time.time() self._share.close() - filelen = self._share.stat() + # XXX should this be self._share.get_used_space() ? + consumed_size = self._share.get_size() self._share = None self.closed = True hunk ./src/allmydata/storage/bucket.py 51 self._canary.dontNotifyOnDisconnect(self._disconnect_marker) - self.ss.bucket_writer_closed(self, filelen) + self.ss.bucket_writer_closed(self, consumed_size) self.ss.add_latency("close", time.time() - start) self.ss.count("close") hunk ./src/allmydata/storage/server.py 182 renew_secret, cancel_secret, sharenums, allocated_size, canary, owner_num=0): - # cancel_secret is no longer used. # owner_num is not for clients to set, but rather it should be # curried into a StorageServer instance dedicated to a particular # owner. hunk ./src/allmydata/storage/server.py 195 # Note that the lease should not be added until the BucketWriter # has been closed. expire_time = time.time() + 31*24*60*60 - lease_info = LeaseInfo(owner_num, renew_secret, + lease_info = LeaseInfo(owner_num, renew_secret, cancel_secret, expire_time, self._serverid) max_space_per_bucket = allocated_size hunk ./src/allmydata/test/no_network.py 349 return self.g.servers_by_number[i] def get_serverdir(self, i): - return self.g.servers_by_number[i].backend.storedir + return self.g.servers_by_number[i].backend._storedir def remove_server(self, i): self.g.remove_server(self.g.servers_by_number[i].get_serverid()) hunk ./src/allmydata/test/no_network.py 357 def iterate_servers(self): for i in sorted(self.g.servers_by_number.keys()): ss = self.g.servers_by_number[i] - yield (i, ss, ss.backend.storedir) + yield (i, ss, ss.backend._storedir) def find_uri_shares(self, uri): si = tahoe_uri.from_string(uri).get_storage_index() hunk ./src/allmydata/test/no_network.py 384 return shares def copy_share(self, from_share, uri, to_server): - si = uri.from_string(self.uri).get_storage_index() + si = tahoe_uri.from_string(uri).get_storage_index() (i_shnum, i_serverid, i_sharefp) = from_share shares_dir = to_server.backend.get_shareset(si)._sharehomedir i_sharefp.copyTo(shares_dir.child(str(i_shnum))) hunk ./src/allmydata/test/test_download.py 127 return d - def _write_shares(self, uri, shares): - si = uri.from_string(uri).get_storage_index() + def _write_shares(self, fileuri, shares): + si = uri.from_string(fileuri).get_storage_index() for i in shares: shares_for_server = shares[i] for shnum in shares_for_server: hunk ./src/allmydata/test/test_hung_server.py 36 def _hang(self, servers, **kwargs): for ss in servers: - self.g.hang_server(ss.get_serverid(), **kwargs) + self.g.hang_server(ss.original.get_serverid(), **kwargs) def _unhang(self, servers, **kwargs): for ss in servers: hunk ./src/allmydata/test/test_hung_server.py 40 - self.g.unhang_server(ss.get_serverid(), **kwargs) + self.g.unhang_server(ss.original.get_serverid(), **kwargs) def _hang_shares(self, shnums, **kwargs): # hang all servers who are holding the given shares hunk ./src/allmydata/test/test_hung_server.py 52 hung_serverids.add(i_serverid) def _delete_all_shares_from(self, servers): - serverids = [ss.get_serverid() for ss in servers] + serverids = [ss.original.get_serverid() for ss in servers] for (i_shnum, i_serverid, i_sharefp) in self.shares: if i_serverid in serverids: i_sharefp.remove() hunk ./src/allmydata/test/test_hung_server.py 58 def _corrupt_all_shares_in(self, servers, corruptor_func): - serverids = [ss.get_serverid() for ss in servers] + serverids = [ss.original.get_serverid() for ss in servers] for (i_shnum, i_serverid, i_sharefp) in self.shares: if i_serverid in serverids: self.corrupt_share((i_shnum, i_serverid, i_sharefp), corruptor_func) hunk ./src/allmydata/test/test_hung_server.py 64 def _copy_all_shares_from(self, from_servers, to_server): - serverids = [ss.get_serverid() for ss in from_servers] + serverids = [ss.original.get_serverid() for ss in from_servers] for (i_shnum, i_serverid, i_sharefp) in self.shares: if i_serverid in serverids: self.copy_share((i_shnum, i_serverid, i_sharefp), self.uri, to_server) hunk ./src/allmydata/test/test_mutable.py 2990 fso = debug.FindSharesOptions() storage_index = base32.b2a(n.get_storage_index()) fso.si_s = storage_index - fso.nodedirs = [unicode(os.path.dirname(os.path.abspath(storedir))) + fso.nodedirs = [unicode(storedir.parent().path) for (i,ss,storedir) in self.iterate_servers()] fso.stdout = StringIO() hunk ./src/allmydata/test/test_upload.py 818 if share_number is not None: self._copy_share_to_server(share_number, server_number) - def _copy_share_to_server(self, share_number, server_number): ss = self.g.servers_by_number[server_number] hunk ./src/allmydata/test/test_upload.py 820 - self.copy_share(self.shares[share_number], ss) + self.copy_share(self.shares[share_number], self.uri, ss) def _setup_grid(self): """ } [docs/backends: document the configuration options for the pluggable backends scheme. refs #999 david-sarah@jacaranda.org**20110920171737 Ignore-this: 5947e864682a43cb04e557334cda7c19 ] { adddir ./docs/backends addfile ./docs/backends/S3.rst hunk ./docs/backends/S3.rst 1 +==================================================== +Storing Shares in Amazon Simple Storage Service (S3) +==================================================== + +S3 is a commercial storage service provided by Amazon, described at +``_. + +The Tahoe-LAFS storage server can be configured to store its shares in +an S3 bucket, rather than on local filesystem. To enable this, add the +following keys to the server's ``tahoe.cfg`` file: + +``[storage]`` + +``backend = s3`` + + This turns off the local filesystem backend and enables use of S3. + +``s3.access_key_id = (string, required)`` +``s3.secret_access_key = (string, required)`` + + These two give the storage server permission to access your Amazon + Web Services account, allowing them to upload and download shares + from S3. + +``s3.bucket = (string, required)`` + + This controls which bucket will be used to hold shares. The Tahoe-LAFS + storage server will only modify and access objects in the configured S3 + bucket. + +``s3.url = (URL string, optional)`` + + This URL tells the storage server how to access the S3 service. It + defaults to ``http://s3.amazonaws.com``, but by setting it to something + else, you may be able to use some other S3-like service if it is + sufficiently compatible. + +``s3.max_space = (str, optional)`` + + This tells the server to limit how much space can be used in the S3 + bucket. Before each share is uploaded, the server will ask S3 for the + current bucket usage, and will only accept the share if it does not cause + the usage to grow above this limit. + + The string contains a number, with an optional case-insensitive scale + suffix like "K" or "M" or "G", and an optional "B" or "iB" suffix. So + "100MB", "100M", "100000000B", "100000000", and "100000kb" all mean the + same thing. Likewise, "1MiB", "1024KiB", and "1048576B" all mean the same + thing. + + If ``s3.max_space`` is omitted, the default behavior is to allow + unlimited usage. + + +Once configured, the WUI "storage server" page will provide information about +how much space is being used and how many shares are being stored. + + +Issues +------ + +Objects in an S3 bucket cannot be read for free. As a result, when Tahoe-LAFS +is configured to store shares in S3 rather than on local disk, some common +operations may behave differently: + +* Lease crawling/expiration is not yet implemented. As a result, shares will + be retained forever, and the Storage Server status web page will not show + information about the number of mutable/immutable shares present. + +* Enabling ``s3.max_space`` causes an extra S3 usage query to be sent for + each share upload, causing the upload process to run slightly slower and + incur more S3 request charges. addfile ./docs/backends/disk.rst hunk ./docs/backends/disk.rst 1 +==================================== +Storing Shares on a Local Filesystem +==================================== + +The "disk" backend stores shares on the local filesystem. Versions of +Tahoe-LAFS <= 1.9.0 always stored shares in this way. + +``[storage]`` + +``backend = disk`` + + This enables use of the disk backend, and is the default. + +``reserved_space = (str, optional)`` + + If provided, this value defines how much disk space is reserved: the + storage server will not accept any share that causes the amount of free + disk space to drop below this value. (The free space is measured by a + call to statvfs(2) on Unix, or GetDiskFreeSpaceEx on Windows, and is the + space available to the user account under which the storage server runs.) + + This string contains a number, with an optional case-insensitive scale + suffix like "K" or "M" or "G", and an optional "B" or "iB" suffix. So + "100MB", "100M", "100000000B", "100000000", and "100000kb" all mean the + same thing. Likewise, "1MiB", "1024KiB", and "1048576B" all mean the same + thing. + + "``tahoe create-node``" generates a tahoe.cfg with + "``reserved_space=1G``", but you may wish to raise, lower, or remove the + reservation to suit your needs. + +``expire.enabled =`` + +``expire.mode =`` + +``expire.override_lease_duration =`` + +``expire.cutoff_date =`` + +``expire.immutable =`` + +``expire.mutable =`` + + These settings control garbage collection, causing the server to + delete shares that no longer have an up-to-date lease on them. Please + see ``_ for full details. hunk ./docs/configuration.rst 436 `_ for the current status of this bug. The default value is ``False``. -``reserved_space = (str, optional)`` +``backend = (string, optional)`` hunk ./docs/configuration.rst 438 - If provided, this value defines how much disk space is reserved: the - storage server will not accept any share that causes the amount of free - disk space to drop below this value. (The free space is measured by a - call to statvfs(2) on Unix, or GetDiskFreeSpaceEx on Windows, and is the - space available to the user account under which the storage server runs.) + Storage servers can store the data into different "backends". Clients + need not be aware of which backend is used by a server. The default + value is ``disk``. hunk ./docs/configuration.rst 442 - This string contains a number, with an optional case-insensitive scale - suffix like "K" or "M" or "G", and an optional "B" or "iB" suffix. So - "100MB", "100M", "100000000B", "100000000", and "100000kb" all mean the - same thing. Likewise, "1MiB", "1024KiB", and "1048576B" all mean the same - thing. +``backend = disk`` hunk ./docs/configuration.rst 444 - "``tahoe create-node``" generates a tahoe.cfg with - "``reserved_space=1G``", but you may wish to raise, lower, or remove the - reservation to suit your needs. + The default is to store shares on the local filesystem (in + BASEDIR/storage/shares/). For configuration details (including how to + reserve a minimum amount of free space), see ``_. hunk ./docs/configuration.rst 448 -``expire.enabled =`` +``backend = S3`` hunk ./docs/configuration.rst 450 -``expire.mode =`` - -``expire.override_lease_duration =`` - -``expire.cutoff_date =`` - -``expire.immutable =`` - -``expire.mutable =`` - - These settings control garbage collection, in which the server will - delete shares that no longer have an up-to-date lease on them. Please see - ``_ for full details. + The storage server can store all shares to an Amazon Simple Storage + Service (S3) bucket. For configuration details, see ``_. Running A Helper } [Fix some incorrect attribute accesses. refs #999 david-sarah@jacaranda.org**20110921031207 Ignore-this: f1ea4c3ea191f6d4b719afaebd2b2bcd ] { hunk ./src/allmydata/client.py 258 backend = DiskBackend(storedir, readonly=readonly, reserved_space=reserved, discard_storage=discard) - ss = StorageServer(nodeid, backend, storedir, + ss = StorageServer(self.nodeid, backend, storedir, stats_provider=self.stats_provider, expiration_policy=expiration_policy) self.add_service(ss) hunk ./src/allmydata/interfaces.py 449 Returns the storage index. """ + def get_storage_index_string(): + """ + Returns the base32-encoded storage index. + """ + def get_shnum(): """ Returns the share number. hunk ./src/allmydata/storage/backends/disk/immutable.py 138 def get_storage_index(self): return self._storageindex + def get_storage_index_string(self): + return si_b2a(self._storageindex) + def get_shnum(self): return self._shnum hunk ./src/allmydata/storage/backends/disk/mutable.py 119 def get_storage_index(self): return self._storageindex + def get_storage_index_string(self): + return si_b2a(self._storageindex) + def get_shnum(self): return self._shnum hunk ./src/allmydata/storage/bucket.py 86 def __init__(self, ss, share): self.ss = ss self._share = share - self.storageindex = share.storageindex - self.shnum = share.shnum + self.storageindex = share.get_storage_index() + self.shnum = share.get_shnum() def __repr__(self): return "<%s %s %s>" % (self.__class__.__name__, hunk ./src/allmydata/storage/expirer.py 6 from twisted.python import log as twlog from allmydata.storage.crawler import ShareCrawler -from allmydata.storage.common import si_b2a, UnknownMutableContainerVersionError, \ +from allmydata.storage.common import UnknownMutableContainerVersionError, \ UnknownImmutableContainerVersionError hunk ./src/allmydata/storage/expirer.py 124 struct.error): twlog.msg("lease-checker error processing %r" % (share,)) twlog.err() - which = (si_b2a(share.storageindex), share.get_shnum()) + which = (share.get_storage_index_string(), share.get_shnum()) self.state["cycle-to-date"]["corrupt-shares"].append(which) wks = (1, 1, 1, "unknown") would_keep_shares.append(wks) hunk ./src/allmydata/storage/server.py 221 alreadygot = set() for share in shareset.get_shares(): share.add_or_renew_lease(lease_info) - alreadygot.add(share.shnum) + alreadygot.add(share.get_shnum()) for shnum in sharenums - alreadygot: if shareset.has_incoming(shnum): hunk ./src/allmydata/storage/server.py 324 try: shareset = self.backend.get_shareset(storageindex) - return shareset.readv(self, shares, readv) + return shareset.readv(shares, readv) finally: self.add_latency("readv", time.time() - start) hunk ./src/allmydata/storage/shares.py 1 -#! /usr/bin/python - -from allmydata.storage.mutable import MutableShareFile -from allmydata.storage.immutable import ShareFile - -def get_share_file(filename): - f = open(filename, "rb") - prefix = f.read(32) - f.close() - if prefix == MutableShareFile.MAGIC: - return MutableShareFile(filename) - # otherwise assume it's immutable - return ShareFile(filename) - rmfile ./src/allmydata/storage/shares.py hunk ./src/allmydata/test/no_network.py 387 si = tahoe_uri.from_string(uri).get_storage_index() (i_shnum, i_serverid, i_sharefp) = from_share shares_dir = to_server.backend.get_shareset(si)._sharehomedir + fileutil.fp_make_dirs(shares_dir) i_sharefp.copyTo(shares_dir.child(str(i_shnum))) def restore_all_shares(self, shares): hunk ./src/allmydata/test/no_network.py 391 - for share, data in shares.items(): - share.home.setContent(data) + for sharepath, data in shares.items(): + FilePath(sharepath).setContent(data) def delete_share(self, (shnum, serverid, sharefp)): sharefp.remove() hunk ./src/allmydata/test/test_upload.py 744 servertoshnums = {} # k: server, v: set(shnum) for i, c in self.g.servers_by_number.iteritems(): - for (dirp, dirns, fns) in os.walk(c.sharedir): + for (dirp, dirns, fns) in os.walk(c.backend._sharedir.path): for fn in fns: try: sharenum = int(fn) } [docs/backends/S3.rst: remove Issues section. refs #999 david-sarah@jacaranda.org**20110921031625 Ignore-this: c83d8f52b790bc32488869e6ee1df8c2 ] hunk ./docs/backends/S3.rst 57 Once configured, the WUI "storage server" page will provide information about how much space is being used and how many shares are being stored. - - -Issues ------- - -Objects in an S3 bucket cannot be read for free. As a result, when Tahoe-LAFS -is configured to store shares in S3 rather than on local disk, some common -operations may behave differently: - -* Lease crawling/expiration is not yet implemented. As a result, shares will - be retained forever, and the Storage Server status web page will not show - information about the number of mutable/immutable shares present. - -* Enabling ``s3.max_space`` causes an extra S3 usage query to be sent for - each share upload, causing the upload process to run slightly slower and - incur more S3 request charges. [docs/backends/S3.rst, disk.rst: describe type of space settings as 'quantity of space', not 'str'. refs #999 david-sarah@jacaranda.org**20110921031705 Ignore-this: a74ed8e01b0a1ab5f07a1487d7bf138 ] { hunk ./docs/backends/S3.rst 38 else, you may be able to use some other S3-like service if it is sufficiently compatible. -``s3.max_space = (str, optional)`` +``s3.max_space = (quantity of space, optional)`` This tells the server to limit how much space can be used in the S3 bucket. Before each share is uploaded, the server will ask S3 for the hunk ./docs/backends/disk.rst 14 This enables use of the disk backend, and is the default. -``reserved_space = (str, optional)`` +``reserved_space = (quantity of space, optional)`` If provided, this value defines how much disk space is reserved: the storage server will not accept any share that causes the amount of free } [More fixes to tests needed for pluggable backends. refs #999 david-sarah@jacaranda.org**20110921184649 Ignore-this: 9be0d3a98e350fd4e17a07d2c00bb4ca ] { hunk ./src/allmydata/scripts/debug.py 8 from twisted.python import usage, failure from twisted.internet import defer from twisted.scripts import trial as twisted_trial +from twisted.python.filepath import FilePath class DumpOptions(usage.Options): hunk ./src/allmydata/scripts/debug.py 38 self['filename'] = argv_to_abspath(filename) def dump_share(options): - from allmydata.storage.mutable import MutableShareFile + from allmydata.storage.backends.disk.disk_backend import get_share from allmydata.util.encodingutil import quote_output out = options.stdout hunk ./src/allmydata/scripts/debug.py 46 # check the version, to see if we have a mutable or immutable share print >>out, "share filename: %s" % quote_output(options['filename']) - f = open(options['filename'], "rb") - prefix = f.read(32) - f.close() - if prefix == MutableShareFile.MAGIC: - return dump_mutable_share(options) - # otherwise assume it's immutable - return dump_immutable_share(options) - -def dump_immutable_share(options): - from allmydata.storage.immutable import ShareFile + share = get_share("", 0, fp) + if share.sharetype == "mutable": + return dump_mutable_share(options, share) + else: + assert share.sharetype == "immutable", share.sharetype + return dump_immutable_share(options) hunk ./src/allmydata/scripts/debug.py 53 +def dump_immutable_share(options, share): out = options.stdout hunk ./src/allmydata/scripts/debug.py 55 - f = ShareFile(options['filename']) if not options["leases-only"]: hunk ./src/allmydata/scripts/debug.py 56 - dump_immutable_chk_share(f, out, options) - dump_immutable_lease_info(f, out) + dump_immutable_chk_share(share, out, options) + dump_immutable_lease_info(share, out) print >>out return 0 hunk ./src/allmydata/scripts/debug.py 166 return when -def dump_mutable_share(options): - from allmydata.storage.mutable import MutableShareFile +def dump_mutable_share(options, m): from allmydata.util import base32, idlib out = options.stdout hunk ./src/allmydata/scripts/debug.py 169 - m = MutableShareFile(options['filename']) f = open(options['filename'], "rb") WE, nodeid = m._read_write_enabler_and_nodeid(f) num_extra_leases = m._read_num_extra_leases(f) hunk ./src/allmydata/scripts/debug.py 641 /home/warner/testnet/node-1/storage/shares/44k/44kai1tui348689nrw8fjegc8c/9 /home/warner/testnet/node-2/storage/shares/44k/44kai1tui348689nrw8fjegc8c/2 """ - from allmydata.storage.server import si_a2b, storage_index_to_dir - from allmydata.util.encodingutil import listdir_unicode + from allmydata.storage.server import si_a2b + from allmydata.storage.backends.disk_backend import si_si2dir + from allmydata.util.encodingutil import quote_filepath out = options.stdout hunk ./src/allmydata/scripts/debug.py 646 - sharedir = storage_index_to_dir(si_a2b(options.si_s)) - for d in options.nodedirs: - d = os.path.join(d, "storage/shares", sharedir) - if os.path.exists(d): - for shnum in listdir_unicode(d): - print >>out, os.path.join(d, shnum) + si = si_a2b(options.si_s) + for nodedir in options.nodedirs: + sharedir = si_si2dir(nodedir.child("storage").child("shares"), si) + if sharedir.exists(): + for sharefp in sharedir.children(): + print >>out, quote_filepath(sharefp, quotemarks=False) return 0 hunk ./src/allmydata/scripts/debug.py 878 print >>err, "Error processing %s" % quote_output(si_dir) failure.Failure().printTraceback(err) + class CorruptShareOptions(usage.Options): def getSynopsis(self): return "Usage: tahoe debug corrupt-share SHARE_FILENAME" hunk ./src/allmydata/scripts/debug.py 902 Obviously, this command should not be used in normal operation. """ return t + def parseArgs(self, filename): self['filename'] = filename hunk ./src/allmydata/scripts/debug.py 907 def corrupt_share(options): + do_corrupt_share(options.stdout, FilePath(options['filename']), options['offset']) + +def do_corrupt_share(out, fp, offset="block-random"): import random hunk ./src/allmydata/scripts/debug.py 911 - from allmydata.storage.mutable import MutableShareFile - from allmydata.storage.immutable import ShareFile + from allmydata.storage.backends.disk.mutable import MutableDiskShare + from allmydata.storage.backends.disk.immutable import ImmutableDiskShare from allmydata.mutable.layout import unpack_header from allmydata.immutable.layout import ReadBucketProxy hunk ./src/allmydata/scripts/debug.py 915 - out = options.stdout - fn = options['filename'] - assert options["offset"] == "block-random", "other offsets not implemented" + + assert offset == "block-random", "other offsets not implemented" + # first, what kind of share is it? def flip_bit(start, end): hunk ./src/allmydata/scripts/debug.py 924 offset = random.randrange(start, end) bit = random.randrange(0, 8) print >>out, "[%d..%d): %d.b%d" % (start, end, offset, bit) - f = open(fn, "rb+") - f.seek(offset) - d = f.read(1) - d = chr(ord(d) ^ 0x01) - f.seek(offset) - f.write(d) - f.close() + f = fp.open("rb+") + try: + f.seek(offset) + d = f.read(1) + d = chr(ord(d) ^ 0x01) + f.seek(offset) + f.write(d) + finally: + f.close() hunk ./src/allmydata/scripts/debug.py 934 - f = open(fn, "rb") - prefix = f.read(32) - f.close() - if prefix == MutableShareFile.MAGIC: - # mutable - m = MutableShareFile(fn) - f = open(fn, "rb") - f.seek(m.DATA_OFFSET) - data = f.read(2000) - # make sure this slot contains an SMDF share - assert data[0] == "\x00", "non-SDMF mutable shares not supported" + f = fp.open("rb") + try: + prefix = f.read(32) + finally: f.close() hunk ./src/allmydata/scripts/debug.py 939 + if prefix == MutableDiskShare.MAGIC: + # mutable + m = MutableDiskShare("", 0, fp) + f = fp.open("rb") + try: + f.seek(m.DATA_OFFSET) + data = f.read(2000) + # make sure this slot contains an SMDF share + assert data[0] == "\x00", "non-SDMF mutable shares not supported" + finally: + f.close() (version, ig_seqnum, ig_roothash, ig_IV, ig_k, ig_N, ig_segsize, ig_datalen, offsets) = unpack_header(data) hunk ./src/allmydata/scripts/debug.py 960 flip_bit(start, end) else: # otherwise assume it's immutable - f = ShareFile(fn) + f = ImmutableDiskShare("", 0, fp) bp = ReadBucketProxy(None, None, '') offsets = bp._parse_offsets(f.read_share_data(0, 0x24)) start = f._data_offset + offsets["data"] hunk ./src/allmydata/storage/backends/base.py 92 (testv, datav, new_length) = test_and_write_vectors[sharenum] if sharenum in shares: if not shares[sharenum].check_testv(testv): - self.log("testv failed: [%d]: %r" % (sharenum, testv)) + storageserver.log("testv failed: [%d]: %r" % (sharenum, testv)) testv_is_good = False break else: hunk ./src/allmydata/storage/backends/base.py 99 # compare the vectors against an empty share, in which all # reads return empty strings if not EmptyShare().check_testv(testv): - self.log("testv failed (empty): [%d] %r" % (sharenum, - testv)) + storageserver.log("testv failed (empty): [%d] %r" % (sharenum, testv)) testv_is_good = False break hunk ./src/allmydata/test/test_cli.py 2892 # delete one, corrupt a second shares = self.find_uri_shares(self.uri) self.failUnlessReallyEqual(len(shares), 10) - os.unlink(shares[0][2]) - cso = debug.CorruptShareOptions() - cso.stdout = StringIO() - cso.parseOptions([shares[1][2]]) + shares[0][2].remove() + stdout = StringIO() + sharefile = shares[1][2] storage_index = uri.from_string(self.uri).get_storage_index() self._corrupt_share_line = " server %s, SI %s, shnum %d" % \ (base32.b2a(shares[1][1]), hunk ./src/allmydata/test/test_cli.py 2900 base32.b2a(storage_index), shares[1][0]) - debug.corrupt_share(cso) + debug.do_corrupt_share(stdout, sharefile) d.addCallback(_clobber_shares) d.addCallback(lambda ign: self.do_cli("check", "--verify", self.uri)) hunk ./src/allmydata/test/test_cli.py 3017 def _clobber_shares(ignored): shares = self.find_uri_shares(self.uris[u"g\u00F6\u00F6d"]) self.failUnlessReallyEqual(len(shares), 10) - os.unlink(shares[0][2]) + shares[0][2].remove() shares = self.find_uri_shares(self.uris["mutable"]) hunk ./src/allmydata/test/test_cli.py 3020 - cso = debug.CorruptShareOptions() - cso.stdout = StringIO() - cso.parseOptions([shares[1][2]]) + stdout = StringIO() + sharefile = shares[1][2] storage_index = uri.from_string(self.uris["mutable"]).get_storage_index() self._corrupt_share_line = " corrupt: server %s, SI %s, shnum %d" % \ (base32.b2a(shares[1][1]), hunk ./src/allmydata/test/test_cli.py 3027 base32.b2a(storage_index), shares[1][0]) - debug.corrupt_share(cso) + debug.do_corrupt_share(stdout, sharefile) d.addCallback(_clobber_shares) # root hunk ./src/allmydata/test/test_client.py 90 "enabled = true\n" + \ "reserved_space = 1000\n") c = client.Client(basedir) - self.failUnlessEqual(c.getServiceNamed("storage").reserved_space, 1000) + self.failUnlessEqual(c.getServiceNamed("storage").backend._reserved_space, 1000) def test_reserved_2(self): basedir = "client.Basic.test_reserved_2" hunk ./src/allmydata/test/test_client.py 101 "enabled = true\n" + \ "reserved_space = 10K\n") c = client.Client(basedir) - self.failUnlessEqual(c.getServiceNamed("storage").reserved_space, 10*1000) + self.failUnlessEqual(c.getServiceNamed("storage").backend._reserved_space, 10*1000) def test_reserved_3(self): basedir = "client.Basic.test_reserved_3" hunk ./src/allmydata/test/test_client.py 112 "enabled = true\n" + \ "reserved_space = 5mB\n") c = client.Client(basedir) - self.failUnlessEqual(c.getServiceNamed("storage").reserved_space, + self.failUnlessEqual(c.getServiceNamed("storage").backend._reserved_space, 5*1000*1000) def test_reserved_4(self): hunk ./src/allmydata/test/test_client.py 124 "enabled = true\n" + \ "reserved_space = 78Gb\n") c = client.Client(basedir) - self.failUnlessEqual(c.getServiceNamed("storage").reserved_space, + self.failUnlessEqual(c.getServiceNamed("storage").backend._reserved_space, 78*1000*1000*1000) def test_reserved_bad(self): hunk ./src/allmydata/test/test_client.py 136 "enabled = true\n" + \ "reserved_space = bogus\n") c = client.Client(basedir) - self.failUnlessEqual(c.getServiceNamed("storage").reserved_space, 0) + self.failUnlessEqual(c.getServiceNamed("storage").backend._reserved_space, 0) def _permute(self, sb, key): return [ s.get_serverid() for s in sb.get_servers_for_psi(key) ] hunk ./src/allmydata/test/test_crawler.py 7 from twisted.trial import unittest from twisted.application import service from twisted.internet import defer +from twisted.python.filepath import FilePath from foolscap.api import eventually, fireEventually from allmydata.util import fileutil, hashutil, pollmixin hunk ./src/allmydata/test/test_crawler.py 13 from allmydata.storage.server import StorageServer, si_b2a from allmydata.storage.crawler import ShareCrawler, TimeSliceExceeded +from allmydata.storage.backends.disk.disk_backend import DiskBackend from allmydata.test.test_storage import FakeCanary from allmydata.test.common_util import StallMixin hunk ./src/allmydata/test/test_crawler.py 115 def test_immediate(self): self.basedir = "crawler/Basic/immediate" - fileutil.make_dirs(self.basedir) serverid = "\x00" * 20 hunk ./src/allmydata/test/test_crawler.py 116 - ss = StorageServer(self.basedir, serverid) + fp = FilePath(self.basedir) + backend = DiskBackend(fp) + ss = StorageServer(serverid, backend, fp) ss.setServiceParent(self.s) sis = [self.write(i, ss, serverid) for i in range(10)] hunk ./src/allmydata/test/test_crawler.py 122 - statefile = os.path.join(self.basedir, "statefile") + statefp = fp.child("statefile") hunk ./src/allmydata/test/test_crawler.py 124 - c = BucketEnumeratingCrawler(ss, statefile, allowed_cpu_percentage=.1) + c = BucketEnumeratingCrawler(backend, statefp, allowed_cpu_percentage=.1) c.load_state() c.start_current_prefix(time.time()) hunk ./src/allmydata/test/test_crawler.py 137 self.failUnlessEqual(sorted(sis), sorted(c.all_buckets)) # check that a new crawler picks up on the state file properly - c2 = BucketEnumeratingCrawler(ss, statefile) + c2 = BucketEnumeratingCrawler(backend, statefp) c2.load_state() c2.start_current_prefix(time.time()) hunk ./src/allmydata/test/test_crawler.py 145 def test_service(self): self.basedir = "crawler/Basic/service" - fileutil.make_dirs(self.basedir) serverid = "\x00" * 20 hunk ./src/allmydata/test/test_crawler.py 146 - ss = StorageServer(self.basedir, serverid) + fp = FilePath(self.basedir) + backend = DiskBackend(fp) + ss = StorageServer(serverid, backend, fp) ss.setServiceParent(self.s) sis = [self.write(i, ss, serverid) for i in range(10)] hunk ./src/allmydata/test/test_crawler.py 153 - statefile = os.path.join(self.basedir, "statefile") - c = BucketEnumeratingCrawler(ss, statefile) + statefp = fp.child("statefile") + c = BucketEnumeratingCrawler(backend, statefp) c.setServiceParent(self.s) # it should be legal to call get_state() and get_progress() right hunk ./src/allmydata/test/test_crawler.py 174 def test_paced(self): self.basedir = "crawler/Basic/paced" - fileutil.make_dirs(self.basedir) serverid = "\x00" * 20 hunk ./src/allmydata/test/test_crawler.py 175 - ss = StorageServer(self.basedir, serverid) + fp = FilePath(self.basedir) + backend = DiskBackend(fp) + ss = StorageServer(serverid, backend, fp) ss.setServiceParent(self.s) # put four buckets in each prefixdir hunk ./src/allmydata/test/test_crawler.py 186 for tail in range(4): sis.append(self.write(i, ss, serverid, tail)) - statefile = os.path.join(self.basedir, "statefile") + statefp = fp.child("statefile") hunk ./src/allmydata/test/test_crawler.py 188 - c = PacedCrawler(ss, statefile) + c = PacedCrawler(backend, statefp) c.load_state() try: c.start_current_prefix(time.time()) hunk ./src/allmydata/test/test_crawler.py 213 del c # start a new crawler, it should start from the beginning - c = PacedCrawler(ss, statefile) + c = PacedCrawler(backend, statefp) c.load_state() try: c.start_current_prefix(time.time()) hunk ./src/allmydata/test/test_crawler.py 226 c.cpu_slice = PacedCrawler.cpu_slice # a third crawler should pick up from where it left off - c2 = PacedCrawler(ss, statefile) + c2 = PacedCrawler(backend, statefp) c2.all_buckets = c.all_buckets[:] c2.load_state() c2.countdown = -1 hunk ./src/allmydata/test/test_crawler.py 237 # now stop it at the end of a bucket (countdown=4), to exercise a # different place that checks the time - c = PacedCrawler(ss, statefile) + c = PacedCrawler(backend, statefp) c.load_state() c.countdown = 4 try: hunk ./src/allmydata/test/test_crawler.py 256 # stop it again at the end of the bucket, check that a new checker # picks up correctly - c = PacedCrawler(ss, statefile) + c = PacedCrawler(backend, statefp) c.load_state() c.countdown = 4 try: hunk ./src/allmydata/test/test_crawler.py 266 # that should stop at the end of one of the buckets. c.save_state() - c2 = PacedCrawler(ss, statefile) + c2 = PacedCrawler(backend, statefp) c2.all_buckets = c.all_buckets[:] c2.load_state() c2.countdown = -1 hunk ./src/allmydata/test/test_crawler.py 277 def test_paced_service(self): self.basedir = "crawler/Basic/paced_service" - fileutil.make_dirs(self.basedir) serverid = "\x00" * 20 hunk ./src/allmydata/test/test_crawler.py 278 - ss = StorageServer(self.basedir, serverid) + fp = FilePath(self.basedir) + backend = DiskBackend(fp) + ss = StorageServer(serverid, backend, fp) ss.setServiceParent(self.s) sis = [self.write(i, ss, serverid) for i in range(10)] hunk ./src/allmydata/test/test_crawler.py 285 - statefile = os.path.join(self.basedir, "statefile") - c = PacedCrawler(ss, statefile) + statefp = fp.child("statefile") + c = PacedCrawler(backend, statefp) did_check_progress = [False] def check_progress(): hunk ./src/allmydata/test/test_crawler.py 345 # and read the stdout when it runs. self.basedir = "crawler/Basic/cpu_usage" - fileutil.make_dirs(self.basedir) serverid = "\x00" * 20 hunk ./src/allmydata/test/test_crawler.py 346 - ss = StorageServer(self.basedir, serverid) + fp = FilePath(self.basedir) + backend = DiskBackend(fp) + ss = StorageServer(serverid, backend, fp) ss.setServiceParent(self.s) for i in range(10): hunk ./src/allmydata/test/test_crawler.py 354 self.write(i, ss, serverid) - statefile = os.path.join(self.basedir, "statefile") - c = ConsumingCrawler(ss, statefile) + statefp = fp.child("statefile") + c = ConsumingCrawler(backend, statefp) c.setServiceParent(self.s) # this will run as fast as it can, consuming about 50ms per call to hunk ./src/allmydata/test/test_crawler.py 391 def test_empty_subclass(self): self.basedir = "crawler/Basic/empty_subclass" - fileutil.make_dirs(self.basedir) serverid = "\x00" * 20 hunk ./src/allmydata/test/test_crawler.py 392 - ss = StorageServer(self.basedir, serverid) + fp = FilePath(self.basedir) + backend = DiskBackend(fp) + ss = StorageServer(serverid, backend, fp) ss.setServiceParent(self.s) for i in range(10): hunk ./src/allmydata/test/test_crawler.py 400 self.write(i, ss, serverid) - statefile = os.path.join(self.basedir, "statefile") - c = ShareCrawler(ss, statefile) + statefp = fp.child("statefile") + c = ShareCrawler(backend, statefp) c.slow_start = 0 c.setServiceParent(self.s) hunk ./src/allmydata/test/test_crawler.py 417 d.addCallback(_done) return d - def test_oneshot(self): self.basedir = "crawler/Basic/oneshot" hunk ./src/allmydata/test/test_crawler.py 419 - fileutil.make_dirs(self.basedir) serverid = "\x00" * 20 hunk ./src/allmydata/test/test_crawler.py 420 - ss = StorageServer(self.basedir, serverid) + fp = FilePath(self.basedir) + backend = DiskBackend(fp) + ss = StorageServer(serverid, backend, fp) ss.setServiceParent(self.s) for i in range(30): hunk ./src/allmydata/test/test_crawler.py 428 self.write(i, ss, serverid) - statefile = os.path.join(self.basedir, "statefile") - c = OneShotCrawler(ss, statefile) + statefp = fp.child("statefile") + c = OneShotCrawler(backend, statefp) c.setServiceParent(self.s) d = c.finished_d hunk ./src/allmydata/test/test_crawler.py 447 self.failUnlessEqual(s["current-cycle"], None) d.addCallback(_check) return d - hunk ./src/allmydata/test/test_deepcheck.py 23 ShouldFailMixin from allmydata.test.common_util import StallMixin from allmydata.test.no_network import GridTestMixin +from allmydata.scripts import debug + timeout = 2400 # One of these took 1046.091s on Zandr's ARM box. hunk ./src/allmydata/test/test_deepcheck.py 905 d.addErrback(self.explain_error) return d - - def set_up_damaged_tree(self): # 6.4s hunk ./src/allmydata/test/test_deepcheck.py 989 return d - def _run_cli(self, argv): - stdout, stderr = StringIO(), StringIO() - # this can only do synchronous operations - assert argv[0] == "debug" - runner.runner(argv, run_by_human=False, stdout=stdout, stderr=stderr) - return stdout.getvalue() - def _delete_some_shares(self, node): self.delete_shares_numbered(node.get_uri(), [0,1]) hunk ./src/allmydata/test/test_deepcheck.py 995 def _corrupt_some_shares(self, node): for (shnum, serverid, sharefile) in self.find_uri_shares(node.get_uri()): if shnum in (0,1): - self._run_cli(["debug", "corrupt-share", sharefile]) + debug.do_corrupt_share(StringIO(), sharefile) def _delete_most_shares(self, node): self.delete_shares_numbered(node.get_uri(), range(1,10)) hunk ./src/allmydata/test/test_deepcheck.py 1000 - def check_is_healthy(self, cr, where): try: self.failUnless(ICheckResults.providedBy(cr), (cr, type(cr), where)) hunk ./src/allmydata/test/test_download.py 134 for shnum in shares_for_server: share_dir = self.get_server(i).backend.get_shareset(si)._sharehomedir fileutil.fp_make_dirs(share_dir) - share_dir.child(str(shnum)).setContent(shares[shnum]) + share_dir.child(str(shnum)).setContent(shares_for_server[shnum]) def load_shares(self, ignored=None): # this uses the data generated by create_shares() to populate the hunk ./src/allmydata/test/test_hung_server.py 32 def _break(self, servers): for ss in servers: - self.g.break_server(ss.get_serverid()) + self.g.break_server(ss.original.get_serverid()) def _hang(self, servers, **kwargs): for ss in servers: hunk ./src/allmydata/test/test_hung_server.py 67 serverids = [ss.original.get_serverid() for ss in from_servers] for (i_shnum, i_serverid, i_sharefp) in self.shares: if i_serverid in serverids: - self.copy_share((i_shnum, i_serverid, i_sharefp), self.uri, to_server) + self.copy_share((i_shnum, i_serverid, i_sharefp), self.uri, to_server.original) self.shares = self.find_uri_shares(self.uri) hunk ./src/allmydata/test/test_mutable.py 3669 # Now execute each assignment by writing the storage. for (share, servernum) in assignments: sharedata = base64.b64decode(self.sdmf_old_shares[share]) - storage_dir = self.get_server(servernum).backend.get_shareset(si).sharehomedir + storage_dir = self.get_server(servernum).backend.get_shareset(si)._sharehomedir fileutil.fp_make_dirs(storage_dir) storage_dir.child("%d" % share).setContent(sharedata) # ...and verify that the shares are there. hunk ./src/allmydata/test/test_no_network.py 10 from allmydata.immutable.upload import Data from allmydata.util.consumer import download_to_data + class Harness(unittest.TestCase): def setUp(self): self.s = service.MultiService() hunk ./src/allmydata/test/test_storage.py 1 -import time, os.path, platform, stat, re, simplejson, struct, shutil +import time, os.path, platform, stat, re, simplejson, struct, shutil, itertools import mock hunk ./src/allmydata/test/test_storage.py 6 from twisted.trial import unittest - from twisted.internet import defer from twisted.application import service hunk ./src/allmydata/test/test_storage.py 8 +from twisted.python.filepath import FilePath from foolscap.api import fireEventually hunk ./src/allmydata/test/test_storage.py 10 -import itertools + from allmydata import interfaces from allmydata.util import fileutil, hashutil, base32, pollmixin, time_format from allmydata.storage.server import StorageServer hunk ./src/allmydata/test/test_storage.py 14 +from allmydata.storage.backends.disk.disk_backend import DiskBackend from allmydata.storage.backends.disk.mutable import MutableDiskShare from allmydata.storage.bucket import BucketWriter, BucketReader from allmydata.storage.common import DataTooLargeError, \ hunk ./src/allmydata/test/test_storage.py 310 return self.sparent.stopService() def workdir(self, name): - basedir = os.path.join("storage", "Server", name) - return basedir + return FilePath("storage").child("Server").child(name) def create(self, name, reserved_space=0, klass=StorageServer): workdir = self.workdir(name) hunk ./src/allmydata/test/test_storage.py 314 - ss = klass(workdir, "\x00" * 20, reserved_space=reserved_space, + backend = DiskBackend(workdir, readonly=False, reserved_space=reserved_space) + ss = klass("\x00" * 20, backend, workdir, stats_provider=FakeStatsProvider()) ss.setServiceParent(self.sparent) return ss hunk ./src/allmydata/test/test_storage.py 1386 def tearDown(self): self.sparent.stopService() - shutil.rmtree(self.workdir("MDMFProxies storage test server")) + fileutil.fp_remove(self.workdir("MDMFProxies storage test server")) def write_enabler(self, we_tag): hunk ./src/allmydata/test/test_storage.py 2781 return self.sparent.stopService() def workdir(self, name): - basedir = os.path.join("storage", "Server", name) - return basedir + return FilePath("storage").child("Server").child(name) def create(self, name): workdir = self.workdir(name) hunk ./src/allmydata/test/test_storage.py 2785 - ss = StorageServer(workdir, "\x00" * 20) + backend = DiskBackend(workdir) + ss = StorageServer("\x00" * 20, backend, workdir) ss.setServiceParent(self.sparent) return ss hunk ./src/allmydata/test/test_storage.py 4061 } basedir = "storage/WebStatus/status_right_disk_stats" - fileutil.make_dirs(basedir) - ss = StorageServer(basedir, "\x00" * 20, reserved_space=reserved_space) - expecteddir = ss.sharedir + fp = FilePath(basedir) + backend = DiskBackend(fp, readonly=False, reserved_space=reserved_space) + ss = StorageServer("\x00" * 20, backend, fp) + expecteddir = backend._sharedir ss.setServiceParent(self.s) w = StorageStatus(ss) html = w.renderSynchronously() hunk ./src/allmydata/test/test_storage.py 4084 def test_readonly(self): basedir = "storage/WebStatus/readonly" - fileutil.make_dirs(basedir) - ss = StorageServer(basedir, "\x00" * 20, readonly_storage=True) + fp = FilePath(basedir) + backend = DiskBackend(fp, readonly=True) + ss = StorageServer("\x00" * 20, backend, fp) ss.setServiceParent(self.s) w = StorageStatus(ss) html = w.renderSynchronously() hunk ./src/allmydata/test/test_storage.py 4096 def test_reserved(self): basedir = "storage/WebStatus/reserved" - fileutil.make_dirs(basedir) - ss = StorageServer(basedir, "\x00" * 20, reserved_space=10e6) - ss.setServiceParent(self.s) - w = StorageStatus(ss) - html = w.renderSynchronously() - self.failUnlessIn("

Storage Server Status

", html) - s = remove_tags(html) - self.failUnlessIn("Reserved space: - 10.00 MB (10000000)", s) - - def test_huge_reserved(self): - basedir = "storage/WebStatus/reserved" - fileutil.make_dirs(basedir) - ss = StorageServer(basedir, "\x00" * 20, reserved_space=10e6) + fp = FilePath(basedir) + backend = DiskBackend(fp, readonly=False, reserved_space=10e6) + ss = StorageServer("\x00" * 20, backend, fp) ss.setServiceParent(self.s) w = StorageStatus(ss) html = w.renderSynchronously() hunk ./src/allmydata/test/test_upload.py 3 # -*- coding: utf-8 -*- -import os, shutil +import os from cStringIO import StringIO from twisted.trial import unittest from twisted.python.failure import Failure hunk ./src/allmydata/test/test_upload.py 14 from allmydata import uri, monitor, client from allmydata.immutable import upload, encode from allmydata.interfaces import FileTooLargeError, UploadUnhappinessError -from allmydata.util import log +from allmydata.util import log, fileutil from allmydata.util.assertutil import precondition from allmydata.util.deferredutil import DeferredListShouldSucceed from allmydata.test.no_network import GridTestMixin hunk ./src/allmydata/test/test_upload.py 972 readonly=True)) # Remove the first share from server 0. def _remove_share_0_from_server_0(): - share_location = self.shares[0][2] - os.remove(share_location) + self.shares[0][2].remove() d.addCallback(lambda ign: _remove_share_0_from_server_0()) # Set happy = 4 in the client. hunk ./src/allmydata/test/test_upload.py 1847 self._copy_share_to_server(3, 1) storedir = self.get_serverdir(0) # remove the storedir, wiping out any existing shares - shutil.rmtree(storedir) + fileutil.fp_remove(storedir) # create an empty storedir to replace the one we just removed hunk ./src/allmydata/test/test_upload.py 1849 - os.mkdir(storedir) + storedir.mkdir() client = self.g.clients[0] client.DEFAULT_ENCODING_PARAMETERS['happy'] = 4 return client hunk ./src/allmydata/test/test_upload.py 1888 self._copy_share_to_server(3, 1) storedir = self.get_serverdir(0) # remove the storedir, wiping out any existing shares - shutil.rmtree(storedir) + fileutil.fp_remove(storedir) # create an empty storedir to replace the one we just removed hunk ./src/allmydata/test/test_upload.py 1890 - os.mkdir(storedir) + storedir.mkdir() client = self.g.clients[0] client.DEFAULT_ENCODING_PARAMETERS['happy'] = 4 return client hunk ./src/allmydata/test/test_web.py 4870 d.addErrback(self.explain_web_error) return d - def _assert_leasecount(self, ignored, which, expected): + def _assert_leasecount(self, which, expected): lease_counts = self.count_leases(self.uris[which]) for (fn, num_leases) in lease_counts: if num_leases != expected: hunk ./src/allmydata/test/test_web.py 4903 self.fileurls[which] = "uri/" + urllib.quote(self.uris[which]) d.addCallback(_compute_fileurls) - d.addCallback(self._assert_leasecount, "one", 1) - d.addCallback(self._assert_leasecount, "two", 1) - d.addCallback(self._assert_leasecount, "mutable", 1) + d.addCallback(lambda ign: self._assert_leasecount("one", 1)) + d.addCallback(lambda ign: self._assert_leasecount("two", 1)) + d.addCallback(lambda ign: self._assert_leasecount("mutable", 1)) d.addCallback(self.CHECK, "one", "t=check") # no add-lease def _got_html_good(res): hunk ./src/allmydata/test/test_web.py 4913 self.failIf("Not Healthy" in res, res) d.addCallback(_got_html_good) - d.addCallback(self._assert_leasecount, "one", 1) - d.addCallback(self._assert_leasecount, "two", 1) - d.addCallback(self._assert_leasecount, "mutable", 1) + d.addCallback(lambda ign: self._assert_leasecount("one", 1)) + d.addCallback(lambda ign: self._assert_leasecount("two", 1)) + d.addCallback(lambda ign: self._assert_leasecount("mutable", 1)) # this CHECK uses the original client, which uses the same # lease-secrets, so it will just renew the original lease hunk ./src/allmydata/test/test_web.py 4922 d.addCallback(self.CHECK, "one", "t=check&add-lease=true") d.addCallback(_got_html_good) - d.addCallback(self._assert_leasecount, "one", 1) - d.addCallback(self._assert_leasecount, "two", 1) - d.addCallback(self._assert_leasecount, "mutable", 1) + d.addCallback(lambda ign: self._assert_leasecount("one", 1)) + d.addCallback(lambda ign: self._assert_leasecount("two", 1)) + d.addCallback(lambda ign: self._assert_leasecount("mutable", 1)) # this CHECK uses an alternate client, which adds a second lease d.addCallback(self.CHECK, "one", "t=check&add-lease=true", clientnum=1) hunk ./src/allmydata/test/test_web.py 4930 d.addCallback(_got_html_good) - d.addCallback(self._assert_leasecount, "one", 2) - d.addCallback(self._assert_leasecount, "two", 1) - d.addCallback(self._assert_leasecount, "mutable", 1) + d.addCallback(lambda ign: self._assert_leasecount("one", 2)) + d.addCallback(lambda ign: self._assert_leasecount("two", 1)) + d.addCallback(lambda ign: self._assert_leasecount("mutable", 1)) d.addCallback(self.CHECK, "mutable", "t=check&add-lease=true") d.addCallback(_got_html_good) hunk ./src/allmydata/test/test_web.py 4937 - d.addCallback(self._assert_leasecount, "one", 2) - d.addCallback(self._assert_leasecount, "two", 1) - d.addCallback(self._assert_leasecount, "mutable", 1) + d.addCallback(lambda ign: self._assert_leasecount("one", 2)) + d.addCallback(lambda ign: self._assert_leasecount("two", 1)) + d.addCallback(lambda ign: self._assert_leasecount("mutable", 1)) d.addCallback(self.CHECK, "mutable", "t=check&add-lease=true", clientnum=1) hunk ./src/allmydata/test/test_web.py 4945 d.addCallback(_got_html_good) - d.addCallback(self._assert_leasecount, "one", 2) - d.addCallback(self._assert_leasecount, "two", 1) - d.addCallback(self._assert_leasecount, "mutable", 2) + d.addCallback(lambda ign: self._assert_leasecount("one", 2)) + d.addCallback(lambda ign: self._assert_leasecount("two", 1)) + d.addCallback(lambda ign: self._assert_leasecount("mutable", 2)) d.addErrback(self.explain_web_error) return d hunk ./src/allmydata/test/test_web.py 4989 self.failUnlessReallyEqual(len(units), 4+1) d.addCallback(_done) - d.addCallback(self._assert_leasecount, "root", 1) - d.addCallback(self._assert_leasecount, "one", 1) - d.addCallback(self._assert_leasecount, "mutable", 1) + d.addCallback(lambda ign: self._assert_leasecount("root", 1)) + d.addCallback(lambda ign: self._assert_leasecount("one", 1)) + d.addCallback(lambda ign: self._assert_leasecount("mutable", 1)) d.addCallback(self.CHECK, "root", "t=stream-deep-check&add-lease=true") d.addCallback(_done) hunk ./src/allmydata/test/test_web.py 4996 - d.addCallback(self._assert_leasecount, "root", 1) - d.addCallback(self._assert_leasecount, "one", 1) - d.addCallback(self._assert_leasecount, "mutable", 1) + d.addCallback(lambda ign: self._assert_leasecount("root", 1)) + d.addCallback(lambda ign: self._assert_leasecount("one", 1)) + d.addCallback(lambda ign: self._assert_leasecount("mutable", 1)) d.addCallback(self.CHECK, "root", "t=stream-deep-check&add-lease=true", clientnum=1) hunk ./src/allmydata/test/test_web.py 5004 d.addCallback(_done) - d.addCallback(self._assert_leasecount, "root", 2) - d.addCallback(self._assert_leasecount, "one", 2) - d.addCallback(self._assert_leasecount, "mutable", 2) + d.addCallback(lambda ign: self._assert_leasecount("root", 2)) + d.addCallback(lambda ign: self._assert_leasecount("one", 2)) + d.addCallback(lambda ign: self._assert_leasecount("mutable", 2)) d.addErrback(self.explain_web_error) return d } [Fix more shallow bugs, mainly FilePathification. Also, remove the max_space_per_bucket parameter from BucketWriter since it can be obtained from the _max_size attribute of the share (via a new get_allocated_size() accessor). refs #999 david-sarah@jacaranda.org**20110921221421 Ignore-this: 600e3ccef8533aa43442fa576c7d88cf ] { hunk ./src/allmydata/scripts/debug.py 642 /home/warner/testnet/node-2/storage/shares/44k/44kai1tui348689nrw8fjegc8c/2 """ from allmydata.storage.server import si_a2b - from allmydata.storage.backends.disk_backend import si_si2dir + from allmydata.storage.backends.disk.disk_backend import si_si2dir from allmydata.util.encodingutil import quote_filepath out = options.stdout hunk ./src/allmydata/scripts/debug.py 648 si = si_a2b(options.si_s) for nodedir in options.nodedirs: - sharedir = si_si2dir(nodedir.child("storage").child("shares"), si) + sharedir = si_si2dir(FilePath(nodedir).child("storage").child("shares"), si) if sharedir.exists(): for sharefp in sharedir.children(): print >>out, quote_filepath(sharefp, quotemarks=False) hunk ./src/allmydata/storage/backends/disk/disk_backend.py 189 incominghome = self._incominghomedir.child(str(shnum)) immsh = ImmutableDiskShare(self.get_storage_index(), shnum, incominghome, finalhome, max_size=max_space_per_bucket) - bw = BucketWriter(storageserver, immsh, max_space_per_bucket, lease_info, canary) + bw = BucketWriter(storageserver, immsh, lease_info, canary) if self._discard_storage: bw.throw_out_all_data = True return bw hunk ./src/allmydata/storage/backends/disk/immutable.py 147 def unlink(self): self._home.remove() + def get_allocated_size(self): + return self._max_size + def get_size(self): return self._home.getsize() hunk ./src/allmydata/storage/bucket.py 15 class BucketWriter(Referenceable): implements(RIBucketWriter) - def __init__(self, ss, immutableshare, max_size, lease_info, canary): + def __init__(self, ss, immutableshare, lease_info, canary): self.ss = ss hunk ./src/allmydata/storage/bucket.py 17 - self._max_size = max_size # don't allow the client to write more than this self._canary = canary self._disconnect_marker = canary.notifyOnDisconnect(self._disconnected) self.closed = False hunk ./src/allmydata/storage/bucket.py 27 self._share.add_lease(lease_info) def allocated_size(self): - return self._max_size + return self._share.get_allocated_size() def remote_write(self, offset, data): start = time.time() hunk ./src/allmydata/storage/crawler.py 480 self.state["bucket-counts"][cycle] = {} self.state["bucket-counts"][cycle][prefix] = len(sharesets) if prefix in self.prefixes[:self.num_sample_prefixes]: - self.state["storage-index-samples"][prefix] = (cycle, sharesets) + si_strings = [shareset.get_storage_index_string() for shareset in sharesets] + self.state["storage-index-samples"][prefix] = (cycle, si_strings) def finished_cycle(self, cycle): last_counts = self.state["bucket-counts"].get(cycle, []) hunk ./src/allmydata/storage/expirer.py 281 # copy() needs to become a deepcopy h["space-recovered"] = s["space-recovered"].copy() - history = pickle.load(self.historyfp.getContent()) + history = pickle.loads(self.historyfp.getContent()) history[cycle] = h while len(history) > 10: oldcycles = sorted(history.keys()) hunk ./src/allmydata/storage/expirer.py 355 progress = self.get_progress() state = ShareCrawler.get_state(self) # does a shallow copy - history = pickle.load(self.historyfp.getContent()) + history = pickle.loads(self.historyfp.getContent()) state["history"] = history if not progress["cycle-in-progress"]: hunk ./src/allmydata/test/test_download.py 199 for shnum in immutable_shares[clientnum]: if s._shnum == shnum: share_dir = self.get_server(clientnum).backend.get_shareset(si)._sharehomedir - share_dir.child(str(shnum)).remove() + fileutil.fp_remove(share_dir.child(str(shnum))) d.addCallback(_clobber_some_shares) d.addCallback(lambda ign: download_to_data(n)) d.addCallback(_got_data) hunk ./src/allmydata/test/test_download.py 224 for clientnum in immutable_shares: for shnum in immutable_shares[clientnum]: share_dir = self.get_server(clientnum).backend.get_shareset(si)._sharehomedir - share_dir.child(str(shnum)).remove() + fileutil.fp_remove(share_dir.child(str(shnum))) # now a new download should fail with NoSharesError. We want a # new ImmutableFileNode so it will forget about the old shares. # If we merely called create_node_from_uri() without first hunk ./src/allmydata/test/test_repairer.py 415 def _test_corrupt(ignored): olddata = {} shares = self.find_uri_shares(self.uri) - for (shnum, serverid, sharefile) in shares: - olddata[ (shnum, serverid) ] = open(sharefile, "rb").read() + for (shnum, serverid, sharefp) in shares: + olddata[ (shnum, serverid) ] = sharefp.getContent() for sh in shares: self.corrupt_share(sh, common._corrupt_uri_extension) hunk ./src/allmydata/test/test_repairer.py 419 - for (shnum, serverid, sharefile) in shares: - newdata = open(sharefile, "rb").read() + for (shnum, serverid, sharefp) in shares: + newdata = sharefp.getContent() self.failIfEqual(olddata[ (shnum, serverid) ], newdata) d.addCallback(_test_corrupt) hunk ./src/allmydata/test/test_storage.py 63 class Bucket(unittest.TestCase): def make_workdir(self, name): - basedir = os.path.join("storage", "Bucket", name) - incoming = os.path.join(basedir, "tmp", "bucket") - final = os.path.join(basedir, "bucket") - fileutil.make_dirs(basedir) - fileutil.make_dirs(os.path.join(basedir, "tmp")) + basedir = FilePath("storage").child("Bucket").child(name) + tmpdir = basedir.child("tmp") + tmpdir.makedirs() + incoming = tmpdir.child("bucket") + final = basedir.child("bucket") return incoming, final def bucket_writer_closed(self, bw, consumed): hunk ./src/allmydata/test/test_storage.py 87 def test_create(self): incoming, final = self.make_workdir("test_create") - bw = BucketWriter(self, incoming, final, 200, self.make_lease(), - FakeCanary()) + share = ImmutableDiskShare("", 0, incoming, final, 200) + bw = BucketWriter(self, share, self.make_lease(), FakeCanary()) bw.remote_write(0, "a"*25) bw.remote_write(25, "b"*25) bw.remote_write(50, "c"*25) hunk ./src/allmydata/test/test_storage.py 97 def test_readwrite(self): incoming, final = self.make_workdir("test_readwrite") - bw = BucketWriter(self, incoming, final, 200, self.make_lease(), - FakeCanary()) + share = ImmutableDiskShare("", 0, incoming, 200) + bw = BucketWriter(self, share, self.make_lease(), FakeCanary()) bw.remote_write(0, "a"*25) bw.remote_write(25, "b"*25) bw.remote_write(50, "c"*7) # last block may be short hunk ./src/allmydata/test/test_storage.py 140 incoming, final = self.make_workdir("test_read_past_end_of_share_data") - fileutil.write(final, share_file_data) + final.setContent(share_file_data) mockstorageserver = mock.Mock() hunk ./src/allmydata/test/test_storage.py 179 class BucketProxy(unittest.TestCase): def make_bucket(self, name, size): - basedir = os.path.join("storage", "BucketProxy", name) - incoming = os.path.join(basedir, "tmp", "bucket") - final = os.path.join(basedir, "bucket") - fileutil.make_dirs(basedir) - fileutil.make_dirs(os.path.join(basedir, "tmp")) - bw = BucketWriter(self, incoming, final, size, self.make_lease(), - FakeCanary()) + basedir = FilePath("storage").child("BucketProxy").child(name) + tmpdir = basedir.child("tmp") + tmpdir.makedirs() + incoming = tmpdir.child("bucket") + final = basedir.child("bucket") + share = ImmutableDiskShare("", 0, incoming, final, size) + bw = BucketWriter(self, share, self.make_lease(), FakeCanary()) rb = RemoteBucket() rb.target = bw return bw, rb, final hunk ./src/allmydata/test/test_storage.py 206 pass def test_create(self): - bw, rb, sharefname = self.make_bucket("test_create", 500) + bw, rb, sharefp = self.make_bucket("test_create", 500) bp = WriteBucketProxy(rb, None, data_size=300, block_size=10, hunk ./src/allmydata/test/test_storage.py 237 for i in (1,9,13)] uri_extension = "s" + "E"*498 + "e" - bw, rb, sharefname = self.make_bucket(name, sharesize) + bw, rb, sharefp = self.make_bucket(name, sharesize) bp = wbp_class(rb, None, data_size=95, block_size=25, hunk ./src/allmydata/test/test_storage.py 258 # now read everything back def _start_reading(res): - br = BucketReader(self, sharefname) + br = BucketReader(self, sharefp) rb = RemoteBucket() rb.target = br server = NoNetworkServer("abc", None) hunk ./src/allmydata/test/test_storage.py 373 for i, wb in writers.items(): wb.remote_write(0, "%10d" % i) wb.remote_close() - storedir = os.path.join(self.workdir("test_dont_overfill_dirs"), - "shares") - children_of_storedir = set(os.listdir(storedir)) + storedir = self.workdir("test_dont_overfill_dirs").child("shares") + children_of_storedir = sorted([child.basename() for child in storedir.children()]) # Now store another one under another storageindex that has leading # chars the same as the first storageindex. hunk ./src/allmydata/test/test_storage.py 382 for i, wb in writers.items(): wb.remote_write(0, "%10d" % i) wb.remote_close() - storedir = os.path.join(self.workdir("test_dont_overfill_dirs"), - "shares") - new_children_of_storedir = set(os.listdir(storedir)) + storedir = self.workdir("test_dont_overfill_dirs").child("shares") + new_children_of_storedir = sorted([child.basename() for child in storedir.children()]) self.failUnlessEqual(children_of_storedir, new_children_of_storedir) def test_remove_incoming(self): hunk ./src/allmydata/test/test_storage.py 390 ss = self.create("test_remove_incoming") already, writers = self.allocate(ss, "vid", range(3), 10) for i,wb in writers.items(): + incoming_share_home = wb._share._home wb.remote_write(0, "%10d" % i) wb.remote_close() hunk ./src/allmydata/test/test_storage.py 393 - incoming_share_dir = wb.incominghome - incoming_bucket_dir = os.path.dirname(incoming_share_dir) - incoming_prefix_dir = os.path.dirname(incoming_bucket_dir) - incoming_dir = os.path.dirname(incoming_prefix_dir) - self.failIf(os.path.exists(incoming_bucket_dir), incoming_bucket_dir) - self.failIf(os.path.exists(incoming_prefix_dir), incoming_prefix_dir) - self.failUnless(os.path.exists(incoming_dir), incoming_dir) + incoming_bucket_dir = incoming_share_home.parent() + incoming_prefix_dir = incoming_bucket_dir.parent() + incoming_dir = incoming_prefix_dir.parent() + self.failIf(incoming_bucket_dir.exists(), incoming_bucket_dir) + self.failIf(incoming_prefix_dir.exists(), incoming_prefix_dir) + self.failUnless(incoming_dir.exists(), incoming_dir) def test_abort(self): # remote_abort, when called on a writer, should make sure that hunk ./src/allmydata/test/test_upload.py 1849 # remove the storedir, wiping out any existing shares fileutil.fp_remove(storedir) # create an empty storedir to replace the one we just removed - storedir.mkdir() + storedir.makedirs() client = self.g.clients[0] client.DEFAULT_ENCODING_PARAMETERS['happy'] = 4 return client hunk ./src/allmydata/test/test_upload.py 1890 # remove the storedir, wiping out any existing shares fileutil.fp_remove(storedir) # create an empty storedir to replace the one we just removed - storedir.mkdir() + storedir.makedirs() client = self.g.clients[0] client.DEFAULT_ENCODING_PARAMETERS['happy'] = 4 return client } [uri.py: resolve a conflict between trunk and the pluggable-backends patches. refs #999 david-sarah@jacaranda.org**20110921222038 Ignore-this: ffeeab60d8e71a6a29a002d024d76fcf ] { hunk ./src/allmydata/uri.py 829 def is_mutable(self): return False + def is_readonly(self): + return True + + def get_readonly(self): + return self + + class DirectoryURIVerifier(_DirectoryBaseURI): implements(IVerifierURI) hunk ./src/allmydata/uri.py 855 def is_mutable(self): return False + def is_readonly(self): + return True + + def get_readonly(self): + return self + class ImmutableDirectoryURIVerifier(DirectoryURIVerifier): implements(IVerifierURI) } [Fix some more test failures. refs #999 david-sarah@jacaranda.org**20110922045451 Ignore-this: b726193cbd03a7c3d343f6e4a0f33ee7 ] { hunk ./src/allmydata/scripts/debug.py 42 from allmydata.util.encodingutil import quote_output out = options.stdout + filename = options['filename'] # check the version, to see if we have a mutable or immutable share hunk ./src/allmydata/scripts/debug.py 45 - print >>out, "share filename: %s" % quote_output(options['filename']) + print >>out, "share filename: %s" % quote_output(filename) hunk ./src/allmydata/scripts/debug.py 47 - share = get_share("", 0, fp) + share = get_share("", 0, FilePath(filename)) if share.sharetype == "mutable": return dump_mutable_share(options, share) else: hunk ./src/allmydata/storage/backends/disk/mutable.py 85 self.parent = parent # for logging def log(self, *args, **kwargs): - return self.parent.log(*args, **kwargs) + if self.parent: + return self.parent.log(*args, **kwargs) def create(self, serverid, write_enabler): assert not self._home.exists() hunk ./src/allmydata/storage/common.py 6 class DataTooLargeError(Exception): pass -class UnknownMutableContainerVersionError(Exception): +class UnknownContainerVersionError(Exception): pass hunk ./src/allmydata/storage/common.py 9 -class UnknownImmutableContainerVersionError(Exception): +class UnknownMutableContainerVersionError(UnknownContainerVersionError): + pass + +class UnknownImmutableContainerVersionError(UnknownContainerVersionError): pass hunk ./src/allmydata/storage/crawler.py 208 try: state = pickle.loads(self.statefp.getContent()) except EnvironmentError: + if self.statefp.exists(): + raise state = {"version": 1, "last-cycle-finished": None, "current-cycle": None, hunk ./src/allmydata/storage/server.py 24 name = 'storage' LeaseCheckerClass = LeaseCheckingCrawler + BucketCounterClass = BucketCountingCrawler DEFAULT_EXPIRATION_POLICY = { 'enabled': False, 'mode': 'age', hunk ./src/allmydata/storage/server.py 70 def _setup_bucket_counter(self): statefp = self._statedir.child("bucket_counter.state") - self.bucket_counter = BucketCountingCrawler(self.backend, statefp) + self.bucket_counter = self.BucketCounterClass(self.backend, statefp) self.bucket_counter.setServiceParent(self) def _setup_lease_checker(self, expiration_policy): hunk ./src/allmydata/storage/server.py 224 share.add_or_renew_lease(lease_info) alreadygot.add(share.get_shnum()) - for shnum in sharenums - alreadygot: + for shnum in set(sharenums) - alreadygot: if shareset.has_incoming(shnum): # Note that we don't create BucketWriters for shnums that # have a partial share (in incoming/), so if a second upload hunk ./src/allmydata/storage/server.py 247 def remote_add_lease(self, storageindex, renew_secret, cancel_secret, owner_num=1): - # cancel_secret is no longer used. start = time.time() self.count("add-lease") new_expire_time = time.time() + 31*24*60*60 hunk ./src/allmydata/storage/server.py 250 - lease_info = LeaseInfo(owner_num, renew_secret, + lease_info = LeaseInfo(owner_num, renew_secret, cancel_secret, new_expire_time, self._serverid) try: hunk ./src/allmydata/storage/server.py 254 - self.backend.add_or_renew_lease(lease_info) + shareset = self.backend.get_shareset(storageindex) + shareset.add_or_renew_lease(lease_info) finally: self.add_latency("add-lease", time.time() - start) hunk ./src/allmydata/test/test_crawler.py 3 import time -import os.path + from twisted.trial import unittest from twisted.application import service from twisted.internet import defer hunk ./src/allmydata/test/test_crawler.py 10 from twisted.python.filepath import FilePath from foolscap.api import eventually, fireEventually -from allmydata.util import fileutil, hashutil, pollmixin +from allmydata.util import hashutil, pollmixin from allmydata.storage.server import StorageServer, si_b2a from allmydata.storage.crawler import ShareCrawler, TimeSliceExceeded from allmydata.storage.backends.disk.disk_backend import DiskBackend hunk ./src/allmydata/test/test_mutable.py 3024 cso.stderr = StringIO() debug.catalog_shares(cso) shares = cso.stdout.getvalue().splitlines() + self.failIf(len(shares) < 1, shares) oneshare = shares[0] # all shares should be MDMF self.failIf(oneshare.startswith("UNKNOWN"), oneshare) self.failUnless(oneshare.startswith("MDMF"), oneshare) hunk ./src/allmydata/test/test_storage.py 1 -import time, os.path, platform, stat, re, simplejson, struct, shutil, itertools +import time, os.path, platform, re, simplejson, struct, itertools import mock hunk ./src/allmydata/test/test_storage.py 15 from allmydata.util import fileutil, hashutil, base32, pollmixin, time_format from allmydata.storage.server import StorageServer from allmydata.storage.backends.disk.disk_backend import DiskBackend +from allmydata.storage.backends.disk.immutable import ImmutableDiskShare from allmydata.storage.backends.disk.mutable import MutableDiskShare from allmydata.storage.bucket import BucketWriter, BucketReader hunk ./src/allmydata/test/test_storage.py 18 -from allmydata.storage.common import DataTooLargeError, \ +from allmydata.storage.common import DataTooLargeError, UnknownContainerVersionError, \ UnknownMutableContainerVersionError, UnknownImmutableContainerVersionError from allmydata.storage.lease import LeaseInfo from allmydata.storage.crawler import BucketCountingCrawler hunk ./src/allmydata/test/test_storage.py 88 def test_create(self): incoming, final = self.make_workdir("test_create") - share = ImmutableDiskShare("", 0, incoming, final, 200) + share = ImmutableDiskShare("", 0, incoming, final, max_size=200) bw = BucketWriter(self, share, self.make_lease(), FakeCanary()) bw.remote_write(0, "a"*25) bw.remote_write(25, "b"*25) hunk ./src/allmydata/test/test_storage.py 98 def test_readwrite(self): incoming, final = self.make_workdir("test_readwrite") - share = ImmutableDiskShare("", 0, incoming, 200) + share = ImmutableDiskShare("", 0, incoming, final, max_size=200) bw = BucketWriter(self, share, self.make_lease(), FakeCanary()) bw.remote_write(0, "a"*25) bw.remote_write(25, "b"*25) hunk ./src/allmydata/test/test_storage.py 106 bw.remote_close() # now read from it - br = BucketReader(self, bw.finalhome) + br = BucketReader(self, share) self.failUnlessEqual(br.remote_read(0, 25), "a"*25) self.failUnlessEqual(br.remote_read(25, 25), "b"*25) self.failUnlessEqual(br.remote_read(50, 7), "c"*7) hunk ./src/allmydata/test/test_storage.py 131 ownernumber = struct.pack('>L', 0) renewsecret = 'THIS LETS ME RENEW YOUR FILE....' assert len(renewsecret) == 32 - cancelsecret = 'THIS LETS ME KILL YOUR FILE HAHA' + cancelsecret = 'THIS USED TO LET ME KILL YR FILE' assert len(cancelsecret) == 32 expirationtime = struct.pack('>L', 60*60*24*31) # 31 days in seconds hunk ./src/allmydata/test/test_storage.py 142 incoming, final = self.make_workdir("test_read_past_end_of_share_data") final.setContent(share_file_data) + share = ImmutableDiskShare("", 0, final) mockstorageserver = mock.Mock() hunk ./src/allmydata/test/test_storage.py 147 # Now read from it. - br = BucketReader(mockstorageserver, final) + br = BucketReader(mockstorageserver, share) self.failUnlessEqual(br.remote_read(0, len(share_data)), share_data) hunk ./src/allmydata/test/test_storage.py 260 # now read everything back def _start_reading(res): - br = BucketReader(self, sharefp) + share = ImmutableDiskShare("", 0, sharefp) + br = BucketReader(self, share) rb = RemoteBucket() rb.target = br server = NoNetworkServer("abc", None) hunk ./src/allmydata/test/test_storage.py 346 if 'cygwin' in syslow or 'windows' in syslow or 'darwin' in syslow: raise unittest.SkipTest("If your filesystem doesn't support efficient sparse files then it is very expensive (Mac OS X and Windows don't support efficient sparse files).") - avail = fileutil.get_available_space('.', 512*2**20) + avail = fileutil.get_available_space(FilePath('.'), 512*2**20) if avail <= 4*2**30: raise unittest.SkipTest("This test will spuriously fail if you have less than 4 GiB free on your filesystem.") hunk ./src/allmydata/test/test_storage.py 476 w[0].remote_write(0, "\xff"*10) w[0].remote_close() - fp = ss.backend.get_shareset("si1").sharehomedir.child("0") + fp = ss.backend.get_shareset("si1")._sharehomedir.child("0") f = fp.open("rb+") hunk ./src/allmydata/test/test_storage.py 478 - f.seek(0) - f.write(struct.pack(">L", 0)) # this is invalid: minimum used is v1 - f.close() + try: + f.seek(0) + f.write(struct.pack(">L", 0)) # this is invalid: minimum used is v1 + finally: + f.close() ss.remote_get_buckets("allocate") hunk ./src/allmydata/test/test_storage.py 575 def test_seek(self): basedir = self.workdir("test_seek_behavior") - fileutil.make_dirs(basedir) - filename = os.path.join(basedir, "testfile") - f = open(filename, "wb") - f.write("start") - f.close() + basedir.makedirs() + fp = basedir.child("testfile") + fp.setContent("start") + # mode="w" allows seeking-to-create-holes, but truncates pre-existing # files. mode="a" preserves previous contents but does not allow # seeking-to-create-holes. mode="r+" allows both. hunk ./src/allmydata/test/test_storage.py 582 - f = open(filename, "rb+") - f.seek(100) - f.write("100") - f.close() - filelen = os.stat(filename)[stat.ST_SIZE] + f = fp.open("rb+") + try: + f.seek(100) + f.write("100") + finally: + f.close() + fp.restat() + filelen = fp.getsize() self.failUnlessEqual(filelen, 100+3) hunk ./src/allmydata/test/test_storage.py 591 - f2 = open(filename, "rb") - self.failUnlessEqual(f2.read(5), "start") - + f2 = fp.open("rb") + try: + self.failUnlessEqual(f2.read(5), "start") + finally: + f2.close() def test_leases(self): ss = self.create("test_leases") hunk ./src/allmydata/test/test_storage.py 693 def test_readonly(self): workdir = self.workdir("test_readonly") - ss = StorageServer(workdir, "\x00" * 20, readonly_storage=True) + backend = DiskBackend(workdir, readonly=True) + ss = StorageServer("\x00" * 20, backend, workdir) ss.setServiceParent(self.sparent) already,writers = self.allocate(ss, "vid", [0,1,2], 75) hunk ./src/allmydata/test/test_storage.py 710 def test_discard(self): # discard is really only used for other tests, but we test it anyways + # XXX replace this with a null backend test workdir = self.workdir("test_discard") hunk ./src/allmydata/test/test_storage.py 712 - ss = StorageServer(workdir, "\x00" * 20, discard_storage=True) + backend = DiskBackend(workdir, readonly=False, discard_storage=True) + ss = StorageServer("\x00" * 20, backend, workdir) ss.setServiceParent(self.sparent) already,writers = self.allocate(ss, "vid", [0,1,2], 75) hunk ./src/allmydata/test/test_storage.py 731 def test_advise_corruption(self): workdir = self.workdir("test_advise_corruption") - ss = StorageServer(workdir, "\x00" * 20, discard_storage=True) + backend = DiskBackend(workdir, readonly=False, discard_storage=True) + ss = StorageServer("\x00" * 20, backend, workdir) ss.setServiceParent(self.sparent) si0_s = base32.b2a("si0") hunk ./src/allmydata/test/test_storage.py 738 ss.remote_advise_corrupt_share("immutable", "si0", 0, "This share smells funny.\n") - reportdir = os.path.join(workdir, "corruption-advisories") - reports = os.listdir(reportdir) + reportdir = workdir.child("corruption-advisories") + reports = [child.basename() for child in reportdir.children()] self.failUnlessEqual(len(reports), 1) report_si0 = reports[0] hunk ./src/allmydata/test/test_storage.py 742 - self.failUnlessIn(si0_s, report_si0) - f = open(os.path.join(reportdir, report_si0), "r") - report = f.read() - f.close() + self.failUnlessIn(si0_s, str(report_si0)) + report = reportdir.child(report_si0).getContent() + self.failUnlessIn("type: immutable", report) self.failUnlessIn("storage_index: %s" % si0_s, report) self.failUnlessIn("share_number: 0", report) hunk ./src/allmydata/test/test_storage.py 762 self.failUnlessEqual(set(b.keys()), set([1])) b[1].remote_advise_corrupt_share("This share tastes like dust.\n") - reports = os.listdir(reportdir) + reports = [child.basename() for child in reportdir.children()] self.failUnlessEqual(len(reports), 2) hunk ./src/allmydata/test/test_storage.py 764 - report_si1 = [r for r in reports if si1_s in r][0] - f = open(os.path.join(reportdir, report_si1), "r") - report = f.read() - f.close() + report_si1 = [r for r in reports if si1_s in str(r)][0] + report = reportdir.child(report_si1).getContent() + self.failUnlessIn("type: immutable", report) self.failUnlessIn("storage_index: %s" % si1_s, report) self.failUnlessIn("share_number: 1", report) hunk ./src/allmydata/test/test_storage.py 783 return self.sparent.stopService() def workdir(self, name): - basedir = os.path.join("storage", "MutableServer", name) - return basedir + return FilePath("storage").child("MutableServer").child(name) def create(self, name): workdir = self.workdir(name) hunk ./src/allmydata/test/test_storage.py 787 - ss = StorageServer(workdir, "\x00" * 20) + backend = DiskBackend(workdir) + ss = StorageServer("\x00" * 20, backend, workdir) ss.setServiceParent(self.sparent) return ss hunk ./src/allmydata/test/test_storage.py 810 cancel_secret = self.cancel_secret(lease_tag) rstaraw = ss.remote_slot_testv_and_readv_and_writev testandwritev = dict( [ (shnum, ([], [], None) ) - for shnum in sharenums ] ) + for shnum in sharenums ] ) readv = [] rc = rstaraw(storage_index, (write_enabler, renew_secret, cancel_secret), hunk ./src/allmydata/test/test_storage.py 824 def test_bad_magic(self): ss = self.create("test_bad_magic") self.allocate(ss, "si1", "we1", self._lease_secret.next(), set([0]), 10) - fp = ss.backend.get_shareset("si1").sharehomedir.child("0") + fp = ss.backend.get_shareset("si1")._sharehomedir.child("0") f = fp.open("rb+") hunk ./src/allmydata/test/test_storage.py 826 - f.seek(0) - f.write("BAD MAGIC") - f.close() + try: + f.seek(0) + f.write("BAD MAGIC") + finally: + f.close() read = ss.remote_slot_readv hunk ./src/allmydata/test/test_storage.py 832 - e = self.failUnlessRaises(UnknownMutableContainerVersionError, + + # This used to test for UnknownMutableContainerVersionError, + # but the current code raises UnknownImmutableContainerVersionError. + # (It changed because remote_slot_readv now works with either + # mutable or immutable shares.) Since the share file doesn't have + # the mutable magic, it's not clear that this is wrong. + # For now, accept either exception. + e = self.failUnlessRaises(UnknownContainerVersionError, read, "si1", [0], [(0,10)]) hunk ./src/allmydata/test/test_storage.py 841 - self.failUnlessIn(" had magic ", str(e)) + self.failUnlessIn(" had ", str(e)) self.failUnlessIn(" but we wanted ", str(e)) def test_container_size(self): hunk ./src/allmydata/test/test_storage.py 1248 # create a random non-numeric file in the bucket directory, to # exercise the code that's supposed to ignore those. - bucket_dir = ss.backend.get_shareset("si1").sharehomedir + bucket_dir = ss.backend.get_shareset("si1")._sharehomedir bucket_dir.child("ignore_me.txt").setContent("you ought to be ignoring me\n") hunk ./src/allmydata/test/test_storage.py 1251 - s0 = MutableDiskShare(os.path.join(bucket_dir, "0")) + s0 = MutableDiskShare("", 0, bucket_dir.child("0")) self.failUnlessEqual(len(list(s0.get_leases())), 1) # add-lease on a missing storage index is silently ignored hunk ./src/allmydata/test/test_storage.py 1365 # note: this is a detail of the storage server implementation, and # may change in the future prefix = si[:2] - prefixdir = os.path.join(self.workdir("test_remove"), "shares", prefix) - bucketdir = os.path.join(prefixdir, si) - self.failUnless(os.path.exists(prefixdir), prefixdir) - self.failIf(os.path.exists(bucketdir), bucketdir) + prefixdir = self.workdir("test_remove").child("shares").child(prefix) + bucketdir = prefixdir.child(si) + self.failUnless(prefixdir.exists(), prefixdir) + self.failIf(bucketdir.exists(), bucketdir) class MDMFProxies(unittest.TestCase, ShouldFailMixin): hunk ./src/allmydata/test/test_storage.py 1420 def workdir(self, name): - basedir = os.path.join("storage", "MutableServer", name) - return basedir - + return FilePath("storage").child("MDMFProxies").child(name) def create(self, name): workdir = self.workdir(name) hunk ./src/allmydata/test/test_storage.py 1424 - ss = StorageServer(workdir, "\x00" * 20) + backend = DiskBackend(workdir) + ss = StorageServer("\x00" * 20, backend, workdir) ss.setServiceParent(self.sparent) return ss hunk ./src/allmydata/test/test_storage.py 2798 return self.sparent.stopService() def workdir(self, name): - return FilePath("storage").child("Server").child(name) + return FilePath("storage").child("Stats").child(name) def create(self, name): workdir = self.workdir(name) hunk ./src/allmydata/test/test_storage.py 2886 d.callback(None) class MyStorageServer(StorageServer): - def add_bucket_counter(self): - statefile = os.path.join(self.storedir, "bucket_counter.state") - self.bucket_counter = MyBucketCountingCrawler(self, statefile) - self.bucket_counter.setServiceParent(self) + BucketCounterClass = MyBucketCountingCrawler + class BucketCounter(unittest.TestCase, pollmixin.PollMixin): hunk ./src/allmydata/test/test_storage.py 2899 def test_bucket_counter(self): basedir = "storage/BucketCounter/bucket_counter" - fileutil.make_dirs(basedir) - ss = StorageServer(basedir, "\x00" * 20) + fp = FilePath(basedir) + backend = DiskBackend(fp) + ss = StorageServer("\x00" * 20, backend, fp) + # to make sure we capture the bucket-counting-crawler in the middle # of a cycle, we reach in and reduce its maximum slice time to 0. We # also make it start sooner than usual. hunk ./src/allmydata/test/test_storage.py 2958 def test_bucket_counter_cleanup(self): basedir = "storage/BucketCounter/bucket_counter_cleanup" - fileutil.make_dirs(basedir) - ss = StorageServer(basedir, "\x00" * 20) + fp = FilePath(basedir) + backend = DiskBackend(fp) + ss = StorageServer("\x00" * 20, backend, fp) + # to make sure we capture the bucket-counting-crawler in the middle # of a cycle, we reach in and reduce its maximum slice time to 0. ss.bucket_counter.slow_start = 0 hunk ./src/allmydata/test/test_storage.py 3002 def test_bucket_counter_eta(self): basedir = "storage/BucketCounter/bucket_counter_eta" - fileutil.make_dirs(basedir) - ss = MyStorageServer(basedir, "\x00" * 20) + fp = FilePath(basedir) + backend = DiskBackend(fp) + ss = MyStorageServer("\x00" * 20, backend, fp) ss.bucket_counter.slow_start = 0 # these will be fired inside finished_prefix() hooks = ss.bucket_counter.hook_ds = [defer.Deferred() for i in range(3)] hunk ./src/allmydata/test/test_storage.py 3125 def test_basic(self): basedir = "storage/LeaseCrawler/basic" - fileutil.make_dirs(basedir) - ss = InstrumentedStorageServer(basedir, "\x00" * 20) + fp = FilePath(basedir) + backend = DiskBackend(fp) + ss = InstrumentedStorageServer("\x00" * 20, backend, fp) + # make it start sooner than usual. lc = ss.lease_checker lc.slow_start = 0 hunk ./src/allmydata/test/test_storage.py 3141 [immutable_si_0, immutable_si_1, mutable_si_2, mutable_si_3] = self.sis # add a non-sharefile to exercise another code path - fp = ss.backend.get_shareset(immutable_si_0).sharehomedir.child("not-a-share") + fp = ss.backend.get_shareset(immutable_si_0)._sharehomedir.child("not-a-share") fp.setContent("I am not a share.\n") # this is before the crawl has started, so we're not in a cycle yet hunk ./src/allmydata/test/test_storage.py 3264 self.failUnlessEqual(rec["configured-sharebytes"], 0) def _get_sharefile(si): - return list(ss._iter_share_files(si))[0] + return list(ss.backend.get_shareset(si).get_shares())[0] def count_leases(si): return len(list(_get_sharefile(si).get_leases())) self.failUnlessEqual(count_leases(immutable_si_0), 1) hunk ./src/allmydata/test/test_storage.py 3296 for i,lease in enumerate(sf.get_leases()): if lease.renew_secret == renew_secret: lease.expiration_time = new_expire_time - f = open(sf.home, 'rb+') - sf._write_lease_record(f, i, lease) - f.close() + f = sf._home.open('rb+') + try: + sf._write_lease_record(f, i, lease) + finally: + f.close() return raise IndexError("unable to renew non-existent lease") hunk ./src/allmydata/test/test_storage.py 3306 def test_expire_age(self): basedir = "storage/LeaseCrawler/expire_age" - fileutil.make_dirs(basedir) + fp = FilePath(basedir) + backend = DiskBackend(fp) + # setting 'override_lease_duration' to 2000 means that any lease that # is more than 2000 seconds old will be expired. expiration_policy = { hunk ./src/allmydata/test/test_storage.py 3317 'override_lease_duration': 2000, 'sharetypes': ('mutable', 'immutable'), } - ss = InstrumentedStorageServer(basedir, "\x00" * 20, expiration_policy) + ss = InstrumentedStorageServer("\x00" * 20, backend, fp, expiration_policy=expiration_policy) + # make it start sooner than usual. lc = ss.lease_checker lc.slow_start = 0 hunk ./src/allmydata/test/test_storage.py 3330 [immutable_si_0, immutable_si_1, mutable_si_2, mutable_si_3] = self.sis def count_shares(si): - return len(list(ss._iter_share_files(si))) + return len(list(ss.backend.get_shareset(si).get_shares())) def _get_sharefile(si): hunk ./src/allmydata/test/test_storage.py 3332 - return list(ss._iter_share_files(si))[0] + return list(ss.backend.get_shareset(si).get_shares())[0] def count_leases(si): return len(list(_get_sharefile(si).get_leases())) hunk ./src/allmydata/test/test_storage.py 3355 sf0 = _get_sharefile(immutable_si_0) self.backdate_lease(sf0, self.renew_secrets[0], now - 1000) - sf0_size = os.stat(sf0.home).st_size + sf0_size = sf0.get_size() # immutable_si_1 gets an extra lease sf1 = _get_sharefile(immutable_si_1) hunk ./src/allmydata/test/test_storage.py 3363 sf2 = _get_sharefile(mutable_si_2) self.backdate_lease(sf2, self.renew_secrets[3], now - 1000) - sf2_size = os.stat(sf2.home).st_size + sf2_size = sf2.get_size() # mutable_si_3 gets an extra lease sf3 = _get_sharefile(mutable_si_3) hunk ./src/allmydata/test/test_storage.py 3450 def test_expire_cutoff_date(self): basedir = "storage/LeaseCrawler/expire_cutoff_date" - fileutil.make_dirs(basedir) + fp = FilePath(basedir) + backend = DiskBackend(fp) + # setting 'cutoff_date' to 2000 seconds ago means that any lease that # is more than 2000 seconds old will be expired. now = time.time() hunk ./src/allmydata/test/test_storage.py 3463 'cutoff_date': then, 'sharetypes': ('mutable', 'immutable'), } - ss = InstrumentedStorageServer(basedir, "\x00" * 20, expiration_policy) + ss = InstrumentedStorageServer("\x00" * 20, backend, fp, expiration_policy=expiration_policy) + # make it start sooner than usual. lc = ss.lease_checker lc.slow_start = 0 hunk ./src/allmydata/test/test_storage.py 3476 [immutable_si_0, immutable_si_1, mutable_si_2, mutable_si_3] = self.sis def count_shares(si): - return len(list(ss._iter_share_files(si))) + return len(list(ss.backend.get_shareset(si).get_shares())) def _get_sharefile(si): hunk ./src/allmydata/test/test_storage.py 3478 - return list(ss._iter_share_files(si))[0] + return list(ss.backend.get_shareset(si).get_shares())[0] def count_leases(si): return len(list(_get_sharefile(si).get_leases())) hunk ./src/allmydata/test/test_storage.py 3505 sf0 = _get_sharefile(immutable_si_0) self.backdate_lease(sf0, self.renew_secrets[0], new_expiration_time) - sf0_size = os.stat(sf0.home).st_size + sf0_size = sf0.get_size() # immutable_si_1 gets an extra lease sf1 = _get_sharefile(immutable_si_1) hunk ./src/allmydata/test/test_storage.py 3513 sf2 = _get_sharefile(mutable_si_2) self.backdate_lease(sf2, self.renew_secrets[3], new_expiration_time) - sf2_size = os.stat(sf2.home).st_size + sf2_size = sf2.get_size() # mutable_si_3 gets an extra lease sf3 = _get_sharefile(mutable_si_3) hunk ./src/allmydata/test/test_storage.py 3605 def test_only_immutable(self): basedir = "storage/LeaseCrawler/only_immutable" - fileutil.make_dirs(basedir) + fp = FilePath(basedir) + backend = DiskBackend(fp) + # setting 'cutoff_date' to 2000 seconds ago means that any lease that # is more than 2000 seconds old will be expired. now = time.time() hunk ./src/allmydata/test/test_storage.py 3618 'cutoff_date': then, 'sharetypes': ('immutable',), } - ss = StorageServer(basedir, "\x00" * 20, expiration_policy) + ss = StorageServer("\x00" * 20, backend, fp, expiration_policy=expiration_policy) lc = ss.lease_checker lc.slow_start = 0 webstatus = StorageStatus(ss) hunk ./src/allmydata/test/test_storage.py 3629 new_expiration_time = now - 3000 + 31*24*60*60 def count_shares(si): - return len(list(ss._iter_share_files(si))) + return len(list(ss.backend.get_shareset(si).get_shares())) def _get_sharefile(si): hunk ./src/allmydata/test/test_storage.py 3631 - return list(ss._iter_share_files(si))[0] + return list(ss.backend.get_shareset(si).get_shares())[0] def count_leases(si): return len(list(_get_sharefile(si).get_leases())) hunk ./src/allmydata/test/test_storage.py 3668 def test_only_mutable(self): basedir = "storage/LeaseCrawler/only_mutable" - fileutil.make_dirs(basedir) + fp = FilePath(basedir) + backend = DiskBackend(fp) + # setting 'cutoff_date' to 2000 seconds ago means that any lease that # is more than 2000 seconds old will be expired. now = time.time() hunk ./src/allmydata/test/test_storage.py 3681 'cutoff_date': then, 'sharetypes': ('mutable',), } - ss = StorageServer(basedir, "\x00" * 20, expiration_policy) + ss = StorageServer("\x00" * 20, backend, fp, expiration_policy=expiration_policy) lc = ss.lease_checker lc.slow_start = 0 webstatus = StorageStatus(ss) hunk ./src/allmydata/test/test_storage.py 3692 new_expiration_time = now - 3000 + 31*24*60*60 def count_shares(si): - return len(list(ss._iter_share_files(si))) + return len(list(ss.backend.get_shareset(si).get_shares())) def _get_sharefile(si): hunk ./src/allmydata/test/test_storage.py 3694 - return list(ss._iter_share_files(si))[0] + return list(ss.backend.get_shareset(si).get_shares())[0] def count_leases(si): return len(list(_get_sharefile(si).get_leases())) hunk ./src/allmydata/test/test_storage.py 3731 def test_bad_mode(self): basedir = "storage/LeaseCrawler/bad_mode" - fileutil.make_dirs(basedir) + fp = FilePath(basedir) + backend = DiskBackend(fp) + + expiration_policy = { + 'enabled': True, + 'mode': 'bogus', + 'override_lease_duration': None, + 'cutoff_date': None, + 'sharetypes': ('mutable', 'immutable'), + } e = self.failUnlessRaises(ValueError, hunk ./src/allmydata/test/test_storage.py 3742 - StorageServer, basedir, "\x00" * 20, - expiration_mode="bogus") + StorageServer, "\x00" * 20, backend, fp, + expiration_policy=expiration_policy) self.failUnlessIn("GC mode 'bogus' must be 'age' or 'cutoff-date'", str(e)) def test_parse_duration(self): hunk ./src/allmydata/test/test_storage.py 3767 def test_limited_history(self): basedir = "storage/LeaseCrawler/limited_history" - fileutil.make_dirs(basedir) - ss = StorageServer(basedir, "\x00" * 20) + fp = FilePath(basedir) + backend = DiskBackend(fp) + ss = StorageServer("\x00" * 20, backend, fp) + # make it start sooner than usual. lc = ss.lease_checker lc.slow_start = 0 hunk ./src/allmydata/test/test_storage.py 3801 def test_unpredictable_future(self): basedir = "storage/LeaseCrawler/unpredictable_future" - fileutil.make_dirs(basedir) - ss = StorageServer(basedir, "\x00" * 20) + fp = FilePath(basedir) + backend = DiskBackend(fp) + ss = StorageServer("\x00" * 20, backend, fp) + # make it start sooner than usual. lc = ss.lease_checker lc.slow_start = 0 hunk ./src/allmydata/test/test_storage.py 3866 def test_no_st_blocks(self): basedir = "storage/LeaseCrawler/no_st_blocks" - fileutil.make_dirs(basedir) + fp = FilePath(basedir) + backend = DiskBackend(fp) + # A negative 'override_lease_duration' means that the "configured-" # space-recovered counts will be non-zero, since all shares will have # expired by then. hunk ./src/allmydata/test/test_storage.py 3878 'override_lease_duration': -1000, 'sharetypes': ('mutable', 'immutable'), } - ss = No_ST_BLOCKS_StorageServer(basedir, "\x00" * 20, expiration_policy) + ss = No_ST_BLOCKS_StorageServer("\x00" * 20, backend, fp, expiration_policy=expiration_policy) # make it start sooner than usual. lc = ss.lease_checker hunk ./src/allmydata/test/test_storage.py 3911 UnknownImmutableContainerVersionError, ] basedir = "storage/LeaseCrawler/share_corruption" - fileutil.make_dirs(basedir) - ss = InstrumentedStorageServer(basedir, "\x00" * 20) + fp = FilePath(basedir) + backend = DiskBackend(fp) + ss = InstrumentedStorageServer("\x00" * 20, backend, fp) w = StorageStatus(ss) # make it start sooner than usual. lc = ss.lease_checker hunk ./src/allmydata/test/test_storage.py 3928 [immutable_si_0, immutable_si_1, mutable_si_2, mutable_si_3] = self.sis first = min(self.sis) first_b32 = base32.b2a(first) - fp = ss.backend.get_shareset(first).sharehomedir.child("0") + fp = ss.backend.get_shareset(first)._sharehomedir.child("0") f = fp.open("rb+") hunk ./src/allmydata/test/test_storage.py 3930 - f.seek(0) - f.write("BAD MAGIC") - f.close() + try: + f.seek(0) + f.write("BAD MAGIC") + finally: + f.close() # if get_share_file() doesn't see the correct mutable magic, it # assumes the file is an immutable share, and then # immutable.ShareFile sees a bad version. So regardless of which kind hunk ./src/allmydata/test/test_storage.py 3943 # also create an empty bucket empty_si = base32.b2a("\x04"*16) - empty_bucket_dir = ss.backend.get_shareset(empty_si).sharehomedir + empty_bucket_dir = ss.backend.get_shareset(empty_si)._sharehomedir fileutil.fp_make_dirs(empty_bucket_dir) ss.setServiceParent(self.s) hunk ./src/allmydata/test/test_storage.py 4031 def test_status(self): basedir = "storage/WebStatus/status" - fileutil.make_dirs(basedir) - ss = StorageServer(basedir, "\x00" * 20) + fp = FilePath(basedir) + backend = DiskBackend(fp) + ss = StorageServer("\x00" * 20, backend, fp) ss.setServiceParent(self.s) w = StorageStatus(ss) d = self.render1(w) hunk ./src/allmydata/test/test_storage.py 4065 # Some platforms may have no disk stats API. Make sure the code can handle that # (test runs on all platforms). basedir = "storage/WebStatus/status_no_disk_stats" - fileutil.make_dirs(basedir) - ss = StorageServer(basedir, "\x00" * 20) + fp = FilePath(basedir) + backend = DiskBackend(fp) + ss = StorageServer("\x00" * 20, backend, fp) ss.setServiceParent(self.s) w = StorageStatus(ss) html = w.renderSynchronously() hunk ./src/allmydata/test/test_storage.py 4085 # If the API to get disk stats exists but a call to it fails, then the status should # show that no shares will be accepted, and get_available_space() should be 0. basedir = "storage/WebStatus/status_bad_disk_stats" - fileutil.make_dirs(basedir) - ss = StorageServer(basedir, "\x00" * 20) + fp = FilePath(basedir) + backend = DiskBackend(fp) + ss = StorageServer("\x00" * 20, backend, fp) ss.setServiceParent(self.s) w = StorageStatus(ss) html = w.renderSynchronously() } [Fix most of the crawler tests. refs #999 david-sarah@jacaranda.org**20110922183008 Ignore-this: 116c0848008f3989ba78d87c07ec783c ] { hunk ./src/allmydata/storage/backends/disk/disk_backend.py 160 self._discard_storage = discard_storage def get_overhead(self): - return (fileutil.get_disk_usage(self._sharehomedir) + - fileutil.get_disk_usage(self._incominghomedir)) + return (fileutil.get_used_space(self._sharehomedir) + + fileutil.get_used_space(self._incominghomedir)) def get_shares(self): """ hunk ./src/allmydata/storage/crawler.py 2 -import time, struct -import cPickle as pickle +import time, pickle, struct from twisted.internet import reactor from twisted.application import service hunk ./src/allmydata/storage/crawler.py 205 # shareset to be processed, or None if we # are sleeping between cycles try: - state = pickle.loads(self.statefp.getContent()) + pickled = self.statefp.getContent() except EnvironmentError: if self.statefp.exists(): raise hunk ./src/allmydata/storage/crawler.py 215 "last-complete-prefix": None, "last-complete-bucket": None, } + else: + state = pickle.loads(pickled) + state.setdefault("current-cycle-start-time", time.time()) # approximate self.state = state lcp = state["last-complete-prefix"] hunk ./src/allmydata/storage/crawler.py 246 else: last_complete_prefix = self.prefixes[lcpi] self.state["last-complete-prefix"] = last_complete_prefix - self.statefp.setContent(pickle.dumps(self.state)) + pickled = pickle.dumps(self.state) + self.statefp.setContent(pickled) def startService(self): # arrange things to look like we were just sleeping, so hunk ./src/allmydata/storage/expirer.py 86 # initialize history if not self.historyfp.exists(): history = {} # cyclenum -> dict - self.historyfp.setContent(pickle.dumps(history)) + pickled = pickle.dumps(history) + self.historyfp.setContent(pickled) def create_empty_cycle_dict(self): recovered = self.create_empty_recovered_dict() hunk ./src/allmydata/storage/expirer.py 111 def started_cycle(self, cycle): self.state["cycle-to-date"] = self.create_empty_cycle_dict() - def process_storage_index(self, cycle, prefix, container): + def process_shareset(self, cycle, prefix, shareset): would_keep_shares = [] wks = None hunk ./src/allmydata/storage/expirer.py 114 - sharetype = None hunk ./src/allmydata/storage/expirer.py 115 - for share in container.get_shares(): - sharetype = share.sharetype + for share in shareset.get_shares(): try: wks = self.process_share(share) except (UnknownMutableContainerVersionError, hunk ./src/allmydata/storage/expirer.py 128 wks = (1, 1, 1, "unknown") would_keep_shares.append(wks) - container_type = None + shareset_type = None if wks: hunk ./src/allmydata/storage/expirer.py 130 - # use the last share's sharetype as the container type - container_type = wks[3] + # use the last share's type as the shareset type + shareset_type = wks[3] rec = self.state["cycle-to-date"]["space-recovered"] self.increment(rec, "examined-buckets", 1) hunk ./src/allmydata/storage/expirer.py 134 - if sharetype: - self.increment(rec, "examined-buckets-"+container_type, 1) + if shareset_type: + self.increment(rec, "examined-buckets-"+shareset_type, 1) hunk ./src/allmydata/storage/expirer.py 137 - container_diskbytes = container.get_overhead() + shareset_diskbytes = shareset.get_overhead() if sum([wks[0] for wks in would_keep_shares]) == 0: hunk ./src/allmydata/storage/expirer.py 140 - self.increment_container_space("original", container_diskbytes, sharetype) + self.increment_shareset_space("original", shareset_diskbytes, shareset_type) if sum([wks[1] for wks in would_keep_shares]) == 0: hunk ./src/allmydata/storage/expirer.py 142 - self.increment_container_space("configured", container_diskbytes, sharetype) + self.increment_shareset_space("configured", shareset_diskbytes, shareset_type) if sum([wks[2] for wks in would_keep_shares]) == 0: hunk ./src/allmydata/storage/expirer.py 144 - self.increment_container_space("actual", container_diskbytes, sharetype) + self.increment_shareset_space("actual", shareset_diskbytes, shareset_type) def process_share(self, share): sharetype = share.sharetype hunk ./src/allmydata/storage/expirer.py 189 so_far = self.state["cycle-to-date"] self.increment(so_far["leases-per-share-histogram"], num_leases, 1) - self.increment_space("examined", diskbytes, sharetype) + self.increment_space("examined", sharebytes, diskbytes, sharetype) would_keep_share = [1, 1, 1, sharetype] hunk ./src/allmydata/storage/expirer.py 220 self.increment(so_far_sr, a+"-sharebytes-"+sharetype, sharebytes) self.increment(so_far_sr, a+"-diskbytes-"+sharetype, diskbytes) - def increment_container_space(self, a, container_diskbytes, container_type): + def increment_shareset_space(self, a, shareset_diskbytes, shareset_type): rec = self.state["cycle-to-date"]["space-recovered"] hunk ./src/allmydata/storage/expirer.py 222 - self.increment(rec, a+"-diskbytes", container_diskbytes) + self.increment(rec, a+"-diskbytes", shareset_diskbytes) self.increment(rec, a+"-buckets", 1) hunk ./src/allmydata/storage/expirer.py 224 - if container_type: - self.increment(rec, a+"-diskbytes-"+container_type, container_diskbytes) - self.increment(rec, a+"-buckets-"+container_type, 1) + if shareset_type: + self.increment(rec, a+"-diskbytes-"+shareset_type, shareset_diskbytes) + self.increment(rec, a+"-buckets-"+shareset_type, 1) def increment(self, d, k, delta=1): if k not in d: hunk ./src/allmydata/storage/expirer.py 280 # copy() needs to become a deepcopy h["space-recovered"] = s["space-recovered"].copy() - history = pickle.loads(self.historyfp.getContent()) + pickled = self.historyfp.getContent() + history = pickle.loads(pickled) history[cycle] = h while len(history) > 10: oldcycles = sorted(history.keys()) hunk ./src/allmydata/storage/expirer.py 286 del history[oldcycles[0]] - self.historyfp.setContent(pickle.dumps(history)) + repickled = pickle.dumps(history) + self.historyfp.setContent(repickled) def get_state(self): """In addition to the crawler state described in hunk ./src/allmydata/storage/expirer.py 356 progress = self.get_progress() state = ShareCrawler.get_state(self) # does a shallow copy - history = pickle.loads(self.historyfp.getContent()) + pickled = self.historyfp.getContent() + history = pickle.loads(pickled) state["history"] = history if not progress["cycle-in-progress"]: hunk ./src/allmydata/test/test_crawler.py 25 ShareCrawler.__init__(self, *args, **kwargs) self.all_buckets = [] self.finished_d = defer.Deferred() - def process_bucket(self, cycle, prefix, prefixdir, storage_index_b32): - self.all_buckets.append(storage_index_b32) + + def process_shareset(self, cycle, prefix, shareset): + self.all_buckets.append(shareset.get_storage_index_string()) + def finished_cycle(self, cycle): eventually(self.finished_d.callback, None) hunk ./src/allmydata/test/test_crawler.py 41 self.all_buckets = [] self.finished_d = defer.Deferred() self.yield_cb = None - def process_bucket(self, cycle, prefix, prefixdir, storage_index_b32): - self.all_buckets.append(storage_index_b32) + + def process_shareset(self, cycle, prefix, shareset): + self.all_buckets.append(shareset.get_storage_index_string()) self.countdown -= 1 if self.countdown == 0: # force a timeout. We restore it in yielding() hunk ./src/allmydata/test/test_crawler.py 66 self.accumulated = 0.0 self.cycles = 0 self.last_yield = 0.0 - def process_bucket(self, cycle, prefix, prefixdir, storage_index_b32): + + def process_shareset(self, cycle, prefix, shareset): start = time.time() time.sleep(0.05) elapsed = time.time() - start hunk ./src/allmydata/test/test_crawler.py 85 ShareCrawler.__init__(self, *args, **kwargs) self.counter = 0 self.finished_d = defer.Deferred() - def process_bucket(self, cycle, prefix, prefixdir, storage_index_b32): + + def process_shareset(self, cycle, prefix, shareset): self.counter += 1 def finished_cycle(self, cycle): self.finished_d.callback(None) hunk ./src/allmydata/test/test_storage.py 3041 class InstrumentedLeaseCheckingCrawler(LeaseCheckingCrawler): stop_after_first_bucket = False - def process_bucket(self, *args, **kwargs): - LeaseCheckingCrawler.process_bucket(self, *args, **kwargs) + + def process_shareset(self, cycle, prefix, shareset): + LeaseCheckingCrawler.process_shareset(self, cycle, prefix, shareset) if self.stop_after_first_bucket: self.stop_after_first_bucket = False self.cpu_slice = -1.0 hunk ./src/allmydata/test/test_storage.py 3051 if not self.stop_after_first_bucket: self.cpu_slice = 500 +class InstrumentedStorageServer(StorageServer): + LeaseCheckerClass = InstrumentedLeaseCheckingCrawler + + class BrokenStatResults: pass class No_ST_BLOCKS_LeaseCheckingCrawler(LeaseCheckingCrawler): hunk ./src/allmydata/test/test_storage.py 3069 setattr(bsr, attrname, getattr(s, attrname)) return bsr -class InstrumentedStorageServer(StorageServer): - LeaseCheckerClass = InstrumentedLeaseCheckingCrawler class No_ST_BLOCKS_StorageServer(StorageServer): LeaseCheckerClass = No_ST_BLOCKS_LeaseCheckingCrawler } [Reinstate the cancel_lease methods of ImmutableDiskShare and MutableDiskShare, since they are needed for lease expiry. refs #999 david-sarah@jacaranda.org**20110922183323 Ignore-this: a11fb0dd0078ff627cb727fc769ec848 ] { hunk ./src/allmydata/storage/backends/disk/immutable.py 260 except IndexError: self.add_lease(lease_info) + def cancel_lease(self, cancel_secret): + """Remove a lease with the given cancel_secret. If the last lease is + cancelled, the file will be removed. Return the number of bytes that + were freed (by truncating the list of leases, and possibly by + deleting the file). Raise IndexError if there was no lease with the + given cancel_secret. + """ + + leases = list(self.get_leases()) + num_leases_removed = 0 + for i, lease in enumerate(leases): + if constant_time_compare(lease.cancel_secret, cancel_secret): + leases[i] = None + num_leases_removed += 1 + if not num_leases_removed: + raise IndexError("unable to find matching lease to cancel") + + space_freed = 0 + if num_leases_removed: + # pack and write out the remaining leases. We write these out in + # the same order as they were added, so that if we crash while + # doing this, we won't lose any non-cancelled leases. + leases = [l for l in leases if l] # remove the cancelled leases + if len(leases) > 0: + f = self._home.open('rb+') + try: + for i, lease in enumerate(leases): + self._write_lease_record(f, i, lease) + self._write_num_leases(f, len(leases)) + self._truncate_leases(f, len(leases)) + finally: + f.close() + space_freed = self.LEASE_SIZE * num_leases_removed + else: + space_freed = fileutil.get_used_space(self._home) + self.unlink() + return space_freed + hunk ./src/allmydata/storage/backends/disk/mutable.py 361 except IndexError: self.add_lease(lease_info) + def cancel_lease(self, cancel_secret): + """Remove any leases with the given cancel_secret. If the last lease + is cancelled, the file will be removed. Return the number of bytes + that were freed (by truncating the list of leases, and possibly by + deleting the file). Raise IndexError if there was no lease with the + given cancel_secret.""" + + # XXX can this be more like ImmutableDiskShare.cancel_lease? + + accepting_nodeids = set() + modified = 0 + remaining = 0 + blank_lease = LeaseInfo(owner_num=0, + renew_secret="\x00"*32, + cancel_secret="\x00"*32, + expiration_time=0, + nodeid="\x00"*20) + f = self._home.open('rb+') + try: + for (leasenum, lease) in self._enumerate_leases(f): + accepting_nodeids.add(lease.nodeid) + if constant_time_compare(lease.cancel_secret, cancel_secret): + self._write_lease_record(f, leasenum, blank_lease) + modified += 1 + else: + remaining += 1 + if modified: + freed_space = self._pack_leases(f) + finally: + f.close() + + if modified > 0: + if remaining == 0: + freed_space = fileutil.get_used_space(self._home) + self.unlink() + return freed_space + + msg = ("Unable to cancel non-existent lease. I have leases " + "accepted by nodeids: ") + msg += ",".join([("'%s'" % idlib.nodeid_b2a(anid)) + for anid in accepting_nodeids]) + msg += " ." + raise IndexError(msg) + + def _pack_leases(self, f): + # TODO: reclaim space from cancelled leases + return 0 + def _read_write_enabler_and_nodeid(self, f): f.seek(0) data = f.read(self.HEADER_SIZE) } [Blank line cleanups. david-sarah@jacaranda.org**20110923012044 Ignore-this: 8e1c4ecb5b0c65673af35872876a8591 ] { hunk ./src/allmydata/interfaces.py 33 LeaseRenewSecret = Hash # used to protect lease renewal requests LeaseCancelSecret = Hash # used to protect lease cancellation requests + class RIStubClient(RemoteInterface): """Each client publishes a service announcement for a dummy object called the StubClient. This object doesn't actually offer any services, but the hunk ./src/allmydata/interfaces.py 42 the grid and the client versions in use). This is the (empty) RemoteInterface for the StubClient.""" + class RIBucketWriter(RemoteInterface): """ Objects of this kind live on the server side. """ def write(offset=Offset, data=ShareData): hunk ./src/allmydata/interfaces.py 61 """ return None + class RIBucketReader(RemoteInterface): def read(offset=Offset, length=ReadSize): return ShareData hunk ./src/allmydata/interfaces.py 78 documentation. """ + TestVector = ListOf(TupleOf(Offset, ReadSize, str, str)) # elements are (offset, length, operator, specimen) # operator is one of "lt, le, eq, ne, ge, gt" hunk ./src/allmydata/interfaces.py 95 ReadData = ListOf(ShareData) # returns data[offset:offset+length] for each element of TestVector + class RIStorageServer(RemoteInterface): __remote_name__ = "RIStorageServer.tahoe.allmydata.com" hunk ./src/allmydata/interfaces.py 2255 def get_storage_index(): """Return a string with the (binary) storage index.""" + def get_storage_index_string(): """Return a string with the (printable) abbreviated storage index.""" hunk ./src/allmydata/interfaces.py 2258 + def get_uri(): """Return the (string) URI of the object that was checked.""" hunk ./src/allmydata/interfaces.py 2353 def get_report(): """Return a list of strings with more detailed results.""" + class ICheckAndRepairResults(Interface): """I contain the detailed results of a check/verify/repair operation. hunk ./src/allmydata/interfaces.py 2363 def get_storage_index(): """Return a string with the (binary) storage index.""" + def get_storage_index_string(): """Return a string with the (printable) abbreviated storage index.""" hunk ./src/allmydata/interfaces.py 2366 + def get_repair_attempted(): """Return a boolean, True if a repair was attempted. We might not attempt to repair the file because it was healthy, or healthy enough hunk ./src/allmydata/interfaces.py 2372 (i.e. some shares were missing but not enough to exceed some threshold), or because we don't know how to repair this object.""" + def get_repair_successful(): """Return a boolean, True if repair was attempted and the file/dir was fully healthy afterwards. False if no repair was attempted or if hunk ./src/allmydata/interfaces.py 2377 a repair attempt failed.""" + def get_pre_repair_results(): """Return an ICheckResults instance that describes the state of the file/dir before any repair was attempted.""" hunk ./src/allmydata/interfaces.py 2381 + def get_post_repair_results(): """Return an ICheckResults instance that describes the state of the file/dir after any repair was attempted. If no repair was attempted, hunk ./src/allmydata/interfaces.py 2615 (childnode, metadata_dict) tuples), the directory will be populated with those children, otherwise it will be empty.""" + class IClientStatus(Interface): def list_all_uploads(): """Return a list of uploader objects, one for each upload that hunk ./src/allmydata/interfaces.py 2621 currently has an object available (tracked with weakrefs). This is intended for debugging purposes.""" + def list_active_uploads(): """Return a list of active IUploadStatus objects.""" hunk ./src/allmydata/interfaces.py 2624 + def list_recent_uploads(): """Return a list of IUploadStatus objects for the most recently started uploads.""" hunk ./src/allmydata/interfaces.py 2633 """Return a list of downloader objects, one for each download that currently has an object available (tracked with weakrefs). This is intended for debugging purposes.""" + def list_active_downloads(): """Return a list of active IDownloadStatus objects.""" hunk ./src/allmydata/interfaces.py 2636 + def list_recent_downloads(): """Return a list of IDownloadStatus objects for the most recently started downloads.""" hunk ./src/allmydata/interfaces.py 2641 + class IUploadStatus(Interface): def get_started(): """Return a timestamp (float with seconds since epoch) indicating hunk ./src/allmydata/interfaces.py 2646 when the operation was started.""" + def get_storage_index(): """Return a string with the (binary) storage index in use on this upload. Returns None if the storage index has not yet been hunk ./src/allmydata/interfaces.py 2651 calculated.""" + def get_size(): """Return an integer with the number of bytes that will eventually be uploaded for this file. Returns None if the size is not yet known. hunk ./src/allmydata/interfaces.py 2656 """ + def using_helper(): """Return True if this upload is using a Helper, False if not.""" hunk ./src/allmydata/interfaces.py 2659 + def get_status(): """Return a string describing the current state of the upload process.""" hunk ./src/allmydata/interfaces.py 2663 + def get_progress(): """Returns a tuple of floats, (chk, ciphertext, encode_and_push), each from 0.0 to 1.0 . 'chk' describes how much progress has been hunk ./src/allmydata/interfaces.py 2675 process has finished: for helper uploads this is dependent upon the helper providing progress reports. It might be reasonable to add all three numbers and report the sum to the user.""" + def get_active(): """Return True if the upload is currently active, False if not.""" hunk ./src/allmydata/interfaces.py 2678 + def get_results(): """Return an instance of UploadResults (which contains timing and sharemap information). Might return None if the upload is not yet hunk ./src/allmydata/interfaces.py 2683 finished.""" + def get_counter(): """Each upload status gets a unique number: this method returns that number. This provides a handle to this particular upload, so a web hunk ./src/allmydata/interfaces.py 2689 page can generate a suitable hyperlink.""" + class IDownloadStatus(Interface): def get_started(): """Return a timestamp (float with seconds since epoch) indicating hunk ./src/allmydata/interfaces.py 2694 when the operation was started.""" + def get_storage_index(): """Return a string with the (binary) storage index in use on this download. This may be None if there is no storage index (i.e. LIT hunk ./src/allmydata/interfaces.py 2699 files).""" + def get_size(): """Return an integer with the number of bytes that will eventually be retrieved for this file. Returns None if the size is not yet known. hunk ./src/allmydata/interfaces.py 2704 """ + def using_helper(): """Return True if this download is using a Helper, False if not.""" hunk ./src/allmydata/interfaces.py 2707 + def get_status(): """Return a string describing the current state of the download process.""" hunk ./src/allmydata/interfaces.py 2711 + def get_progress(): """Returns a float (from 0.0 to 1.0) describing the amount of the download that has completed. This value will remain at 0.0 until the hunk ./src/allmydata/interfaces.py 2716 first byte of plaintext is pushed to the download target.""" + def get_active(): """Return True if the download is currently active, False if not.""" hunk ./src/allmydata/interfaces.py 2719 + def get_counter(): """Each download status gets a unique number: this method returns that number. This provides a handle to this particular download, so a hunk ./src/allmydata/interfaces.py 2725 web page can generate a suitable hyperlink.""" + class IServermapUpdaterStatus(Interface): pass hunk ./src/allmydata/interfaces.py 2728 + + class IPublishStatus(Interface): pass hunk ./src/allmydata/interfaces.py 2732 + + class IRetrieveStatus(Interface): pass hunk ./src/allmydata/interfaces.py 2737 + class NotCapableError(Exception): """You have tried to write to a read-only node.""" hunk ./src/allmydata/interfaces.py 2741 + class BadWriteEnablerError(Exception): pass hunk ./src/allmydata/interfaces.py 2745 -class RIControlClient(RemoteInterface): hunk ./src/allmydata/interfaces.py 2746 +class RIControlClient(RemoteInterface): def wait_for_client_connections(num_clients=int): """Do not return until we have connections to at least NUM_CLIENTS storage servers. hunk ./src/allmydata/interfaces.py 2801 return DictOf(str, float) + UploadResults = Any() #DictOf(str, str) hunk ./src/allmydata/interfaces.py 2804 + class RIEncryptedUploadable(RemoteInterface): __remote_name__ = "RIEncryptedUploadable.tahoe.allmydata.com" hunk ./src/allmydata/interfaces.py 2877 """ return DictOf(str, DictOf(str, ChoiceOf(float, int, long, None))) + class RIStatsGatherer(RemoteInterface): __remote_name__ = "RIStatsGatherer.tahoe.allmydata.com" """ hunk ./src/allmydata/interfaces.py 2917 class FileTooLargeError(Exception): pass + class IValidatedThingProxy(Interface): def start(): """ Acquire a thing and validate it. Return a deferred that is hunk ./src/allmydata/interfaces.py 2924 eventually fired with self if the thing is valid or errbacked if it can't be acquired or validated.""" + class InsufficientVersionError(Exception): def __init__(self, needed, got): self.needed = needed hunk ./src/allmydata/interfaces.py 2933 return "InsufficientVersionError(need '%s', got %s)" % (self.needed, self.got) + class EmptyPathnameComponentError(Exception): """The webapi disallows empty pathname components.""" hunk ./src/allmydata/test/test_crawler.py 21 class BucketEnumeratingCrawler(ShareCrawler): cpu_slice = 500 # make sure it can complete in a single slice slow_start = 0 + def __init__(self, *args, **kwargs): ShareCrawler.__init__(self, *args, **kwargs) self.all_buckets = [] hunk ./src/allmydata/test/test_crawler.py 33 def finished_cycle(self, cycle): eventually(self.finished_d.callback, None) + class PacedCrawler(ShareCrawler): cpu_slice = 500 # make sure it can complete in a single slice slow_start = 0 hunk ./src/allmydata/test/test_crawler.py 37 + def __init__(self, *args, **kwargs): ShareCrawler.__init__(self, *args, **kwargs) self.countdown = 6 hunk ./src/allmydata/test/test_crawler.py 51 if self.countdown == 0: # force a timeout. We restore it in yielding() self.cpu_slice = -1.0 + def yielding(self, sleep_time): self.cpu_slice = 500 if self.yield_cb: hunk ./src/allmydata/test/test_crawler.py 56 self.yield_cb() + def finished_cycle(self, cycle): eventually(self.finished_d.callback, None) hunk ./src/allmydata/test/test_crawler.py 60 + class ConsumingCrawler(ShareCrawler): cpu_slice = 0.5 allowed_cpu_percentage = 0.5 hunk ./src/allmydata/test/test_crawler.py 79 elapsed = time.time() - start self.accumulated += elapsed self.last_yield += elapsed + def finished_cycle(self, cycle): self.cycles += 1 hunk ./src/allmydata/test/test_crawler.py 82 + def yielding(self, sleep_time): self.last_yield = 0.0 hunk ./src/allmydata/test/test_crawler.py 86 + class OneShotCrawler(ShareCrawler): cpu_slice = 500 # make sure it can complete in a single slice slow_start = 0 hunk ./src/allmydata/test/test_crawler.py 90 + def __init__(self, *args, **kwargs): ShareCrawler.__init__(self, *args, **kwargs) self.counter = 0 hunk ./src/allmydata/test/test_crawler.py 98 def process_shareset(self, cycle, prefix, shareset): self.counter += 1 + def finished_cycle(self, cycle): self.finished_d.callback(None) self.disownServiceParent() hunk ./src/allmydata/test/test_crawler.py 103 + class Basic(unittest.TestCase, StallMixin, pollmixin.PollMixin): def setUp(self): self.s = service.MultiService() hunk ./src/allmydata/test/test_crawler.py 114 def si(self, i): return hashutil.storage_index_hash(str(i)) + def rs(self, i, serverid): return hashutil.bucket_renewal_secret_hash(str(i), serverid) hunk ./src/allmydata/test/test_crawler.py 117 + def cs(self, i, serverid): return hashutil.bucket_cancel_secret_hash(str(i), serverid) hunk ./src/allmydata/test/test_storage.py 39 from allmydata.test.no_network import NoNetworkServer from allmydata.web.storage import StorageStatus, remove_prefix + class Marker: pass hunk ./src/allmydata/test/test_storage.py 42 + + class FakeCanary: def __init__(self, ignore_disconnectors=False): self.ignore = ignore_disconnectors hunk ./src/allmydata/test/test_storage.py 59 return del self.disconnectors[marker] + class FakeStatsProvider: def count(self, name, delta=1): pass hunk ./src/allmydata/test/test_storage.py 66 def register_producer(self, producer): pass + class Bucket(unittest.TestCase): def make_workdir(self, name): basedir = FilePath("storage").child("Bucket").child(name) hunk ./src/allmydata/test/test_storage.py 165 result_of_read = br.remote_read(0, len(share_data)+1) self.failUnlessEqual(result_of_read, share_data) + class RemoteBucket: def __init__(self): hunk ./src/allmydata/test/test_storage.py 309 return self._do_test_readwrite("test_readwrite_v2", 0x44, WriteBucketProxy_v2, ReadBucketProxy) + class Server(unittest.TestCase): def setUp(self): hunk ./src/allmydata/test/test_storage.py 780 self.failUnlessIn("This share tastes like dust.", report) - class MutableServer(unittest.TestCase): def setUp(self): hunk ./src/allmydata/test/test_storage.py 1407 # header. self.salt_hash_tree_s = self.serialize_blockhashes(self.salt_hash_tree[1:]) - def tearDown(self): self.sparent.stopService() fileutil.fp_remove(self.workdir("MDMFProxies storage test server")) hunk ./src/allmydata/test/test_storage.py 1411 - def write_enabler(self, we_tag): return hashutil.tagged_hash("we_blah", we_tag) hunk ./src/allmydata/test/test_storage.py 1414 - def renew_secret(self, tag): return hashutil.tagged_hash("renew_blah", str(tag)) hunk ./src/allmydata/test/test_storage.py 1417 - def cancel_secret(self, tag): return hashutil.tagged_hash("cancel_blah", str(tag)) hunk ./src/allmydata/test/test_storage.py 1420 - def workdir(self, name): return FilePath("storage").child("MDMFProxies").child(name) hunk ./src/allmydata/test/test_storage.py 1430 ss.setServiceParent(self.sparent) return ss - def build_test_mdmf_share(self, tail_segment=False, empty=False): # Start with the checkstring data = struct.pack(">BQ32s", hunk ./src/allmydata/test/test_storage.py 1527 data += self.block_hash_tree_s return data - def write_test_share_to_server(self, storage_index, tail_segment=False, hunk ./src/allmydata/test/test_storage.py 1548 results = write(storage_index, self.secrets, tws, readv) self.failUnless(results[0]) - def build_test_sdmf_share(self, empty=False): if empty: sharedata = "" hunk ./src/allmydata/test/test_storage.py 1598 self.offsets['EOF'] = eof_offset return final_share - def write_sdmf_share_to_server(self, storage_index, empty=False): hunk ./src/allmydata/test/test_storage.py 1613 results = write(storage_index, self.secrets, tws, readv) self.failUnless(results[0]) - def test_read(self): self.write_test_share_to_server("si1") mr = MDMFSlotReadProxy(self.rref, "si1", 0) hunk ./src/allmydata/test/test_storage.py 1682 self.failUnlessEqual(checkstring, checkstring)) return d - def test_read_with_different_tail_segment_size(self): self.write_test_share_to_server("si1", tail_segment=True) mr = MDMFSlotReadProxy(self.rref, "si1", 0) hunk ./src/allmydata/test/test_storage.py 1693 d.addCallback(_check_tail_segment) return d - def test_get_block_with_invalid_segnum(self): self.write_test_share_to_server("si1") mr = MDMFSlotReadProxy(self.rref, "si1", 0) hunk ./src/allmydata/test/test_storage.py 1703 mr.get_block_and_salt, 7)) return d - def test_get_encoding_parameters_first(self): self.write_test_share_to_server("si1") mr = MDMFSlotReadProxy(self.rref, "si1", 0) hunk ./src/allmydata/test/test_storage.py 1715 d.addCallback(_check_encoding_parameters) return d - def test_get_seqnum_first(self): self.write_test_share_to_server("si1") mr = MDMFSlotReadProxy(self.rref, "si1", 0) hunk ./src/allmydata/test/test_storage.py 1723 self.failUnlessEqual(seqnum, 0)) return d - def test_get_root_hash_first(self): self.write_test_share_to_server("si1") mr = MDMFSlotReadProxy(self.rref, "si1", 0) hunk ./src/allmydata/test/test_storage.py 1731 self.failUnlessEqual(root_hash, self.root_hash)) return d - def test_get_checkstring_first(self): self.write_test_share_to_server("si1") mr = MDMFSlotReadProxy(self.rref, "si1", 0) hunk ./src/allmydata/test/test_storage.py 1739 self.failUnlessEqual(checkstring, self.checkstring)) return d - def test_write_read_vectors(self): # When writing for us, the storage server will return to us a # read vector, along with its result. If a write fails because hunk ./src/allmydata/test/test_storage.py 1777 # The checkstring remains the same for the rest of the process. return d - def test_private_key_after_share_hash_chain(self): mw = self._make_new_mw("si1", 0) d = defer.succeed(None) hunk ./src/allmydata/test/test_storage.py 1795 mw.put_encprivkey, self.encprivkey)) return d - def test_signature_after_verification_key(self): mw = self._make_new_mw("si1", 0) d = defer.succeed(None) hunk ./src/allmydata/test/test_storage.py 1821 mw.put_signature, self.signature)) return d - def test_uncoordinated_write(self): # Make two mutable writers, both pointing to the same storage # server, both at the same storage index, and try writing to the hunk ./src/allmydata/test/test_storage.py 1853 d.addCallback(_check_failure) return d - def test_invalid_salt_size(self): # Salts need to be 16 bytes in size. Writes that attempt to # write more or less than this should be rejected. hunk ./src/allmydata/test/test_storage.py 1871 another_invalid_salt)) return d - def test_write_test_vectors(self): # If we give the write proxy a bogus test vector at # any point during the process, it should fail to write when we hunk ./src/allmydata/test/test_storage.py 1904 d.addCallback(_check_success) return d - def serialize_blockhashes(self, blockhashes): return "".join(blockhashes) hunk ./src/allmydata/test/test_storage.py 1907 - def serialize_sharehashes(self, sharehashes): ret = "".join([struct.pack(">H32s", i, sharehashes[i]) for i in sorted(sharehashes.keys())]) hunk ./src/allmydata/test/test_storage.py 1912 return ret - def test_write(self): # This translates to a file with 6 6-byte segments, and with 2-byte # blocks. hunk ./src/allmydata/test/test_storage.py 2043 6, datalength) return mw - def test_write_rejected_with_too_many_blocks(self): mw = self._make_new_mw("si0", 0) hunk ./src/allmydata/test/test_storage.py 2059 mw.put_block, self.block, 7, self.salt)) return d - def test_write_rejected_with_invalid_salt(self): # Try writing an invalid salt. Salts are 16 bytes -- any more or # less should cause an error. hunk ./src/allmydata/test/test_storage.py 2070 None, mw.put_block, self.block, 7, bad_salt)) return d - def test_write_rejected_with_invalid_root_hash(self): # Try writing an invalid root hash. This should be SHA256d, and # 32 bytes long as a result. hunk ./src/allmydata/test/test_storage.py 2095 None, mw.put_root_hash, invalid_root_hash)) return d - def test_write_rejected_with_invalid_blocksize(self): # The blocksize implied by the writer that we get from # _make_new_mw is 2bytes -- any more or any less than this hunk ./src/allmydata/test/test_storage.py 2128 mw.put_block(valid_block, 5, self.salt)) return d - def test_write_enforces_order_constraints(self): # We require that the MDMFSlotWriteProxy be interacted with in a # specific way. hunk ./src/allmydata/test/test_storage.py 2213 mw0.put_verification_key(self.verification_key)) return d - def test_end_to_end(self): mw = self._make_new_mw("si1", 0) # Write a share using the mutable writer, and make sure that the hunk ./src/allmydata/test/test_storage.py 2378 self.failUnlessEqual(root_hash, self.root_hash, root_hash)) return d - def test_only_reads_one_segment_sdmf(self): # SDMF shares have only one segment, so it doesn't make sense to # read more segments than that. The reader should know this and hunk ./src/allmydata/test/test_storage.py 2395 mr.get_block_and_salt, 1)) return d - def test_read_with_prefetched_mdmf_data(self): # The MDMFSlotReadProxy will prefill certain fields if you pass # it data that you have already fetched. This is useful for hunk ./src/allmydata/test/test_storage.py 2459 d.addCallback(_check_block_and_salt) return d - def test_read_with_prefetched_sdmf_data(self): sdmf_data = self.build_test_sdmf_share() self.write_sdmf_share_to_server("si1") hunk ./src/allmydata/test/test_storage.py 2522 d.addCallback(_check_block_and_salt) return d - def test_read_with_empty_mdmf_file(self): # Some tests upload a file with no contents to test things # unrelated to the actual handling of the content of the file. hunk ./src/allmydata/test/test_storage.py 2550 mr.get_block_and_salt, 0)) return d - def test_read_with_empty_sdmf_file(self): self.write_sdmf_share_to_server("si1", empty=True) mr = MDMFSlotReadProxy(self.rref, "si1", 0) hunk ./src/allmydata/test/test_storage.py 2575 mr.get_block_and_salt, 0)) return d - def test_verinfo_with_sdmf_file(self): self.write_sdmf_share_to_server("si1") mr = MDMFSlotReadProxy(self.rref, "si1", 0) hunk ./src/allmydata/test/test_storage.py 2615 d.addCallback(_check_verinfo) return d - def test_verinfo_with_mdmf_file(self): self.write_test_share_to_server("si1") mr = MDMFSlotReadProxy(self.rref, "si1", 0) hunk ./src/allmydata/test/test_storage.py 2653 d.addCallback(_check_verinfo) return d - def test_sdmf_writer(self): # Go through the motions of writing an SDMF share to the storage # server. Then read the storage server to see that the share got hunk ./src/allmydata/test/test_storage.py 2696 d.addCallback(_then) return d - def test_sdmf_writer_preexisting_share(self): data = self.build_test_sdmf_share() self.write_sdmf_share_to_server("si1") hunk ./src/allmydata/test/test_storage.py 2839 self.failUnless(output["get"]["99_0_percentile"] is None, output) self.failUnless(output["get"]["99_9_percentile"] is None, output) + def remove_tags(s): s = re.sub(r'<[^>]*>', ' ', s) s = re.sub(r'\s+', ' ', s) hunk ./src/allmydata/test/test_storage.py 2845 return s + class MyBucketCountingCrawler(BucketCountingCrawler): def finished_prefix(self, cycle, prefix): BucketCountingCrawler.finished_prefix(self, cycle, prefix) hunk ./src/allmydata/test/test_storage.py 2974 backend = DiskBackend(fp) ss = MyStorageServer("\x00" * 20, backend, fp) ss.bucket_counter.slow_start = 0 + # these will be fired inside finished_prefix() hooks = ss.bucket_counter.hook_ds = [defer.Deferred() for i in range(3)] w = StorageStatus(ss) hunk ./src/allmydata/test/test_storage.py 3008 ss.setServiceParent(self.s) return d + class InstrumentedLeaseCheckingCrawler(LeaseCheckingCrawler): stop_after_first_bucket = False hunk ./src/allmydata/test/test_storage.py 3017 if self.stop_after_first_bucket: self.stop_after_first_bucket = False self.cpu_slice = -1.0 + def yielding(self, sleep_time): if not self.stop_after_first_bucket: self.cpu_slice = 500 hunk ./src/allmydata/test/test_storage.py 3028 class BrokenStatResults: pass + class No_ST_BLOCKS_LeaseCheckingCrawler(LeaseCheckingCrawler): def stat(self, fn): s = os.stat(fn) hunk ./src/allmydata/test/test_storage.py 3044 class No_ST_BLOCKS_StorageServer(StorageServer): LeaseCheckerClass = No_ST_BLOCKS_LeaseCheckingCrawler + class LeaseCrawler(unittest.TestCase, pollmixin.PollMixin, WebRenderingMixin): def setUp(self): hunk ./src/allmydata/test/test_storage.py 3891 backend = DiskBackend(fp) ss = InstrumentedStorageServer("\x00" * 20, backend, fp) w = StorageStatus(ss) + # make it start sooner than usual. lc = ss.lease_checker lc.stop_after_first_bucket = True hunk ./src/allmydata/util/fileutil.py 460 'avail': avail, } + def get_available_space(whichdirfp, reserved_space): """Returns available space for share storage in bytes, or None if no API to get this information is available. } [mutable/publish.py: elements should not be removed from a dictionary while it is being iterated over. refs #393 david-sarah@jacaranda.org**20110923040825 Ignore-this: 135da94bd344db6ccd59a576b54901c1 ] { hunk ./src/allmydata/mutable/publish.py 6 import os, time from StringIO import StringIO from itertools import count +from copy import copy from zope.interface import implements from twisted.internet import defer from twisted.python import failure merger 0.0 ( hunk ./src/allmydata/mutable/publish.py 868 - - # TODO: Bad, since we remove from this same dict. We need to - # make a copy, or just use a non-iterated value. - for (shnum, writer) in self.writers.iteritems(): + for (shnum, writer) in self.writers.copy().iteritems(): hunk ./src/allmydata/mutable/publish.py 868 - - # TODO: Bad, since we remove from this same dict. We need to - # make a copy, or just use a non-iterated value. - for (shnum, writer) in self.writers.iteritems(): + for (shnum, writer) in copy(self.writers).iteritems(): ) } [A few comment cleanups. refs #999 david-sarah@jacaranda.org**20110923041003 Ignore-this: f574b4a3954b6946016646011ad15edf ] { hunk ./src/allmydata/storage/backends/disk/disk_backend.py 17 # storage/ # storage/shares/incoming -# incoming/ holds temp dirs named $START/$STORAGEINDEX/$SHARENUM which will -# be moved to storage/shares/$START/$STORAGEINDEX/$SHARENUM upon success -# storage/shares/$START/$STORAGEINDEX -# storage/shares/$START/$STORAGEINDEX/$SHARENUM +# incoming/ holds temp dirs named $PREFIX/$STORAGEINDEX/$SHNUM which will +# be moved to storage/shares/$PREFIX/$STORAGEINDEX/$SHNUM upon success +# storage/shares/$PREFIX/$STORAGEINDEX +# storage/shares/$PREFIX/$STORAGEINDEX/$SHNUM hunk ./src/allmydata/storage/backends/disk/disk_backend.py 22 -# Where "$START" denotes the first 10 bits worth of $STORAGEINDEX (that's 2 +# Where "$PREFIX" denotes the first 10 bits worth of $STORAGEINDEX (that's 2 # base-32 chars). # $SHARENUM matches this regex: NUM_RE=re.compile("^[0-9]+$") hunk ./src/allmydata/storage/backends/disk/immutable.py 16 from allmydata.storage.lease import LeaseInfo -# each share file (in storage/shares/$SI/$SHNUM) contains lease information -# and share data. The share data is accessed by RIBucketWriter.write and -# RIBucketReader.read . The lease information is not accessible through these -# interfaces. +# Each share file (in storage/shares/$PREFIX/$STORAGEINDEX/$SHNUM) contains +# lease information and share data. The share data is accessed by +# RIBucketWriter.write and RIBucketReader.read . The lease information is not +# accessible through these remote interfaces. # The share file has the following layout: # 0x00: share file version number, four bytes, current version is 1 hunk ./src/allmydata/storage/backends/disk/immutable.py 211 # These lease operations are intended for use by disk_backend.py. # Other clients should not depend on the fact that the disk backend - # stores leases in share files. XXX bucket.py also relies on this. + # stores leases in share files. + # XXX BucketWriter in bucket.py also relies on add_lease. def get_leases(self): """Yields a LeaseInfo instance for all leases.""" } [Move advise_corrupt_share to allmydata/storage/backends/base.py, since it will be common to the disk and S3 backends. refs #999 david-sarah@jacaranda.org**20110923041115 Ignore-this: 782b49f243bd98fcb6c249f8e40fd9f ] { hunk ./src/allmydata/storage/backends/base.py 4 from twisted.application import service +from allmydata.util import fileutil, log, time_format from allmydata.storage.common import si_b2a from allmydata.storage.lease import LeaseInfo from allmydata.storage.bucket import BucketReader hunk ./src/allmydata/storage/backends/base.py 13 class Backend(service.MultiService): def __init__(self): service.MultiService.__init__(self) + self._corruption_advisory_dir = None + + def advise_corrupt_share(self, sharetype, storageindex, shnum, reason): + if self._corruption_advisory_dir is not None: + fileutil.fp_make_dirs(self._corruption_advisory_dir) + now = time_format.iso_utc(sep="T") + si_s = si_b2a(storageindex) + + # Windows can't handle colons in the filename. + name = ("%s--%s-%d" % (now, si_s, shnum)).replace(":", "") + f = self._corruption_advisory_dir.child(name).open("w") + try: + f.write("report: Share Corruption\n") + f.write("type: %s\n" % sharetype) + f.write("storage_index: %s\n" % si_s) + f.write("share_number: %d\n" % shnum) + f.write("\n") + f.write(reason) + f.write("\n") + finally: + f.close() + + log.msg(format=("client claims corruption in (%(share_type)s) " + + "%(si)s-%(shnum)d: %(reason)s"), + share_type=sharetype, si=si_s, shnum=shnum, reason=reason, + level=log.SCARY, umid="2fASGx") class ShareSet(object): hunk ./src/allmydata/storage/backends/disk/disk_backend.py 8 from zope.interface import implements from allmydata.interfaces import IStorageBackend, IShareSet -from allmydata.util import fileutil, log, time_format +from allmydata.util import fileutil, log from allmydata.storage.common import si_b2a, si_a2b from allmydata.storage.bucket import BucketWriter from allmydata.storage.backends.base import Backend, ShareSet hunk ./src/allmydata/storage/backends/disk/disk_backend.py 125 return 0 return fileutil.get_available_space(self._sharedir, self._reserved_space) - def advise_corrupt_share(self, sharetype, storageindex, shnum, reason): - fileutil.fp_make_dirs(self._corruption_advisory_dir) - now = time_format.iso_utc(sep="T") - si_s = si_b2a(storageindex) - - # Windows can't handle colons in the filename. - name = ("%s--%s-%d" % (now, si_s, shnum)).replace(":", "") - f = self._corruption_advisory_dir.child(name).open("w") - try: - f.write("report: Share Corruption\n") - f.write("type: %s\n" % sharetype) - f.write("storage_index: %s\n" % si_s) - f.write("share_number: %d\n" % shnum) - f.write("\n") - f.write(reason) - f.write("\n") - finally: - f.close() - - log.msg(format=("client claims corruption in (%(share_type)s) " + - "%(si)s-%(shnum)d: %(reason)s"), - share_type=sharetype, si=si_s, shnum=shnum, reason=reason, - level=log.SCARY, umid="SGx2fA") - class DiskShareSet(ShareSet): implements(IShareSet) } [Add incomplete S3 backend. refs #999 david-sarah@jacaranda.org**20110923041314 Ignore-this: b48df65699e3926dcbb87b5f755cdbf1 ] { adddir ./src/allmydata/storage/backends/s3 addfile ./src/allmydata/storage/backends/s3/__init__.py addfile ./src/allmydata/storage/backends/s3/immutable.py hunk ./src/allmydata/storage/backends/s3/immutable.py 1 + +import struct + +from zope.interface import implements + +from allmydata.interfaces import IStoredShare +from allmydata.util.assertutil import precondition +from allmydata.storage.common import si_b2a, UnknownImmutableContainerVersionError, DataTooLargeError + + +# Each share file (in storage/shares/$PREFIX/$STORAGEINDEX/$SHNUM) contains +# lease information [currently inaccessible] and share data. The share data is +# accessed by RIBucketWriter.write and RIBucketReader.read . + +# The share file has the following layout: +# 0x00: share file version number, four bytes, current version is 1 +# 0x04: always zero (was share data length prior to Tahoe-LAFS v1.3.0) +# 0x08: number of leases, four bytes big-endian +# 0x0c: beginning of share data (see immutable.layout.WriteBucketProxy) +# data_length+0x0c: first lease. Each lease record is 72 bytes. + + +class ImmutableS3Share(object): + implements(IStoredShare) + + sharetype = "immutable" + LEASE_SIZE = struct.calcsize(">L32s32sL") # for compatibility + + + def __init__(self, storageindex, shnum, s3bucket, create=False, max_size=None): + """ + If max_size is not None then I won't allow more than max_size to be written to me. + """ + precondition((max_size is not None) or not create, max_size, create) + self._storageindex = storageindex + self._max_size = max_size + + self._s3bucket = s3bucket + si_s = si_b2a(storageindex) + self._key = "storage/shares/%s/%s/%d" % (si_s[:2], si_s, shnum) + self._shnum = shnum + + if create: + # The second field, which was the four-byte share data length in + # Tahoe-LAFS versions prior to 1.3.0, is not used; we always write 0. + # We also write 0 for the number of leases. + self._home.setContent(struct.pack(">LLL", 1, 0, 0) ) + self._end_offset = max_size + 0x0c + + # TODO: start write to S3. + else: + # TODO: get header + header = "\x00"*12 + (version, unused, num_leases) = struct.unpack(">LLL", header) + + if version != 1: + msg = "sharefile %s had version %d but we wanted 1" % \ + (self._home, version) + raise UnknownImmutableContainerVersionError(msg) + + # We cannot write leases in share files, but allow them to be present + # in case a share file is copied from a disk backend, or in case we + # need them in future. + # TODO: filesize = size of S3 object + self._end_offset = filesize - (num_leases * self.LEASE_SIZE) + self._data_offset = 0xc + + def __repr__(self): + return ("" + % (si_b2a(self._storageindex), self._shnum, self._key)) + + def close(self): + # TODO: finalize write to S3. + pass + + def get_used_space(self): + return self._size + + def get_storage_index(self): + return self._storageindex + + def get_storage_index_string(self): + return si_b2a(self._storageindex) + + def get_shnum(self): + return self._shnum + + def unlink(self): + # TODO: remove the S3 object. + pass + + def get_allocated_size(self): + return self._max_size + + def get_size(self): + return self._size + + def get_data_length(self): + return self._end_offset - self._data_offset + + def read_share_data(self, offset, length): + precondition(offset >= 0) + + # Reads beyond the end of the data are truncated. Reads that start + # beyond the end of the data return an empty string. + seekpos = self._data_offset+offset + actuallength = max(0, min(length, self._end_offset-seekpos)) + if actuallength == 0: + return "" + + # TODO: perform an S3 GET request, possibly with a Content-Range header. + return "\x00"*actuallength + + def write_share_data(self, offset, data): + assert offset >= self._size, "offset = %r, size = %r" % (offset, self._size) + + # TODO: write data to S3. If offset > self._size, fill the space + # between with zeroes. + + self._size = offset + len(data) + + def add_lease(self, lease_info): + pass addfile ./src/allmydata/storage/backends/s3/mutable.py hunk ./src/allmydata/storage/backends/s3/mutable.py 1 + +import struct + +from zope.interface import implements + +from allmydata.interfaces import IStoredMutableShare, BadWriteEnablerError +from allmydata.util import fileutil, idlib, log +from allmydata.util.assertutil import precondition +from allmydata.util.hashutil import constant_time_compare +from allmydata.util.encodingutil import quote_filepath +from allmydata.storage.common import si_b2a, UnknownMutableContainerVersionError, \ + DataTooLargeError +from allmydata.storage.lease import LeaseInfo +from allmydata.storage.backends.base import testv_compare + + +# The MutableDiskShare is like the ImmutableDiskShare, but used for mutable data. +# It has a different layout. See docs/mutable.rst for more details. + +# # offset size name +# 1 0 32 magic verstr "tahoe mutable container v1" plus binary +# 2 32 20 write enabler's nodeid +# 3 52 32 write enabler +# 4 84 8 data size (actual share data present) (a) +# 5 92 8 offset of (8) count of extra leases (after data) +# 6 100 368 four leases, 92 bytes each +# 0 4 ownerid (0 means "no lease here") +# 4 4 expiration timestamp +# 8 32 renewal token +# 40 32 cancel token +# 72 20 nodeid that accepted the tokens +# 7 468 (a) data +# 8 ?? 4 count of extra leases +# 9 ?? n*92 extra leases + + +# The struct module doc says that L's are 4 bytes in size, and that Q's are +# 8 bytes in size. Since compatibility depends upon this, double-check it. +assert struct.calcsize(">L") == 4, struct.calcsize(">L") +assert struct.calcsize(">Q") == 8, struct.calcsize(">Q") + + +class MutableDiskShare(object): + implements(IStoredMutableShare) + + sharetype = "mutable" + DATA_LENGTH_OFFSET = struct.calcsize(">32s20s32s") + EXTRA_LEASE_OFFSET = DATA_LENGTH_OFFSET + 8 + HEADER_SIZE = struct.calcsize(">32s20s32sQQ") # doesn't include leases + LEASE_SIZE = struct.calcsize(">LL32s32s20s") + assert LEASE_SIZE == 92 + DATA_OFFSET = HEADER_SIZE + 4*LEASE_SIZE + assert DATA_OFFSET == 468, DATA_OFFSET + + # our sharefiles share with a recognizable string, plus some random + # binary data to reduce the chance that a regular text file will look + # like a sharefile. + MAGIC = "Tahoe mutable container v1\n" + "\x75\x09\x44\x03\x8e" + assert len(MAGIC) == 32 + MAX_SIZE = 2*1000*1000*1000 # 2GB, kind of arbitrary + # TODO: decide upon a policy for max share size + + def __init__(self, storageindex, shnum, home, parent=None): + self._storageindex = storageindex + self._shnum = shnum + self._home = home + if self._home.exists(): + # we don't cache anything, just check the magic + f = self._home.open('rb') + try: + data = f.read(self.HEADER_SIZE) + (magic, + write_enabler_nodeid, write_enabler, + data_length, extra_least_offset) = \ + struct.unpack(">32s20s32sQQ", data) + if magic != self.MAGIC: + msg = "sharefile %s had magic '%r' but we wanted '%r'" % \ + (quote_filepath(self._home), magic, self.MAGIC) + raise UnknownMutableContainerVersionError(msg) + finally: + f.close() + self.parent = parent # for logging + + def log(self, *args, **kwargs): + if self.parent: + return self.parent.log(*args, **kwargs) + + def create(self, serverid, write_enabler): + assert not self._home.exists() + data_length = 0 + extra_lease_offset = (self.HEADER_SIZE + + 4 * self.LEASE_SIZE + + data_length) + assert extra_lease_offset == self.DATA_OFFSET # true at creation + num_extra_leases = 0 + f = self._home.open('wb') + try: + header = struct.pack(">32s20s32sQQ", + self.MAGIC, serverid, write_enabler, + data_length, extra_lease_offset, + ) + leases = ("\x00"*self.LEASE_SIZE) * 4 + f.write(header + leases) + # data goes here, empty after creation + f.write(struct.pack(">L", num_extra_leases)) + # extra leases go here, none at creation + finally: + f.close() + + def __repr__(self): + return ("" + % (si_b2a(self._storageindex), self._shnum, quote_filepath(self._home))) + + def get_used_space(self): + return fileutil.get_used_space(self._home) + + def get_storage_index(self): + return self._storageindex + + def get_storage_index_string(self): + return si_b2a(self._storageindex) + + def get_shnum(self): + return self._shnum + + def unlink(self): + self._home.remove() + + def _read_data_length(self, f): + f.seek(self.DATA_LENGTH_OFFSET) + (data_length,) = struct.unpack(">Q", f.read(8)) + return data_length + + def _write_data_length(self, f, data_length): + f.seek(self.DATA_LENGTH_OFFSET) + f.write(struct.pack(">Q", data_length)) + + def _read_share_data(self, f, offset, length): + precondition(offset >= 0) + data_length = self._read_data_length(f) + if offset+length > data_length: + # reads beyond the end of the data are truncated. Reads that + # start beyond the end of the data return an empty string. + length = max(0, data_length-offset) + if length == 0: + return "" + precondition(offset+length <= data_length) + f.seek(self.DATA_OFFSET+offset) + data = f.read(length) + return data + + def _read_extra_lease_offset(self, f): + f.seek(self.EXTRA_LEASE_OFFSET) + (extra_lease_offset,) = struct.unpack(">Q", f.read(8)) + return extra_lease_offset + + def _write_extra_lease_offset(self, f, offset): + f.seek(self.EXTRA_LEASE_OFFSET) + f.write(struct.pack(">Q", offset)) + + def _read_num_extra_leases(self, f): + offset = self._read_extra_lease_offset(f) + f.seek(offset) + (num_extra_leases,) = struct.unpack(">L", f.read(4)) + return num_extra_leases + + def _write_num_extra_leases(self, f, num_leases): + extra_lease_offset = self._read_extra_lease_offset(f) + f.seek(extra_lease_offset) + f.write(struct.pack(">L", num_leases)) + + def _change_container_size(self, f, new_container_size): + if new_container_size > self.MAX_SIZE: + raise DataTooLargeError() + old_extra_lease_offset = self._read_extra_lease_offset(f) + new_extra_lease_offset = self.DATA_OFFSET + new_container_size + if new_extra_lease_offset < old_extra_lease_offset: + # TODO: allow containers to shrink. For now they remain large. + return + num_extra_leases = self._read_num_extra_leases(f) + f.seek(old_extra_lease_offset) + leases_size = 4 + num_extra_leases * self.LEASE_SIZE + extra_lease_data = f.read(leases_size) + + # Zero out the old lease info (in order to minimize the chance that + # it could accidentally be exposed to a reader later, re #1528). + f.seek(old_extra_lease_offset) + f.write('\x00' * leases_size) + f.flush() + + # An interrupt here will corrupt the leases. + + f.seek(new_extra_lease_offset) + f.write(extra_lease_data) + self._write_extra_lease_offset(f, new_extra_lease_offset) + + def _write_share_data(self, f, offset, data): + length = len(data) + precondition(offset >= 0) + data_length = self._read_data_length(f) + extra_lease_offset = self._read_extra_lease_offset(f) + + if offset+length >= data_length: + # They are expanding their data size. + + if self.DATA_OFFSET+offset+length > extra_lease_offset: + # TODO: allow containers to shrink. For now, they remain + # large. + + # Their new data won't fit in the current container, so we + # have to move the leases. With luck, they're expanding it + # more than the size of the extra lease block, which will + # minimize the corrupt-the-share window + self._change_container_size(f, offset+length) + extra_lease_offset = self._read_extra_lease_offset(f) + + # an interrupt here is ok.. the container has been enlarged + # but the data remains untouched + + assert self.DATA_OFFSET+offset+length <= extra_lease_offset + # Their data now fits in the current container. We must write + # their new data and modify the recorded data size. + + # Fill any newly exposed empty space with 0's. + if offset > data_length: + f.seek(self.DATA_OFFSET+data_length) + f.write('\x00'*(offset - data_length)) + f.flush() + + new_data_length = offset+length + self._write_data_length(f, new_data_length) + # an interrupt here will result in a corrupted share + + # now all that's left to do is write out their data + f.seek(self.DATA_OFFSET+offset) + f.write(data) + return + + def _write_lease_record(self, f, lease_number, lease_info): + extra_lease_offset = self._read_extra_lease_offset(f) + num_extra_leases = self._read_num_extra_leases(f) + if lease_number < 4: + offset = self.HEADER_SIZE + lease_number * self.LEASE_SIZE + elif (lease_number-4) < num_extra_leases: + offset = (extra_lease_offset + + 4 + + (lease_number-4)*self.LEASE_SIZE) + else: + # must add an extra lease record + self._write_num_extra_leases(f, num_extra_leases+1) + offset = (extra_lease_offset + + 4 + + (lease_number-4)*self.LEASE_SIZE) + f.seek(offset) + assert f.tell() == offset + f.write(lease_info.to_mutable_data()) + + def _read_lease_record(self, f, lease_number): + # returns a LeaseInfo instance, or None + extra_lease_offset = self._read_extra_lease_offset(f) + num_extra_leases = self._read_num_extra_leases(f) + if lease_number < 4: + offset = self.HEADER_SIZE + lease_number * self.LEASE_SIZE + elif (lease_number-4) < num_extra_leases: + offset = (extra_lease_offset + + 4 + + (lease_number-4)*self.LEASE_SIZE) + else: + raise IndexError("No such lease number %d" % lease_number) + f.seek(offset) + assert f.tell() == offset + data = f.read(self.LEASE_SIZE) + lease_info = LeaseInfo().from_mutable_data(data) + if lease_info.owner_num == 0: + return None + return lease_info + + def _get_num_lease_slots(self, f): + # how many places do we have allocated for leases? Not all of them + # are filled. + num_extra_leases = self._read_num_extra_leases(f) + return 4+num_extra_leases + + def _get_first_empty_lease_slot(self, f): + # return an int with the index of an empty slot, or None if we do not + # currently have an empty slot + + for i in range(self._get_num_lease_slots(f)): + if self._read_lease_record(f, i) is None: + return i + return None + + def get_leases(self): + """Yields a LeaseInfo instance for all leases.""" + f = self._home.open('rb') + try: + for i, lease in self._enumerate_leases(f): + yield lease + finally: + f.close() + + def _enumerate_leases(self, f): + for i in range(self._get_num_lease_slots(f)): + try: + data = self._read_lease_record(f, i) + if data is not None: + yield i, data + except IndexError: + return + + # These lease operations are intended for use by disk_backend.py. + # Other non-test clients should not depend on the fact that the disk + # backend stores leases in share files. + + def add_lease(self, lease_info): + precondition(lease_info.owner_num != 0) # 0 means "no lease here" + f = self._home.open('rb+') + try: + num_lease_slots = self._get_num_lease_slots(f) + empty_slot = self._get_first_empty_lease_slot(f) + if empty_slot is not None: + self._write_lease_record(f, empty_slot, lease_info) + else: + self._write_lease_record(f, num_lease_slots, lease_info) + finally: + f.close() + + def renew_lease(self, renew_secret, new_expire_time): + accepting_nodeids = set() + f = self._home.open('rb+') + try: + for (leasenum, lease) in self._enumerate_leases(f): + if constant_time_compare(lease.renew_secret, renew_secret): + # yup. See if we need to update the owner time. + if new_expire_time > lease.expiration_time: + # yes + lease.expiration_time = new_expire_time + self._write_lease_record(f, leasenum, lease) + return + accepting_nodeids.add(lease.nodeid) + finally: + f.close() + # Return the accepting_nodeids set, to give the client a chance to + # update the leases on a share that has been migrated from its + # original server to a new one. + msg = ("Unable to renew non-existent lease. I have leases accepted by" + " nodeids: ") + msg += ",".join([("'%s'" % idlib.nodeid_b2a(anid)) + for anid in accepting_nodeids]) + msg += " ." + raise IndexError(msg) + + def add_or_renew_lease(self, lease_info): + precondition(lease_info.owner_num != 0) # 0 means "no lease here" + try: + self.renew_lease(lease_info.renew_secret, + lease_info.expiration_time) + except IndexError: + self.add_lease(lease_info) + + def cancel_lease(self, cancel_secret): + """Remove any leases with the given cancel_secret. If the last lease + is cancelled, the file will be removed. Return the number of bytes + that were freed (by truncating the list of leases, and possibly by + deleting the file). Raise IndexError if there was no lease with the + given cancel_secret.""" + + # XXX can this be more like ImmutableDiskShare.cancel_lease? + + accepting_nodeids = set() + modified = 0 + remaining = 0 + blank_lease = LeaseInfo(owner_num=0, + renew_secret="\x00"*32, + cancel_secret="\x00"*32, + expiration_time=0, + nodeid="\x00"*20) + f = self._home.open('rb+') + try: + for (leasenum, lease) in self._enumerate_leases(f): + accepting_nodeids.add(lease.nodeid) + if constant_time_compare(lease.cancel_secret, cancel_secret): + self._write_lease_record(f, leasenum, blank_lease) + modified += 1 + else: + remaining += 1 + if modified: + freed_space = self._pack_leases(f) + finally: + f.close() + + if modified > 0: + if remaining == 0: + freed_space = fileutil.get_used_space(self._home) + self.unlink() + return freed_space + + msg = ("Unable to cancel non-existent lease. I have leases " + "accepted by nodeids: ") + msg += ",".join([("'%s'" % idlib.nodeid_b2a(anid)) + for anid in accepting_nodeids]) + msg += " ." + raise IndexError(msg) + + def _pack_leases(self, f): + # TODO: reclaim space from cancelled leases + return 0 + + def _read_write_enabler_and_nodeid(self, f): + f.seek(0) + data = f.read(self.HEADER_SIZE) + (magic, + write_enabler_nodeid, write_enabler, + data_length, extra_least_offset) = \ + struct.unpack(">32s20s32sQQ", data) + assert magic == self.MAGIC + return (write_enabler, write_enabler_nodeid) + + def readv(self, readv): + datav = [] + f = self._home.open('rb') + try: + for (offset, length) in readv: + datav.append(self._read_share_data(f, offset, length)) + finally: + f.close() + return datav + + def get_size(self): + return self._home.getsize() + + def get_data_length(self): + f = self._home.open('rb') + try: + data_length = self._read_data_length(f) + finally: + f.close() + return data_length + + def check_write_enabler(self, write_enabler, si_s): + f = self._home.open('rb+') + try: + (real_write_enabler, write_enabler_nodeid) = self._read_write_enabler_and_nodeid(f) + finally: + f.close() + # avoid a timing attack + #if write_enabler != real_write_enabler: + if not constant_time_compare(write_enabler, real_write_enabler): + # accomodate share migration by reporting the nodeid used for the + # old write enabler. + self.log(format="bad write enabler on SI %(si)s," + " recorded by nodeid %(nodeid)s", + facility="tahoe.storage", + level=log.WEIRD, umid="cE1eBQ", + si=si_s, nodeid=idlib.nodeid_b2a(write_enabler_nodeid)) + msg = "The write enabler was recorded by nodeid '%s'." % \ + (idlib.nodeid_b2a(write_enabler_nodeid),) + raise BadWriteEnablerError(msg) + + def check_testv(self, testv): + test_good = True + f = self._home.open('rb+') + try: + for (offset, length, operator, specimen) in testv: + data = self._read_share_data(f, offset, length) + if not testv_compare(data, operator, specimen): + test_good = False + break + finally: + f.close() + return test_good + + def writev(self, datav, new_length): + f = self._home.open('rb+') + try: + for (offset, data) in datav: + self._write_share_data(f, offset, data) + if new_length is not None: + cur_length = self._read_data_length(f) + if new_length < cur_length: + self._write_data_length(f, new_length) + # TODO: if we're going to shrink the share file when the + # share data has shrunk, then call + # self._change_container_size() here. + finally: + f.close() + + def close(self): + pass + + +def create_mutable_disk_share(storageindex, shnum, fp, serverid, write_enabler, parent): + ms = MutableDiskShare(storageindex, shnum, fp, parent) + ms.create(serverid, write_enabler) + del ms + return MutableDiskShare(storageindex, shnum, fp, parent) addfile ./src/allmydata/storage/backends/s3/s3_backend.py hunk ./src/allmydata/storage/backends/s3/s3_backend.py 1 + +from zope.interface import implements +from allmydata.interfaces import IStorageBackend, IShareSet +from allmydata.storage.common import si_b2a, si_a2b +from allmydata.storage.bucket import BucketWriter +from allmydata.storage.backends.base import Backend, ShareSet +from allmydata.storage.backends.s3.immutable import ImmutableS3Share +from allmydata.storage.backends.s3.mutable import MutableS3Share + +# The S3 bucket has keys of the form shares/$STORAGEINDEX/$SHARENUM + + +class S3Backend(Backend): + implements(IStorageBackend) + + def __init__(self, s3bucket, readonly=False, max_space=None, corruption_advisory_dir=None): + Backend.__init__(self) + self._s3bucket = s3bucket + self._readonly = readonly + if max_space is None: + self._max_space = 2**64 + else: + self._max_space = int(max_space) + + # TODO: any set-up for S3? + + # we don't actually create the corruption-advisory dir until necessary + self._corruption_advisory_dir = corruption_advisory_dir + + def get_sharesets_for_prefix(self, prefix): + # TODO: query S3 for keys matching prefix + return [] + + def get_shareset(self, storageindex): + return S3ShareSet(storageindex, self._s3bucket) + + def fill_in_space_stats(self, stats): + stats['storage_server.max_space'] = self._max_space + + # TODO: query space usage of S3 bucket + stats['storage_server.accepting_immutable_shares'] = int(not self._readonly) + + def get_available_space(self): + if self._readonly: + return 0 + # TODO: query space usage of S3 bucket + return self._max_space + + +class S3ShareSet(ShareSet): + implements(IShareSet) + + def __init__(self, storageindex, s3bucket): + ShareSet.__init__(self, storageindex) + self._s3bucket = s3bucket + + def get_overhead(self): + return 0 + + def get_shares(self): + """ + Generate IStorageBackendShare objects for shares we have for this storage index. + ("Shares we have" means completed ones, excluding incoming ones.) + """ + pass + + def has_incoming(self, shnum): + # TODO: this might need to be more like the disk backend; review callers + return False + + def make_bucket_writer(self, storageserver, shnum, max_space_per_bucket, lease_info, canary): + immsh = ImmutableS3Share(self.get_storage_index(), shnum, self._s3bucket, + max_size=max_space_per_bucket) + bw = BucketWriter(storageserver, immsh, lease_info, canary) + return bw + + def _create_mutable_share(self, storageserver, shnum, write_enabler): + # TODO + serverid = storageserver.get_serverid() + return MutableS3Share(self.get_storage_index(), shnum, self._s3bucket, serverid, write_enabler, storageserver) + + def _clean_up_after_unlink(self): + pass + } [interfaces.py: add fill_in_space_stats method to IStorageBackend. refs #999 david-sarah@jacaranda.org**20110923203723 Ignore-this: 59371c150532055939794fed6c77dcb6 ] { hunk ./src/allmydata/interfaces.py 304 def get_sharesets_for_prefix(prefix): """ Generates IShareSet objects for all storage indices matching the - given prefix for which this backend holds shares. + given base-32 prefix for which this backend holds shares. """ def get_shareset(storageindex): hunk ./src/allmydata/interfaces.py 312 Get an IShareSet object for the given storage index. """ + def fill_in_space_stats(stats): + """ + Fill in the 'stats' dict with space statistics for this backend, in + 'storage_server.*' keys. + """ + def advise_corrupt_share(storageindex, sharetype, shnum, reason): """ Clients who discover hash failures in shares that they have } [Remove redundant si_s argument from check_write_enabler. refs #999 david-sarah@jacaranda.org**20110923204425 Ignore-this: 25be760118dbce2eb661137f7d46dd20 ] { hunk ./src/allmydata/interfaces.py 500 class IStoredMutableShare(IStoredShare): - def check_write_enabler(write_enabler, si_s): + def check_write_enabler(write_enabler): """ XXX """ hunk ./src/allmydata/storage/backends/base.py 102 if len(secrets) > 2: cancel_secret = secrets[2] - si_s = self.get_storage_index_string() shares = {} for share in self.get_shares(): # XXX is it correct to ignore immutable shares? Maybe get_shares should hunk ./src/allmydata/storage/backends/base.py 107 # have a parameter saying what type it's expecting. if share.sharetype == "mutable": - share.check_write_enabler(write_enabler, si_s) + share.check_write_enabler(write_enabler) shares[share.get_shnum()] = share # write_enabler is good for all existing shares hunk ./src/allmydata/storage/backends/disk/mutable.py 440 f.close() return data_length - def check_write_enabler(self, write_enabler, si_s): + def check_write_enabler(self, write_enabler): f = self._home.open('rb+') try: (real_write_enabler, write_enabler_nodeid) = self._read_write_enabler_and_nodeid(f) hunk ./src/allmydata/storage/backends/disk/mutable.py 447 finally: f.close() # avoid a timing attack - #if write_enabler != real_write_enabler: if not constant_time_compare(write_enabler, real_write_enabler): # accomodate share migration by reporting the nodeid used for the # old write enabler. hunk ./src/allmydata/storage/backends/disk/mutable.py 454 " recorded by nodeid %(nodeid)s", facility="tahoe.storage", level=log.WEIRD, umid="cE1eBQ", - si=si_s, nodeid=idlib.nodeid_b2a(write_enabler_nodeid)) + si=self.get_storage_index_string(), + nodeid=idlib.nodeid_b2a(write_enabler_nodeid)) msg = "The write enabler was recorded by nodeid '%s'." % \ (idlib.nodeid_b2a(write_enabler_nodeid),) raise BadWriteEnablerError(msg) hunk ./src/allmydata/storage/backends/s3/mutable.py 440 f.close() return data_length - def check_write_enabler(self, write_enabler, si_s): + def check_write_enabler(self, write_enabler): f = self._home.open('rb+') try: (real_write_enabler, write_enabler_nodeid) = self._read_write_enabler_and_nodeid(f) hunk ./src/allmydata/storage/backends/s3/mutable.py 447 finally: f.close() # avoid a timing attack - #if write_enabler != real_write_enabler: if not constant_time_compare(write_enabler, real_write_enabler): # accomodate share migration by reporting the nodeid used for the # old write enabler. hunk ./src/allmydata/storage/backends/s3/mutable.py 454 " recorded by nodeid %(nodeid)s", facility="tahoe.storage", level=log.WEIRD, umid="cE1eBQ", - si=si_s, nodeid=idlib.nodeid_b2a(write_enabler_nodeid)) + si=self.get_storage_index_string(), + nodeid=idlib.nodeid_b2a(write_enabler_nodeid)) msg = "The write enabler was recorded by nodeid '%s'." % \ (idlib.nodeid_b2a(write_enabler_nodeid),) raise BadWriteEnablerError(msg) } [Implement readv for immutable shares. refs #999 david-sarah@jacaranda.org**20110923204611 Ignore-this: 24f14b663051169d66293020e40c5a05 ] { hunk ./src/allmydata/storage/backends/disk/immutable.py 156 def get_data_length(self): return self._lease_offset - self._data_offset - #def readv(self, read_vector): - # ... + def readv(self, readv): + datav = [] + f = self._home.open('rb') + try: + for (offset, length) in readv: + datav.append(self._read_share_data(f, offset, length)) + finally: + f.close() + return datav hunk ./src/allmydata/storage/backends/disk/immutable.py 166 - def read_share_data(self, offset, length): + def _read_share_data(self, f, offset, length): precondition(offset >= 0) # Reads beyond the end of the data are truncated. Reads that start hunk ./src/allmydata/storage/backends/disk/immutable.py 175 actuallength = max(0, min(length, self._lease_offset-seekpos)) if actuallength == 0: return "" + f.seek(seekpos) + return f.read(actuallength) + + def read_share_data(self, offset, length): f = self._home.open(mode='rb') try: hunk ./src/allmydata/storage/backends/disk/immutable.py 181 - f.seek(seekpos) - sharedata = f.read(actuallength) + return self._read_share_data(f, offset, length) finally: f.close() hunk ./src/allmydata/storage/backends/disk/immutable.py 184 - return sharedata def write_share_data(self, offset, data): length = len(data) hunk ./src/allmydata/storage/backends/null/null_backend.py 89 return self.shnum def unlink(self): - os.unlink(self.fname) + pass + + def readv(self, readv): + datav = [] + for (offset, length) in readv: + datav.append("") + return datav def read_share_data(self, offset, length): precondition(offset >= 0) hunk ./src/allmydata/storage/backends/s3/immutable.py 101 def get_data_length(self): return self._end_offset - self._data_offset + def readv(self, readv): + datav = [] + for (offset, length) in readv: + datav.append(self.read_share_data(offset, length)) + return datav + def read_share_data(self, offset, length): precondition(offset >= 0) } [The cancel secret needs to be unique, even if it isn't explicitly provided. refs #999 david-sarah@jacaranda.org**20110923204914 Ignore-this: 6c44bb908dd4c0cdc59506b2d87a47b0 ] { hunk ./src/allmydata/storage/backends/base.py 98 write_enabler = secrets[0] renew_secret = secrets[1] - cancel_secret = '\x00'*32 if len(secrets) > 2: cancel_secret = secrets[2] hunk ./src/allmydata/storage/backends/base.py 100 + else: + cancel_secret = renew_secret shares = {} for share in self.get_shares(): } [Make EmptyShare.check_testv a simple function. refs #999 david-sarah@jacaranda.org**20110923204945 Ignore-this: d0132c085f40c39815fa920b77fc39ab ] { hunk ./src/allmydata/storage/backends/base.py 125 else: # compare the vectors against an empty share, in which all # reads return empty strings - if not EmptyShare().check_testv(testv): + if not empty_check_testv(testv): storageserver.log("testv failed (empty): [%d] %r" % (sharenum, testv)) testv_is_good = False break hunk ./src/allmydata/storage/backends/base.py 195 # never reached -class EmptyShare: - def check_testv(self, testv): - test_good = True - for (offset, length, operator, specimen) in testv: - data = "" - if not testv_compare(data, operator, specimen): - test_good = False - break - return test_good +def empty_check_testv(testv): + test_good = True + for (offset, length, operator, specimen) in testv: + data = "" + if not testv_compare(data, operator, specimen): + test_good = False + break + return test_good } [Update the null backend to take into account interface changes. Also, it now records which shares are present, but not their contents. refs #999 david-sarah@jacaranda.org**20110923205219 Ignore-this: 42a23d7e253255003dc63facea783251 ] { hunk ./src/allmydata/storage/backends/null/null_backend.py 2 -import os, struct - from zope.interface import implements from allmydata.interfaces import IStorageBackend, IShareSet, IStoredShare, IStoredMutableShare hunk ./src/allmydata/storage/backends/null/null_backend.py 6 from allmydata.util.assertutil import precondition -from allmydata.util.hashutil import constant_time_compare -from allmydata.storage.backends.base import Backend, ShareSet -from allmydata.storage.bucket import BucketWriter +from allmydata.storage.backends.base import Backend, empty_check_testv +from allmydata.storage.bucket import BucketWriter, BucketReader from allmydata.storage.common import si_b2a hunk ./src/allmydata/storage/backends/null/null_backend.py 9 -from allmydata.storage.lease import LeaseInfo class NullBackend(Backend): hunk ./src/allmydata/storage/backends/null/null_backend.py 13 implements(IStorageBackend) + """ + I am a test backend that records (in memory) which shares exist, but not their contents, leases, + or write-enablers. + """ def __init__(self): Backend.__init__(self) hunk ./src/allmydata/storage/backends/null/null_backend.py 20 + # mapping from storageindex to NullShareSet + self._sharesets = {} hunk ./src/allmydata/storage/backends/null/null_backend.py 23 - def get_available_space(self, reserved_space): + def get_available_space(self): return None def get_sharesets_for_prefix(self, prefix): hunk ./src/allmydata/storage/backends/null/null_backend.py 27 - pass + sharesets = [] + for (si, shareset) in self._sharesets.iteritems(): + if si_b2a(si).startswith(prefix): + sharesets.append(shareset) + + def _by_base32si(b): + return b.get_storage_index_string() + sharesets.sort(key=_by_base32si) + return sharesets def get_shareset(self, storageindex): hunk ./src/allmydata/storage/backends/null/null_backend.py 38 - return NullShareSet(storageindex) + shareset = self._sharesets.get(storageindex, None) + if shareset is None: + shareset = NullShareSet(storageindex) + self._sharesets[storageindex] = shareset + return shareset def fill_in_space_stats(self, stats): pass hunk ./src/allmydata/storage/backends/null/null_backend.py 47 - def set_storage_server(self, ss): - self.ss = ss hunk ./src/allmydata/storage/backends/null/null_backend.py 48 - def advise_corrupt_share(self, sharetype, storageindex, shnum, reason): - pass - - -class NullShareSet(ShareSet): +class NullShareSet(object): implements(IShareSet) def __init__(self, storageindex): hunk ./src/allmydata/storage/backends/null/null_backend.py 53 self.storageindex = storageindex + self._incoming_shnums = set() + self._immutable_shnums = set() + self._mutable_shnums = set() + + def close_shnum(self, shnum): + self._incoming_shnums.remove(shnum) + self._immutable_shnums.add(shnum) def get_overhead(self): return 0 hunk ./src/allmydata/storage/backends/null/null_backend.py 64 - def get_incoming_shnums(self): - return frozenset() - def get_shares(self): hunk ./src/allmydata/storage/backends/null/null_backend.py 65 + for shnum in self._immutable_shnums: + yield ImmutableNullShare(self, shnum) + for shnum in self._mutable_shnums: + yield MutableNullShare(self, shnum) + + def renew_lease(self, renew_secret, new_expiration_time): + raise IndexError("no such lease to renew") + + def get_leases(self): pass hunk ./src/allmydata/storage/backends/null/null_backend.py 76 - def get_share(self, shnum): - return None + def add_or_renew_lease(self, lease_info): + pass + + def has_incoming(self, shnum): + return shnum in self._incoming_shnums def get_storage_index(self): return self.storageindex hunk ./src/allmydata/storage/backends/null/null_backend.py 89 return si_b2a(self.storageindex) def make_bucket_writer(self, storageserver, shnum, max_space_per_bucket, lease_info, canary): - immutableshare = ImmutableNullShare() - return BucketWriter(self.ss, immutableshare, max_space_per_bucket, lease_info, canary) + self._incoming_shnums.add(shnum) + immutableshare = ImmutableNullShare(self, shnum) + bw = BucketWriter(storageserver, immutableshare, lease_info, canary) + bw.throw_out_all_data = True + return bw hunk ./src/allmydata/storage/backends/null/null_backend.py 95 - def _create_mutable_share(self, storageserver, shnum, write_enabler): - return MutableNullShare() + def make_bucket_reader(self, storageserver, share): + return BucketReader(storageserver, share) hunk ./src/allmydata/storage/backends/null/null_backend.py 98 - def _clean_up_after_unlink(self): - pass + def testv_and_readv_and_writev(self, storageserver, secrets, + test_and_write_vectors, read_vector, + expiration_time): + # evaluate test vectors + testv_is_good = True + for sharenum in test_and_write_vectors: + # compare the vectors against an empty share, in which all + # reads return empty strings + (testv, datav, new_length) = test_and_write_vectors[sharenum] + if not empty_check_testv(testv): + storageserver.log("testv failed (empty): [%d] %r" % (sharenum, testv)) + testv_is_good = False + break hunk ./src/allmydata/storage/backends/null/null_backend.py 112 + # gather the read vectors + read_data = {} + for shnum in self._mutable_shnums: + read_data[shnum] = "" hunk ./src/allmydata/storage/backends/null/null_backend.py 117 -class ImmutableNullShare: - implements(IStoredShare) - sharetype = "immutable" + if testv_is_good: + # now apply the write vectors + for shnum in test_and_write_vectors: + (testv, datav, new_length) = test_and_write_vectors[shnum] + if new_length == 0: + self._mutable_shnums.remove(shnum) + else: + self._mutable_shnums.add(shnum) hunk ./src/allmydata/storage/backends/null/null_backend.py 126 - def __init__(self): - """ If max_size is not None then I won't allow more than - max_size to be written to me. If create=True then max_size - must not be None. """ - pass + return (testv_is_good, read_data) + + def readv(self, wanted_shnums, read_vector): + return {} + + +class NullShareBase(object): + def __init__(self, shareset, shnum): + self.shareset = shareset + self.shnum = shnum + + def get_storage_index(self): + return self.shareset.get_storage_index() + + def get_storage_index_string(self): + return self.shareset.get_storage_index_string() def get_shnum(self): return self.shnum hunk ./src/allmydata/storage/backends/null/null_backend.py 146 + def get_data_length(self): + return 0 + + def get_size(self): + return 0 + + def get_used_space(self): + return 0 + def unlink(self): pass hunk ./src/allmydata/storage/backends/null/null_backend.py 166 def read_share_data(self, offset, length): precondition(offset >= 0) - # Reads beyond the end of the data are truncated. Reads that start - # beyond the end of the data return an empty string. - seekpos = self._data_offset+offset - fsize = os.path.getsize(self.fname) - actuallength = max(0, min(length, fsize-seekpos)) # XXX #1528 - if actuallength == 0: - return "" - f = open(self.fname, 'rb') - f.seek(seekpos) - return f.read(actuallength) + return "" def write_share_data(self, offset, data): pass hunk ./src/allmydata/storage/backends/null/null_backend.py 171 - def _write_lease_record(self, f, lease_number, lease_info): - offset = self._lease_offset + lease_number * self.LEASE_SIZE - f.seek(offset) - assert f.tell() == offset - f.write(lease_info.to_immutable_data()) - - def _read_num_leases(self, f): - f.seek(0x08) - (num_leases,) = struct.unpack(">L", f.read(4)) - return num_leases - - def _write_num_leases(self, f, num_leases): - f.seek(0x08) - f.write(struct.pack(">L", num_leases)) - - def _truncate_leases(self, f, num_leases): - f.truncate(self._lease_offset + num_leases * self.LEASE_SIZE) - def get_leases(self): hunk ./src/allmydata/storage/backends/null/null_backend.py 172 - """Yields a LeaseInfo instance for all leases.""" - f = open(self.fname, 'rb') - (version, unused, num_leases) = struct.unpack(">LLL", f.read(0xc)) - f.seek(self._lease_offset) - for i in range(num_leases): - data = f.read(self.LEASE_SIZE) - if data: - yield LeaseInfo().from_immutable_data(data) + pass def add_lease(self, lease): pass hunk ./src/allmydata/storage/backends/null/null_backend.py 178 def renew_lease(self, renew_secret, new_expire_time): - for i,lease in enumerate(self.get_leases()): - if constant_time_compare(lease.renew_secret, renew_secret): - # yup. See if we need to update the owner time. - if new_expire_time > lease.expiration_time: - # yes - lease.expiration_time = new_expire_time - f = open(self.fname, 'rb+') - self._write_lease_record(f, i, lease) - f.close() - return raise IndexError("unable to renew non-existent lease") def add_or_renew_lease(self, lease_info): hunk ./src/allmydata/storage/backends/null/null_backend.py 181 - try: - self.renew_lease(lease_info.renew_secret, - lease_info.expiration_time) - except IndexError: - self.add_lease(lease_info) + pass hunk ./src/allmydata/storage/backends/null/null_backend.py 184 -class MutableNullShare: +class ImmutableNullShare(NullShareBase): + implements(IStoredShare) + sharetype = "immutable" + + def close(self): + self.shareset.close_shnum(self.shnum) + + +class MutableNullShare(NullShareBase): implements(IStoredMutableShare) sharetype = "mutable" hunk ./src/allmydata/storage/backends/null/null_backend.py 195 + + def check_write_enabler(self, write_enabler): + # Null backend doesn't check write enablers. + pass + + def check_testv(self, testv): + return empty_check_testv(testv) + + def writev(self, datav, new_length): + pass + + def close(self): + pass hunk ./src/allmydata/storage/backends/null/null_backend.py 209 - """ XXX: TODO """ } [Update the S3 backend. refs #999 david-sarah@jacaranda.org**20110923205345 Ignore-this: 5ca623a17e09ddad4cab2f51b49aec0a ] { hunk ./src/allmydata/storage/backends/s3/immutable.py 11 from allmydata.storage.common import si_b2a, UnknownImmutableContainerVersionError, DataTooLargeError -# Each share file (in storage/shares/$PREFIX/$STORAGEINDEX/$SHNUM) contains +# Each share file (with key 'shares/$PREFIX/$STORAGEINDEX/$SHNUM') contains # lease information [currently inaccessible] and share data. The share data is # accessed by RIBucketWriter.write and RIBucketReader.read . hunk ./src/allmydata/storage/backends/s3/immutable.py 65 # in case a share file is copied from a disk backend, or in case we # need them in future. # TODO: filesize = size of S3 object + filesize = 0 self._end_offset = filesize - (num_leases * self.LEASE_SIZE) self._data_offset = 0xc hunk ./src/allmydata/storage/backends/s3/immutable.py 122 return "\x00"*actuallength def write_share_data(self, offset, data): - assert offset >= self._size, "offset = %r, size = %r" % (offset, self._size) + length = len(data) + precondition(offset >= self._size, "offset = %r, size = %r" % (offset, self._size)) + if self._max_size is not None and offset+length > self._max_size: + raise DataTooLargeError(self._max_size, offset, length) # TODO: write data to S3. If offset > self._size, fill the space # between with zeroes. hunk ./src/allmydata/storage/backends/s3/mutable.py 17 from allmydata.storage.backends.base import testv_compare -# The MutableDiskShare is like the ImmutableDiskShare, but used for mutable data. +# The MutableS3Share is like the ImmutableS3Share, but used for mutable data. # It has a different layout. See docs/mutable.rst for more details. # # offset size name hunk ./src/allmydata/storage/backends/s3/mutable.py 43 assert struct.calcsize(">Q") == 8, struct.calcsize(">Q") -class MutableDiskShare(object): +class MutableS3Share(object): implements(IStoredMutableShare) sharetype = "mutable" hunk ./src/allmydata/storage/backends/s3/mutable.py 111 f.close() def __repr__(self): - return ("" + return ("" % (si_b2a(self._storageindex), self._shnum, quote_filepath(self._home))) def get_used_space(self): hunk ./src/allmydata/storage/backends/s3/mutable.py 311 except IndexError: return - # These lease operations are intended for use by disk_backend.py. - # Other non-test clients should not depend on the fact that the disk - # backend stores leases in share files. - - def add_lease(self, lease_info): - precondition(lease_info.owner_num != 0) # 0 means "no lease here" - f = self._home.open('rb+') - try: - num_lease_slots = self._get_num_lease_slots(f) - empty_slot = self._get_first_empty_lease_slot(f) - if empty_slot is not None: - self._write_lease_record(f, empty_slot, lease_info) - else: - self._write_lease_record(f, num_lease_slots, lease_info) - finally: - f.close() - - def renew_lease(self, renew_secret, new_expire_time): - accepting_nodeids = set() - f = self._home.open('rb+') - try: - for (leasenum, lease) in self._enumerate_leases(f): - if constant_time_compare(lease.renew_secret, renew_secret): - # yup. See if we need to update the owner time. - if new_expire_time > lease.expiration_time: - # yes - lease.expiration_time = new_expire_time - self._write_lease_record(f, leasenum, lease) - return - accepting_nodeids.add(lease.nodeid) - finally: - f.close() - # Return the accepting_nodeids set, to give the client a chance to - # update the leases on a share that has been migrated from its - # original server to a new one. - msg = ("Unable to renew non-existent lease. I have leases accepted by" - " nodeids: ") - msg += ",".join([("'%s'" % idlib.nodeid_b2a(anid)) - for anid in accepting_nodeids]) - msg += " ." - raise IndexError(msg) - - def add_or_renew_lease(self, lease_info): - precondition(lease_info.owner_num != 0) # 0 means "no lease here" - try: - self.renew_lease(lease_info.renew_secret, - lease_info.expiration_time) - except IndexError: - self.add_lease(lease_info) - - def cancel_lease(self, cancel_secret): - """Remove any leases with the given cancel_secret. If the last lease - is cancelled, the file will be removed. Return the number of bytes - that were freed (by truncating the list of leases, and possibly by - deleting the file). Raise IndexError if there was no lease with the - given cancel_secret.""" - - # XXX can this be more like ImmutableDiskShare.cancel_lease? - - accepting_nodeids = set() - modified = 0 - remaining = 0 - blank_lease = LeaseInfo(owner_num=0, - renew_secret="\x00"*32, - cancel_secret="\x00"*32, - expiration_time=0, - nodeid="\x00"*20) - f = self._home.open('rb+') - try: - for (leasenum, lease) in self._enumerate_leases(f): - accepting_nodeids.add(lease.nodeid) - if constant_time_compare(lease.cancel_secret, cancel_secret): - self._write_lease_record(f, leasenum, blank_lease) - modified += 1 - else: - remaining += 1 - if modified: - freed_space = self._pack_leases(f) - finally: - f.close() - - if modified > 0: - if remaining == 0: - freed_space = fileutil.get_used_space(self._home) - self.unlink() - return freed_space - - msg = ("Unable to cancel non-existent lease. I have leases " - "accepted by nodeids: ") - msg += ",".join([("'%s'" % idlib.nodeid_b2a(anid)) - for anid in accepting_nodeids]) - msg += " ." - raise IndexError(msg) - - def _pack_leases(self, f): - # TODO: reclaim space from cancelled leases - return 0 - def _read_write_enabler_and_nodeid(self, f): f.seek(0) data = f.read(self.HEADER_SIZE) hunk ./src/allmydata/storage/backends/s3/mutable.py 394 pass -def create_mutable_disk_share(storageindex, shnum, fp, serverid, write_enabler, parent): - ms = MutableDiskShare(storageindex, shnum, fp, parent) +def create_mutable_s3_share(storageindex, shnum, fp, serverid, write_enabler, parent): + ms = MutableS3Share(storageindex, shnum, fp, parent) ms.create(serverid, write_enabler) del ms hunk ./src/allmydata/storage/backends/s3/mutable.py 398 - return MutableDiskShare(storageindex, shnum, fp, parent) + return MutableS3Share(storageindex, shnum, fp, parent) hunk ./src/allmydata/storage/backends/s3/s3_backend.py 10 from allmydata.storage.backends.s3.immutable import ImmutableS3Share from allmydata.storage.backends.s3.mutable import MutableS3Share -# The S3 bucket has keys of the form shares/$STORAGEINDEX/$SHARENUM - +# The S3 bucket has keys of the form shares/$PREFIX/$STORAGEINDEX/$SHNUM . class S3Backend(Backend): implements(IStorageBackend) } [Minor cleanup to disk backend. refs #999 david-sarah@jacaranda.org**20110923205510 Ignore-this: 79f92d7c2edb14cfedb167247c3f0d08 ] { hunk ./src/allmydata/storage/backends/disk/immutable.py 87 (version, unused, num_leases) = struct.unpack(">LLL", f.read(0xc)) finally: f.close() - filesize = self._home.getsize() if version != 1: msg = "sharefile %s had version %d but we wanted 1" % \ (self._home, version) hunk ./src/allmydata/storage/backends/disk/immutable.py 91 raise UnknownImmutableContainerVersionError(msg) + + filesize = self._home.getsize() self._num_leases = num_leases self._lease_offset = filesize - (num_leases * self.LEASE_SIZE) self._data_offset = 0xc } [Add 'has-immutable-readv' to server version information. refs #999 david-sarah@jacaranda.org**20110923220935 Ignore-this: c3c4358f2ab8ac503f99c968ace8efcf ] { hunk ./src/allmydata/storage/server.py 174 "delete-mutable-shares-with-zero-length-writev": True, "fills-holes-with-zero-bytes": True, "prevents-read-past-end-of-share-data": True, + "has-immutable-readv": True, }, "application-version": str(allmydata.__full_version__), } hunk ./src/allmydata/test/test_storage.py 339 sv1 = ver['http://allmydata.org/tahoe/protocols/storage/v1'] self.failUnless(sv1.get('prevents-read-past-end-of-share-data'), sv1) + def test_has_immutable_readv(self): + ss = self.create("test_has_immutable_readv") + ver = ss.remote_get_version() + sv1 = ver['http://allmydata.org/tahoe/protocols/storage/v1'] + self.failUnless(sv1.get('has-immutable-readv'), sv1) + + # TODO: test that we actually support it + def allocate(self, ss, storage_index, sharenums, size, canary=None): renew_secret = hashutil.tagged_hash("blah", "%d" % self._lease_secret.next()) cancel_secret = hashutil.tagged_hash("blah", "%d" % self._lease_secret.next()) } [util/deferredutil.py: add some utilities for asynchronous iteration. refs #999 david-sarah@jacaranda.org**20110927070947 Ignore-this: ac4946c1e5779ea64b85a1a420d34c9e ] { hunk ./src/allmydata/util/deferredutil.py 1 + +from foolscap.api import fireEventually from twisted.internet import defer # utility wrapper for DeferredList hunk ./src/allmydata/util/deferredutil.py 38 d.addCallbacks(_parseDListResult, _unwrapFirstError) return d + +def async_accumulate(accumulator, body): + """ + I execute an asynchronous loop in which, for each iteration, I eventually + call 'body' with the current value of an accumulator. 'body' should return a + (possibly deferred) pair: (result, should_continue). If should_continue is + a (possibly deferred) True value, the loop will continue with result as the + new accumulator, otherwise it will terminate. + + I return a Deferred that fires with the final result, or that fails with + the first failure of 'body'. + """ + d = defer.succeed(accumulator) + d.addCallback(body) + def _iterate((result, should_continue)): + if not should_continue: + return result + d2 = fireEventually(result) + d2.addCallback(async_accumulate, body) + return d2 + d.addCallback(_iterate) + return d + +def async_iterate(process, iterable): + """ + I iterate over the elements of 'iterable' (which may be deferred), eventually + applying 'process' to each one. 'process' should return a (possibly deferred) + boolean: True to continue the iteration, False to stop. + + I return a Deferred that fires with True if all elements of the iterable + were processed (i.e. 'process' only returned True values); with False if + the iteration was stopped by 'process' returning False; or that fails with + the first failure of either 'process' or the iterator. + """ + iterator = iter(iterable) + + def _body(accumulator): + d = defer.maybeDeferred(iterator.next) + def _cb(item): + d2 = defer.maybeDeferred(process, item) + d2.addCallback(lambda res: (res, res)) + return d2 + def _eb(f): + if f.trap(StopIteration): + return (True, False) + d.addCallbacks(_cb, _eb) + return d + + return async_accumulate(False, _body) + +def async_foldl(process, unit, iterable): + """ + I perform an asynchronous left fold, similar to Haskell 'foldl process unit iterable'. + Each call to process is eventual. + + I return a Deferred that fires with the result of the fold, or that fails with + the first failure of either 'process' or the iterator. + """ + iterator = iter(iterable) + + def _body(accumulator): + d = defer.maybeDeferred(iterator.next) + def _cb(item): + d2 = defer.maybeDeferred(process, accumulator, item) + d2.addCallback(lambda res: (res, True)) + return d2 + def _eb(f): + if f.trap(StopIteration): + return (accumulator, False) + d.addCallbacks(_cb, _eb) + return d + + return async_accumulate(unit, _body) } [test_storage.py: fix test_status_bad_disk_stats. refs #999 david-sarah@jacaranda.org**20110927071403 Ignore-this: 6108fee69a60962be2df2ad11b483a11 ] hunk ./src/allmydata/storage/backends/disk/disk_backend.py 123 def get_available_space(self): if self._readonly: return 0 - return fileutil.get_available_space(self._sharedir, self._reserved_space) + try: + return fileutil.get_available_space(self._sharedir, self._reserved_space) + except EnvironmentError: + return 0 class DiskShareSet(ShareSet): [Cleanups to disk backend. refs #999 david-sarah@jacaranda.org**20110927071544 Ignore-this: e9d3fd0e85aaf301c04342fffdc8f26 ] { hunk ./src/allmydata/storage/backends/disk/immutable.py 46 sharetype = "immutable" LEASE_SIZE = struct.calcsize(">L32s32sL") - + HEADER = ">LLL" + HEADER_SIZE = struct.calcsize(HEADER) def __init__(self, storageindex, shnum, home, finalhome=None, max_size=None): """ hunk ./src/allmydata/storage/backends/disk/immutable.py 79 # the largest length that can fit into the field. That way, even # if this does happen, the old < v1.3.0 server will still allow # clients to read the first part of the share. - self._home.setContent(struct.pack(">LLL", 1, min(2**32-1, max_size), 0) ) - self._lease_offset = max_size + 0x0c + self._home.setContent(struct.pack(self.HEADER, 1, min(2**32-1, max_size), 0) ) + self._lease_offset = self.HEADER_SIZE + max_size self._num_leases = 0 else: f = self._home.open(mode='rb') hunk ./src/allmydata/storage/backends/disk/immutable.py 85 try: - (version, unused, num_leases) = struct.unpack(">LLL", f.read(0xc)) + (version, unused, num_leases) = struct.unpack(self.HEADER, f.read(self.HEADER_SIZE)) finally: f.close() if version != 1: hunk ./src/allmydata/storage/backends/disk/immutable.py 229 """Yields a LeaseInfo instance for all leases.""" f = self._home.open(mode='rb') try: - (version, unused, num_leases) = struct.unpack(">LLL", f.read(0xc)) + (version, unused, num_leases) = struct.unpack(self.HEADER, f.read(self.HEADER_SIZE)) f.seek(self._lease_offset) for i in range(num_leases): data = f.read(self.LEASE_SIZE) } [Cleanups to S3 backend (not including Deferred changes). refs #999 david-sarah@jacaranda.org**20110927071855 Ignore-this: f0dca788190d92b1edb1ee1498fb34dc ] { hunk ./src/allmydata/storage/backends/s3/immutable.py 7 from zope.interface import implements from allmydata.interfaces import IStoredShare + from allmydata.util.assertutil import precondition from allmydata.storage.common import si_b2a, UnknownImmutableContainerVersionError, DataTooLargeError hunk ./src/allmydata/storage/backends/s3/immutable.py 29 sharetype = "immutable" LEASE_SIZE = struct.calcsize(">L32s32sL") # for compatibility + HEADER = ">LLL" + HEADER_SIZE = struct.calcsize(HEADER) hunk ./src/allmydata/storage/backends/s3/immutable.py 32 - - def __init__(self, storageindex, shnum, s3bucket, create=False, max_size=None): + def __init__(self, storageindex, shnum, s3bucket, max_size=None, data=None): """ If max_size is not None then I won't allow more than max_size to be written to me. """ hunk ./src/allmydata/storage/backends/s3/immutable.py 36 - precondition((max_size is not None) or not create, max_size, create) + precondition((max_size is not None) or (data is not None), max_size, data) self._storageindex = storageindex hunk ./src/allmydata/storage/backends/s3/immutable.py 38 + self._shnum = shnum + self._s3bucket = s3bucket self._max_size = max_size hunk ./src/allmydata/storage/backends/s3/immutable.py 41 + self._data = data hunk ./src/allmydata/storage/backends/s3/immutable.py 43 - self._s3bucket = s3bucket - si_s = si_b2a(storageindex) - self._key = "storage/shares/%s/%s/%d" % (si_s[:2], si_s, shnum) - self._shnum = shnum + sistr = self.get_storage_index_string() + self._key = "shares/%s/%s/%d" % (sistr[:2], sistr, shnum) hunk ./src/allmydata/storage/backends/s3/immutable.py 46 - if create: + if data is None: # creating share # The second field, which was the four-byte share data length in # Tahoe-LAFS versions prior to 1.3.0, is not used; we always write 0. # We also write 0 for the number of leases. hunk ./src/allmydata/storage/backends/s3/immutable.py 50 - self._home.setContent(struct.pack(">LLL", 1, 0, 0) ) - self._end_offset = max_size + 0x0c - - # TODO: start write to S3. + self._home.setContent(struct.pack(self.HEADER, 1, 0, 0) ) + self._end_offset = self.HEADER_SIZE + max_size + self._size = self.HEADER_SIZE + self._writes = [] else: hunk ./src/allmydata/storage/backends/s3/immutable.py 55 - # TODO: get header - header = "\x00"*12 - (version, unused, num_leases) = struct.unpack(">LLL", header) + (version, unused, num_leases) = struct.unpack(self.HEADER, data[:self.HEADER_SIZE]) if version != 1: hunk ./src/allmydata/storage/backends/s3/immutable.py 58 - msg = "sharefile %s had version %d but we wanted 1" % \ - (self._home, version) + msg = "%r had version %d but we wanted 1" % (self, version) raise UnknownImmutableContainerVersionError(msg) # We cannot write leases in share files, but allow them to be present hunk ./src/allmydata/storage/backends/s3/immutable.py 64 # in case a share file is copied from a disk backend, or in case we # need them in future. - # TODO: filesize = size of S3 object - filesize = 0 - self._end_offset = filesize - (num_leases * self.LEASE_SIZE) - self._data_offset = 0xc + self._size = len(data) + self._end_offset = self._size - (num_leases * self.LEASE_SIZE) + self._data_offset = self.HEADER_SIZE def __repr__(self): hunk ./src/allmydata/storage/backends/s3/immutable.py 69 - return ("" - % (si_b2a(self._storageindex), self._shnum, self._key)) + return ("" % (self._key,)) def close(self): # TODO: finalize write to S3. hunk ./src/allmydata/storage/backends/s3/immutable.py 88 return self._shnum def unlink(self): - # TODO: remove the S3 object. - pass + self._data = None + self._writes = None + return self._s3bucket.delete_object(self._key) def get_allocated_size(self): return self._max_size hunk ./src/allmydata/storage/backends/s3/immutable.py 126 if self._max_size is not None and offset+length > self._max_size: raise DataTooLargeError(self._max_size, offset, length) - # TODO: write data to S3. If offset > self._size, fill the space - # between with zeroes. - + if offset > self._size: + self._writes.append("\x00" * (offset - self._size)) + self._writes.append(data) self._size = offset + len(data) def add_lease(self, lease_info): hunk ./src/allmydata/storage/backends/s3/s3_backend.py 2 -from zope.interface import implements +import re + +from zope.interface import implements, Interface from allmydata.interfaces import IStorageBackend, IShareSet hunk ./src/allmydata/storage/backends/s3/s3_backend.py 6 -from allmydata.storage.common import si_b2a, si_a2b + +from allmydata.storage.common import si_a2b from allmydata.storage.bucket import BucketWriter from allmydata.storage.backends.base import Backend, ShareSet from allmydata.storage.backends.s3.immutable import ImmutableS3Share hunk ./src/allmydata/storage/backends/s3/s3_backend.py 15 # The S3 bucket has keys of the form shares/$PREFIX/$STORAGEINDEX/$SHNUM . +NUM_RE=re.compile("^[0-9]+$") + + +class IS3Bucket(Interface): + """ + I represent an S3 bucket. + """ + def create(self): + """ + Create this bucket. + """ + + def delete(self): + """ + Delete this bucket. + The bucket must be empty before it can be deleted. + """ + + def list_objects(self, prefix=""): + """ + Get a list of all the objects in this bucket whose object names start with + the given prefix. + """ + + def put_object(self, object_name, data, content_type=None, metadata={}): + """ + Put an object in this bucket. + Any existing object of the same name will be replaced. + """ + + def get_object(self, object_name): + """ + Get an object from this bucket. + """ + + def head_object(self, object_name): + """ + Retrieve object metadata only. + """ + + def delete_object(self, object_name): + """ + Delete an object from this bucket. + Once deleted, there is no method to restore or undelete an object. + """ + + class S3Backend(Backend): implements(IStorageBackend) hunk ./src/allmydata/storage/backends/s3/s3_backend.py 74 else: self._max_space = int(max_space) - # TODO: any set-up for S3? - # we don't actually create the corruption-advisory dir until necessary self._corruption_advisory_dir = corruption_advisory_dir hunk ./src/allmydata/storage/backends/s3/s3_backend.py 103 def __init__(self, storageindex, s3bucket): ShareSet.__init__(self, storageindex) self._s3bucket = s3bucket + sistr = self.get_storage_index_string() + self._key = 'shares/%s/%s/' % (sistr[:2], sistr) def get_overhead(self): return 0 hunk ./src/allmydata/storage/backends/s3/s3_backend.py 129 def _create_mutable_share(self, storageserver, shnum, write_enabler): # TODO serverid = storageserver.get_serverid() - return MutableS3Share(self.get_storage_index(), shnum, self._s3bucket, serverid, write_enabler, storageserver) + return MutableS3Share(self.get_storage_index(), shnum, self._s3bucket, serverid, + write_enabler, storageserver) def _clean_up_after_unlink(self): pass } [test_storage.py: fix test_no_st_blocks. refs #999 david-sarah@jacaranda.org**20110927072848 Ignore-this: 5f12b784920f87d09c97c676d0afa6f8 ] { hunk ./src/allmydata/test/test_storage.py 3034 LeaseCheckerClass = InstrumentedLeaseCheckingCrawler -class BrokenStatResults: - pass - -class No_ST_BLOCKS_LeaseCheckingCrawler(LeaseCheckingCrawler): - def stat(self, fn): - s = os.stat(fn) - bsr = BrokenStatResults() - for attrname in dir(s): - if attrname.startswith("_"): - continue - if attrname == "st_blocks": - continue - setattr(bsr, attrname, getattr(s, attrname)) - return bsr - -class No_ST_BLOCKS_StorageServer(StorageServer): - LeaseCheckerClass = No_ST_BLOCKS_LeaseCheckingCrawler - - class LeaseCrawler(unittest.TestCase, pollmixin.PollMixin, WebRenderingMixin): def setUp(self): hunk ./src/allmydata/test/test_storage.py 3830 return d def test_no_st_blocks(self): - basedir = "storage/LeaseCrawler/no_st_blocks" - fp = FilePath(basedir) - backend = DiskBackend(fp) + # TODO: replace with @patch that supports Deferreds. hunk ./src/allmydata/test/test_storage.py 3832 - # A negative 'override_lease_duration' means that the "configured-" - # space-recovered counts will be non-zero, since all shares will have - # expired by then. - expiration_policy = { - 'enabled': True, - 'mode': 'age', - 'override_lease_duration': -1000, - 'sharetypes': ('mutable', 'immutable'), - } - ss = No_ST_BLOCKS_StorageServer("\x00" * 20, backend, fp, expiration_policy=expiration_policy) + class BrokenStatResults: + pass hunk ./src/allmydata/test/test_storage.py 3835 - # make it start sooner than usual. - lc = ss.lease_checker - lc.slow_start = 0 + def call_stat(fn): + s = self.old_os_stat(fn) + bsr = BrokenStatResults() + for attrname in dir(s): + if attrname.startswith("_"): + continue + if attrname == "st_blocks": + continue + setattr(bsr, attrname, getattr(s, attrname)) + return bsr hunk ./src/allmydata/test/test_storage.py 3846 - self.make_shares(ss) - ss.setServiceParent(self.s) - def _wait(): - return bool(lc.get_state()["last-cycle-finished"] is not None) - d = self.poll(_wait) + def _cleanup(res): + os.stat = self.old_os_stat + return res hunk ./src/allmydata/test/test_storage.py 3850 - def _check(ignored): - s = lc.get_state() - last = s["history"][0] - rec = last["space-recovered"] - self.failUnlessEqual(rec["configured-buckets"], 4) - self.failUnlessEqual(rec["configured-shares"], 4) - self.failUnless(rec["configured-sharebytes"] > 0, - rec["configured-sharebytes"]) - # without the .st_blocks field in os.stat() results, we should be - # reporting diskbytes==sharebytes - self.failUnlessEqual(rec["configured-sharebytes"], - rec["configured-diskbytes"]) - d.addCallback(_check) - return d + self.old_os_stat = os.stat + try: + os.stat = call_stat + + basedir = "storage/LeaseCrawler/no_st_blocks" + fp = FilePath(basedir) + backend = DiskBackend(fp) + + # A negative 'override_lease_duration' means that the "configured-" + # space-recovered counts will be non-zero, since all shares will have + # expired by then. + expiration_policy = { + 'enabled': True, + 'mode': 'age', + 'override_lease_duration': -1000, + 'sharetypes': ('mutable', 'immutable'), + } + ss = StorageServer("\x00" * 20, backend, fp, expiration_policy=expiration_policy) + + # make it start sooner than usual. + lc = ss.lease_checker + lc.slow_start = 0 + + d = defer.succeed(None) + d.addCallback(lambda ign: self.make_shares(ss)) + d.addCallback(lambda ign: ss.setServiceParent(self.s)) + def _wait(): + return bool(lc.get_state()["last-cycle-finished"] is not None) + d.addCallback(lambda ign: self.poll(_wait)) + + def _check(ignored): + s = lc.get_state() + last = s["history"][0] + rec = last["space-recovered"] + self.failUnlessEqual(rec["configured-buckets"], 4) + self.failUnlessEqual(rec["configured-shares"], 4) + self.failUnless(rec["configured-sharebytes"] > 0, + rec["configured-sharebytes"]) + # without the .st_blocks field in os.stat() results, we should be + # reporting diskbytes==sharebytes + self.failUnlessEqual(rec["configured-sharebytes"], + rec["configured-diskbytes"]) + d.addCallback(_check) + d.addBoth(_cleanup) + return d + finally: + _cleanup(None) def test_share_corruption(self): self._poll_should_ignore_these_errors = [ } [mutable/publish.py: resolve conflicting patches. refs #999 david-sarah@jacaranda.org**20110927073530 Ignore-this: 6154a113723dc93148151288bd032439 ] { hunk ./src/allmydata/mutable/publish.py 6 import os, time from StringIO import StringIO from itertools import count -from copy import copy from zope.interface import implements from twisted.internet import defer from twisted.python import failure hunk ./src/allmydata/mutable/publish.py 867 ds = [] verification_key = self._pubkey.serialize() - - # TODO: Bad, since we remove from this same dict. We need to - # make a copy, or just use a non-iterated value. - for (shnum, writer) in self.writers.iteritems(): + for (shnum, writer) in self.writers.copy().iteritems(): writer.put_verification_key(verification_key) self.num_outstanding += 1 def _no_longer_outstanding(res): } [Work in progress for asyncifying the backend interface (necessary to call txaws methods that return Deferreds). This is incomplete so lots of tests fail. refs #999 david-sarah@jacaranda.org**20110927073903 Ignore-this: ebdc6c06c3baa9460af128ec8f5b418b ] { hunk ./src/allmydata/interfaces.py 303 def get_sharesets_for_prefix(prefix): """ - Generates IShareSet objects for all storage indices matching the - given base-32 prefix for which this backend holds shares. + Return a Deferred for an iterable containing IShareSet objects for + all storage indices matching the given base-32 prefix, for which + this backend holds shares. """ def get_shareset(storageindex): hunk ./src/allmydata/interfaces.py 311 """ Get an IShareSet object for the given storage index. + This method is synchronous. """ def fill_in_space_stats(stats): hunk ./src/allmydata/interfaces.py 325 Clients who discover hash failures in shares that they have downloaded from me will use this method to inform me about the failures. I will record their concern so that my operator can - manually inspect the shares in question. + manually inspect the shares in question. This method is synchronous. 'sharetype' is either 'mutable' or 'immutable'. 'shnum' is the integer share number. 'reason' is a human-readable explanation of the problem, hunk ./src/allmydata/interfaces.py 361 def get_shares(): """ - Generates IStoredShare objects for all completed shares in this shareset. + Returns a Deferred that fires with an iterable of IStoredShare objects + for all completed shares in this shareset. """ def has_incoming(shnum): hunk ./src/allmydata/interfaces.py 367 """ - Returns True if this shareset has an incoming (partial) share with this number, otherwise False. + Returns True if this shareset has an incoming (partial) share with this + number, otherwise False. """ def make_bucket_writer(storageserver, shnum, max_space_per_bucket, lease_info, canary): hunk ./src/allmydata/interfaces.py 398 """ Read a vector from the numbered shares in this shareset. An empty wanted_shnums list means to return data from all known shares. + Return a Deferred that fires with a dict mapping the share number + to the corresponding ReadData. @param wanted_shnums=ListOf(int) @param read_vector=ReadVector hunk ./src/allmydata/interfaces.py 403 - @return DictOf(int, ReadData): shnum -> results, with one key per share + @return DeferredOf(DictOf(int, ReadData)): shnum -> results, with one key per share """ def testv_and_readv_and_writev(storageserver, secrets, test_and_write_vectors, read_vector, expiration_time): hunk ./src/allmydata/interfaces.py 412 Perform a bunch of comparisons against the existing shares in this shareset. If they all pass: use the read vectors to extract data from all the shares, then apply a bunch of write vectors to those shares. - Return the read data, which does not include any modifications made by - the writes. + Return a Deferred that fires with a pair consisting of a boolean that is + True iff the test vectors passed, and a dict mapping the share number + to the corresponding ReadData. Reads do not include any modifications + made by the writes. See the similar method in RIStorageServer for more detail. hunk ./src/allmydata/interfaces.py 424 @param test_and_write_vectors=TestAndWriteVectorsForShares @param read_vector=ReadVector @param expiration_time=int - @return TupleOf(bool, DictOf(int, ReadData)) + @return DeferredOf(TupleOf(bool, DictOf(int, ReadData))) """ def add_or_renew_lease(lease_info): hunk ./src/allmydata/storage/backends/base.py 3 from twisted.application import service +from twisted.internet import defer from allmydata.util import fileutil, log, time_format hunk ./src/allmydata/storage/backends/base.py 6 +from allmydata.util.deferredutil import async_iterate, gatherResults from allmydata.storage.common import si_b2a from allmydata.storage.lease import LeaseInfo from allmydata.storage.bucket import BucketReader hunk ./src/allmydata/storage/backends/base.py 105 else: cancel_secret = renew_secret - shares = {} - for share in self.get_shares(): - # XXX is it correct to ignore immutable shares? Maybe get_shares should - # have a parameter saying what type it's expecting. - if share.sharetype == "mutable": - share.check_write_enabler(write_enabler) - shares[share.get_shnum()] = share - - # write_enabler is good for all existing shares - - # now evaluate test vectors - testv_is_good = True - for sharenum in test_and_write_vectors: - (testv, datav, new_length) = test_and_write_vectors[sharenum] - if sharenum in shares: - if not shares[sharenum].check_testv(testv): - storageserver.log("testv failed: [%d]: %r" % (sharenum, testv)) - testv_is_good = False - break - else: - # compare the vectors against an empty share, in which all - # reads return empty strings - if not empty_check_testv(testv): - storageserver.log("testv failed (empty): [%d] %r" % (sharenum, testv)) - testv_is_good = False - break + sharemap = {} + d = self.get_shares() + def _got_shares(shares): + d2 = defer.succeed(None) + for share in shares: + # XXX is it correct to ignore immutable shares? Maybe get_shares should + # have a parameter saying what type it's expecting. + if share.sharetype == "mutable": + d2.addCallback(lambda ign: share.check_write_enabler(write_enabler)) + sharemap[share.get_shnum()] = share hunk ./src/allmydata/storage/backends/base.py 116 - # gather the read vectors, before we do any writes - read_data = {} - for shnum, share in shares.items(): - read_data[shnum] = share.readv(read_vector) + shnums = sorted(sharemap.keys()) hunk ./src/allmydata/storage/backends/base.py 118 - ownerid = 1 # TODO - lease_info = LeaseInfo(ownerid, renew_secret, cancel_secret, - expiration_time, storageserver.get_serverid()) + # if d2 does not fail, write_enabler is good for all existing shares hunk ./src/allmydata/storage/backends/base.py 120 - if testv_is_good: - # now apply the write vectors - for shnum in test_and_write_vectors: + # now evaluate test vectors + def _check_testv(shnum): (testv, datav, new_length) = test_and_write_vectors[shnum] hunk ./src/allmydata/storage/backends/base.py 123 - if new_length == 0: - if shnum in shares: - shares[shnum].unlink() + if shnum in sharemap: + d3 = sharemap[shnum].check_testv(testv) else: hunk ./src/allmydata/storage/backends/base.py 126 - if shnum not in shares: - # allocate a new share - share = self._create_mutable_share(storageserver, shnum, write_enabler) - shares[shnum] = share - shares[shnum].writev(datav, new_length) - # and update the lease - shares[shnum].add_or_renew_lease(lease_info) + # compare the vectors against an empty share, in which all + # reads return empty strings + d3 = defer.succeed(empty_check_testv(testv)) + + def _check_result(res): + if not res: + storageserver.log("testv failed: [%d] %r" % (shnum, testv)) + return res + d3.addCallback(_check_result) + return d3 + + d2.addCallback(lambda ign: async_iterate(_check_testv, test_and_write_vectors)) hunk ./src/allmydata/storage/backends/base.py 139 - if new_length == 0: - self._clean_up_after_unlink() + def _gather(testv_is_good): + # gather the read vectors, before we do any writes + d3 = gatherResults([sharemap[shnum].readv(read_vector) for shnum in shnums]) hunk ./src/allmydata/storage/backends/base.py 143 - return (testv_is_good, read_data) + def _do_writes(reads): + read_data = {} + for i in range(len(shnums)): + read_data[shnums[i]] = reads[i] + + ownerid = 1 # TODO + lease_info = LeaseInfo(ownerid, renew_secret, cancel_secret, + expiration_time, storageserver.get_serverid()) + + d4 = defer.succeed(None) + if testv_is_good: + # now apply the write vectors + for shnum in test_and_write_vectors: + (testv, datav, new_length) = test_and_write_vectors[shnum] + if new_length == 0: + if shnum in sharemap: + d4.addCallback(lambda ign: sharemap[shnum].unlink()) + else: + if shnum not in shares: + # allocate a new share + share = self._create_mutable_share(storageserver, shnum, + write_enabler) + sharemap[shnum] = share + d4.addCallback(lambda ign: + sharemap[shnum].writev(datav, new_length)) + # and update the lease + d4.addCallback(lambda ign: + sharemap[shnum].add_or_renew_lease(lease_info)) + if new_length == 0: + d4.addCallback(lambda ign: self._clean_up_after_unlink()) + + d4.addCallback(lambda ign: (testv_is_good, read_data)) + return d4 + d3.addCallback(_do_writes) + return d3 + d2.addCallback(_gather) + return d2 + d.addCallback(_got_shares) + return d def readv(self, wanted_shnums, read_vector): """ hunk ./src/allmydata/storage/backends/base.py 192 @param read_vector=ReadVector @return DictOf(int, ReadData): shnum -> results, with one key per share """ - datavs = {} - for share in self.get_shares(): - shnum = share.get_shnum() - if not wanted_shnums or shnum in wanted_shnums: - datavs[shnum] = share.readv(read_vector) + shnums = [] + dreads = [] + d = self.get_shares() + def _got_shares(shares): + for share in shares: + # XXX is it correct to ignore immutable shares? Maybe get_shares should + # have a parameter saying what type it's expecting. + if share.sharetype == "mutable": + shnum = share.get_shnum() + if not wanted_shnums or shnum in wanted_shnums: + shnums.add(share.get_shnum()) + dreads.add(share.readv(read_vector)) + return gatherResults(dreads) + d.addCallback(_got_shares) hunk ./src/allmydata/storage/backends/base.py 207 - return datavs + def _got_reads(reads): + datavs = {} + for i in range(len(shnums)): + datavs[shnums[i]] = reads[i] + return datavs + d.addCallback(_got_reads) + return d def testv_compare(a, op, b): hunk ./src/allmydata/storage/backends/disk/disk_backend.py 5 import re from twisted.python.filepath import UnlistableError +from twisted.internet import defer from zope.interface import implements from allmydata.interfaces import IStorageBackend, IShareSet hunk ./src/allmydata/storage/backends/disk/disk_backend.py 90 sharesets.sort(key=_by_base32si) except EnvironmentError: sharesets = [] - return sharesets + return defer.succeed(sharesets) def get_shareset(self, storageindex): sharehomedir = si_si2dir(self._sharedir, storageindex) hunk ./src/allmydata/storage/backends/disk/disk_backend.py 144 fileutil.get_used_space(self._incominghomedir)) def get_shares(self): + return defer.succeed(list(self._get_shares())) + + def _get_shares(self): """ Generate IStorageBackendShare objects for shares we have for this storage index. ("Shares we have" means completed ones, excluding incoming ones.) hunk ./src/allmydata/storage/backends/disk/immutable.py 4 import struct -from zope.interface import implements +from twisted.internet import defer hunk ./src/allmydata/storage/backends/disk/immutable.py 6 +from zope.interface import implements from allmydata.interfaces import IStoredShare hunk ./src/allmydata/storage/backends/disk/immutable.py 8 + from allmydata.util import fileutil from allmydata.util.assertutil import precondition from allmydata.util.fileutil import fp_make_dirs hunk ./src/allmydata/storage/backends/disk/immutable.py 134 # allow lease changes after closing. self._home = self._finalhome self._finalhome = None + return defer.succeed(None) def get_used_space(self): hunk ./src/allmydata/storage/backends/disk/immutable.py 137 - return (fileutil.get_used_space(self._finalhome) + - fileutil.get_used_space(self._home)) + return defer.succeed(fileutil.get_used_space(self._finalhome) + + fileutil.get_used_space(self._home)) def get_storage_index(self): return self._storageindex hunk ./src/allmydata/storage/backends/disk/immutable.py 151 def unlink(self): self._home.remove() + return defer.succeed(None) def get_allocated_size(self): return self._max_size hunk ./src/allmydata/storage/backends/disk/immutable.py 157 def get_size(self): - return self._home.getsize() + return defer.succeed(self._home.getsize()) def get_data_length(self): hunk ./src/allmydata/storage/backends/disk/immutable.py 160 - return self._lease_offset - self._data_offset + return defer.succeed(self._lease_offset - self._data_offset) def readv(self, readv): datav = [] hunk ./src/allmydata/storage/backends/disk/immutable.py 170 datav.append(self._read_share_data(f, offset, length)) finally: f.close() - return datav + return defer.succeed(datav) def _read_share_data(self, f, offset, length): precondition(offset >= 0) hunk ./src/allmydata/storage/backends/disk/immutable.py 187 def read_share_data(self, offset, length): f = self._home.open(mode='rb') try: - return self._read_share_data(f, offset, length) + return defer.succeed(self._read_share_data(f, offset, length)) finally: f.close() hunk ./src/allmydata/storage/backends/disk/immutable.py 202 f.seek(real_offset) assert f.tell() == real_offset f.write(data) + return defer.succeed(None) finally: f.close() hunk ./src/allmydata/storage/backends/disk/mutable.py 4 import struct -from zope.interface import implements +from twisted.internet import defer hunk ./src/allmydata/storage/backends/disk/mutable.py 6 +from zope.interface import implements from allmydata.interfaces import IStoredMutableShare, BadWriteEnablerError hunk ./src/allmydata/storage/backends/disk/mutable.py 8 + from allmydata.util import fileutil, idlib, log from allmydata.util.assertutil import precondition from allmydata.util.hashutil import constant_time_compare hunk ./src/allmydata/storage/backends/disk/mutable.py 111 # extra leases go here, none at creation finally: f.close() + return defer.succeed(None) def __repr__(self): return ("" hunk ./src/allmydata/storage/backends/disk/mutable.py 118 % (si_b2a(self._storageindex), self._shnum, quote_filepath(self._home))) def get_used_space(self): - return fileutil.get_used_space(self._home) + return defer.succeed(fileutil.get_used_space(self._home)) def get_storage_index(self): return self._storageindex hunk ./src/allmydata/storage/backends/disk/mutable.py 131 def unlink(self): self._home.remove() + return defer.succeed(None) def _read_data_length(self, f): f.seek(self.DATA_LENGTH_OFFSET) hunk ./src/allmydata/storage/backends/disk/mutable.py 431 datav.append(self._read_share_data(f, offset, length)) finally: f.close() - return datav + return defer.succeed(datav) def get_size(self): hunk ./src/allmydata/storage/backends/disk/mutable.py 434 - return self._home.getsize() + return defer.succeed(self._home.getsize()) def get_data_length(self): f = self._home.open('rb') hunk ./src/allmydata/storage/backends/disk/mutable.py 442 data_length = self._read_data_length(f) finally: f.close() - return data_length + return defer.succeed(data_length) def check_write_enabler(self, write_enabler): f = self._home.open('rb+') hunk ./src/allmydata/storage/backends/disk/mutable.py 463 msg = "The write enabler was recorded by nodeid '%s'." % \ (idlib.nodeid_b2a(write_enabler_nodeid),) raise BadWriteEnablerError(msg) + return defer.succeed(None) def check_testv(self, testv): test_good = True hunk ./src/allmydata/storage/backends/disk/mutable.py 476 break finally: f.close() - return test_good + return defer.succeed(test_good) def writev(self, datav, new_length): f = self._home.open('rb+') hunk ./src/allmydata/storage/backends/disk/mutable.py 492 # self._change_container_size() here. finally: f.close() + return defer.succeed(None) def close(self): hunk ./src/allmydata/storage/backends/disk/mutable.py 495 - pass + return defer.succeed(None) def create_mutable_disk_share(storageindex, shnum, fp, serverid, write_enabler, parent): hunk ./src/allmydata/storage/backends/null/null_backend.py 2 -from zope.interface import implements +from twisted.internet import defer hunk ./src/allmydata/storage/backends/null/null_backend.py 4 +from zope.interface import implements from allmydata.interfaces import IStorageBackend, IShareSet, IStoredShare, IStoredMutableShare hunk ./src/allmydata/storage/backends/null/null_backend.py 6 + from allmydata.util.assertutil import precondition from allmydata.storage.backends.base import Backend, empty_check_testv from allmydata.storage.bucket import BucketWriter, BucketReader hunk ./src/allmydata/storage/backends/null/null_backend.py 37 def _by_base32si(b): return b.get_storage_index_string() sharesets.sort(key=_by_base32si) - return sharesets + return defer.succeed(sharesets) def get_shareset(self, storageindex): shareset = self._sharesets.get(storageindex, None) hunk ./src/allmydata/storage/backends/null/null_backend.py 67 return 0 def get_shares(self): + shares = [] for shnum in self._immutable_shnums: hunk ./src/allmydata/storage/backends/null/null_backend.py 69 - yield ImmutableNullShare(self, shnum) + shares.append(ImmutableNullShare(self, shnum)) for shnum in self._mutable_shnums: hunk ./src/allmydata/storage/backends/null/null_backend.py 71 - yield MutableNullShare(self, shnum) + shares.append(MutableNullShare(self, shnum)) + return defer.succeed(shares) def renew_lease(self, renew_secret, new_expiration_time): raise IndexError("no such lease to renew") hunk ./src/allmydata/storage/backends/null/null_backend.py 130 else: self._mutable_shnums.add(shnum) - return (testv_is_good, read_data) + return defer.succeed((testv_is_good, read_data)) def readv(self, wanted_shnums, read_vector): hunk ./src/allmydata/storage/backends/null/null_backend.py 133 - return {} + return defer.succeed({}) class NullShareBase(object): hunk ./src/allmydata/storage/backends/null/null_backend.py 151 return self.shnum def get_data_length(self): - return 0 + return defer.succeed(0) def get_size(self): hunk ./src/allmydata/storage/backends/null/null_backend.py 154 - return 0 + return defer.succeed(0) def get_used_space(self): hunk ./src/allmydata/storage/backends/null/null_backend.py 157 - return 0 + return defer.succeed(0) def unlink(self): hunk ./src/allmydata/storage/backends/null/null_backend.py 160 - pass + return defer.succeed(None) def readv(self, readv): datav = [] hunk ./src/allmydata/storage/backends/null/null_backend.py 166 for (offset, length) in readv: datav.append("") - return datav + return defer.succeed(datav) def read_share_data(self, offset, length): precondition(offset >= 0) hunk ./src/allmydata/storage/backends/null/null_backend.py 170 - return "" + return defer.succeed("") def write_share_data(self, offset, data): hunk ./src/allmydata/storage/backends/null/null_backend.py 173 - pass + return defer.succeed(None) def get_leases(self): pass hunk ./src/allmydata/storage/backends/null/null_backend.py 193 sharetype = "immutable" def close(self): - self.shareset.close_shnum(self.shnum) + return self.shareset.close_shnum(self.shnum) class MutableNullShare(NullShareBase): hunk ./src/allmydata/storage/backends/null/null_backend.py 202 def check_write_enabler(self, write_enabler): # Null backend doesn't check write enablers. - pass + return defer.succeed(None) def check_testv(self, testv): hunk ./src/allmydata/storage/backends/null/null_backend.py 205 - return empty_check_testv(testv) + return defer.succeed(empty_check_testv(testv)) def writev(self, datav, new_length): hunk ./src/allmydata/storage/backends/null/null_backend.py 208 - pass + return defer.succeed(None) def close(self): hunk ./src/allmydata/storage/backends/null/null_backend.py 211 - pass + return defer.succeed(None) hunk ./src/allmydata/storage/backends/s3/immutable.py 4 import struct -from zope.interface import implements +from twisted.internet import defer hunk ./src/allmydata/storage/backends/s3/immutable.py 6 +from zope.interface import implements from allmydata.interfaces import IStoredShare from allmydata.util.assertutil import precondition hunk ./src/allmydata/storage/backends/s3/immutable.py 73 return ("" % (self._key,)) def close(self): - # TODO: finalize write to S3. - pass + # This will briefly use memory equal to double the share size. + # We really want to stream writes to S3, but I don't think txaws supports that yet + # (and neither does IS3Bucket, since that's a very thin wrapper over the txaws S3 API). + self._data = "".join(self._writes) + self._writes = None + self._s3bucket.put_object(self._key, self._data) + return defer.succeed(None) def get_used_space(self): hunk ./src/allmydata/storage/backends/s3/immutable.py 82 - return self._size + return defer.succeed(self._size) def get_storage_index(self): return self._storageindex hunk ./src/allmydata/storage/backends/s3/immutable.py 102 return self._max_size def get_size(self): - return self._size + return defer.succeed(self._size) def get_data_length(self): hunk ./src/allmydata/storage/backends/s3/immutable.py 105 - return self._end_offset - self._data_offset + return defer.succeed(self._end_offset - self._data_offset) def readv(self, readv): datav = [] hunk ./src/allmydata/storage/backends/s3/immutable.py 111 for (offset, length) in readv: datav.append(self.read_share_data(offset, length)) - return datav + return defer.succeed(datav) def read_share_data(self, offset, length): precondition(offset >= 0) hunk ./src/allmydata/storage/backends/s3/immutable.py 121 seekpos = self._data_offset+offset actuallength = max(0, min(length, self._end_offset-seekpos)) if actuallength == 0: - return "" - - # TODO: perform an S3 GET request, possibly with a Content-Range header. - return "\x00"*actuallength + return defer.succeed("") + return defer.succeed(self._data[offset:offset+actuallength]) def write_share_data(self, offset, data): length = len(data) hunk ./src/allmydata/storage/backends/s3/immutable.py 134 self._writes.append("\x00" * (offset - self._size)) self._writes.append(data) self._size = offset + len(data) + return defer.succeed(None) def add_lease(self, lease_info): pass hunk ./src/allmydata/storage/backends/s3/s3_backend.py 78 self._corruption_advisory_dir = corruption_advisory_dir def get_sharesets_for_prefix(self, prefix): - # TODO: query S3 for keys matching prefix - return [] + d = self._s3bucket.list_objects('shares/%s/' % (prefix,), '/') + def _get_sharesets(res): + # XXX this enumerates all shares to get the set of SIs. + # Is there a way to enumerate SIs more efficiently? + si_strings = set() + for item in res.contents: + # XXX better error handling + path = item.key.split('/') + assert path[0:2] == ["shares", prefix] + si_strings.add(path[2]) + + # XXX we want this to be deterministic, so we return the sharesets sorted + # by their si_strings, but we shouldn't need to explicitly re-sort them + # because list_objects returns a sorted list. + return [S3ShareSet(si_a2b(s), self._s3bucket) for s in sorted(si_strings)] + d.addCallback(_get_sharesets) + return d def get_shareset(self, storageindex): return S3ShareSet(storageindex, self._s3bucket) hunk ./src/allmydata/storage/backends/s3/s3_backend.py 129 Generate IStorageBackendShare objects for shares we have for this storage index. ("Shares we have" means completed ones, excluding incoming ones.) """ - pass + d = self._s3bucket.list_objects(self._key, '/') + def _get_shares(res): + # XXX this enumerates all shares to get the set of SIs. + # Is there a way to enumerate SIs more efficiently? + shnums = [] + for item in res.contents: + # XXX better error handling + assert item.key.startswith(self._key), item.key + path = item.key.split('/') + assert len(path) == 4, path + shnumstr = path[3] + if NUM_RE.matches(shnumstr): + shnums.add(int(shnumstr)) + + return [self._get_share(shnum) for shnum in sorted(shnums)] + d.addCallback(_get_shares) + return d + + def _get_share(self, shnum): + d = self._s3bucket.get_object("%s%d" % (self._key, shnum)) + def _make_share(data): + if data.startswith(MutableS3Share.MAGIC): + return MutableS3Share(self._storageindex, shnum, self._s3bucket, data=data) + else: + # assume it's immutable + return ImmutableS3Share(self._storageindex, shnum, self._s3bucket, data=data) + d.addCallback(_make_share) + return d def has_incoming(self, shnum): # TODO: this might need to be more like the disk backend; review callers hunk ./src/allmydata/storage/bucket.py 5 import time from foolscap.api import Referenceable +from twisted.internet import defer from zope.interface import implements from allmydata.interfaces import RIBucketWriter, RIBucketReader hunk ./src/allmydata/storage/bucket.py 9 + from allmydata.util import base32, log from allmydata.util.assertutil import precondition hunk ./src/allmydata/storage/bucket.py 31 def allocated_size(self): return self._share.get_allocated_size() + def _add_latency(self, res, name, start): + self.ss.add_latency(name, time.time() - start) + self.ss.count(name) + return res + def remote_write(self, offset, data): start = time.time() precondition(not self.closed) hunk ./src/allmydata/storage/bucket.py 40 if self.throw_out_all_data: - return - self._share.write_share_data(offset, data) - self.ss.add_latency("write", time.time() - start) - self.ss.count("write") + return defer.succeed(None) + d = self._share.write_share_data(offset, data) + d.addBoth(self._add_latency, "write", start) + return d def remote_close(self): precondition(not self.closed) hunk ./src/allmydata/storage/bucket.py 49 start = time.time() - self._share.close() + d = self._share.close() # XXX should this be self._share.get_used_space() ? hunk ./src/allmydata/storage/bucket.py 51 - consumed_size = self._share.get_size() - self._share = None - - self.closed = True - self._canary.dontNotifyOnDisconnect(self._disconnect_marker) + d.addCallback(lambda ign: self._share.get_size()) + def _got_size(consumed_size): + self._share = None + self.closed = True + self._canary.dontNotifyOnDisconnect(self._disconnect_marker) hunk ./src/allmydata/storage/bucket.py 57 - self.ss.bucket_writer_closed(self, consumed_size) - self.ss.add_latency("close", time.time() - start) - self.ss.count("close") + self.ss.bucket_writer_closed(self, consumed_size) + d.addCallback(_got_size) + d.addBoth(self._add_latency, "close", start) + return d def _disconnected(self): if not self.closed: hunk ./src/allmydata/storage/bucket.py 64 - self._abort() + return self._abort() + return defer.succeed(None) def remote_abort(self): log.msg("storage: aborting write to share %r" % self._share, hunk ./src/allmydata/storage/bucket.py 72 facility="tahoe.storage", level=log.UNUSUAL) if not self.closed: self._canary.dontNotifyOnDisconnect(self._disconnect_marker) - self._abort() - self.ss.count("abort") + d = self._abort() + def _count(ign): + self.ss.count("abort") + d.addBoth(_count) + return d def _abort(self): if self.closed: hunk ./src/allmydata/storage/bucket.py 80 - return - self._share.unlink() - self._share = None + return defer.succeed(None) + d = self._share.unlink() + def _unlinked(ign): + self._share = None hunk ./src/allmydata/storage/bucket.py 85 - # We are now considered closed for further writing. We must tell - # the storage server about this so that it stops expecting us to - # use the space it allocated for us earlier. - self.closed = True - self.ss.bucket_writer_closed(self, 0) + # We are now considered closed for further writing. We must tell + # the storage server about this so that it stops expecting us to + # use the space it allocated for us earlier. + self.closed = True + self.ss.bucket_writer_closed(self, 0) + d.addCallback(_unlinked) + return d class BucketReader(Referenceable): hunk ./src/allmydata/storage/bucket.py 108 base32.b2a_l(self.storageindex[:8], 60), self.shnum) + def _add_latency(self, res, name, start): + self.ss.add_latency(name, time.time() - start) + self.ss.count(name) + return res + def remote_read(self, offset, length): start = time.time() hunk ./src/allmydata/storage/bucket.py 115 - data = self._share.read_share_data(offset, length) - self.ss.add_latency("read", time.time() - start) - self.ss.count("read") - return data + d = self._share.read_share_data(offset, length) + d.addBoth(self._add_latency, "read", start) + return d def remote_advise_corrupt_share(self, reason): return self.ss.remote_advise_corrupt_share("immutable", hunk ./src/allmydata/storage/server.py 180 } return version + def _add_latency(self, res, name, start): + self.add_latency(name, time.time() - start) + return res + def remote_allocate_buckets(self, storageindex, renew_secret, cancel_secret, sharenums, allocated_size, hunk ./src/allmydata/storage/server.py 225 # XXX should we be making the assumption here that lease info is # duplicated in all shares? alreadygot = set() - for share in shareset.get_shares(): - share.add_or_renew_lease(lease_info) - alreadygot.add(share.get_shnum()) + d = shareset.get_shares() + def _got_shares(shares): + remaining = remaining_space + for share in shares: + share.add_or_renew_lease(lease_info) + alreadygot.add(share.get_shnum()) hunk ./src/allmydata/storage/server.py 232 - for shnum in set(sharenums) - alreadygot: - if shareset.has_incoming(shnum): - # Note that we don't create BucketWriters for shnums that - # have a partial share (in incoming/), so if a second upload - # occurs while the first is still in progress, the second - # uploader will use different storage servers. - pass - elif (not limited) or (remaining_space >= max_space_per_bucket): - bw = shareset.make_bucket_writer(self, shnum, max_space_per_bucket, - lease_info, canary) - bucketwriters[shnum] = bw - self._active_writers[bw] = 1 - if limited: - remaining_space -= max_space_per_bucket - else: - # Bummer not enough space to accept this share. - pass + for shnum in set(sharenums) - alreadygot: + if shareset.has_incoming(shnum): + # Note that we don't create BucketWriters for shnums that + # have a partial share (in incoming/), so if a second upload + # occurs while the first is still in progress, the second + # uploader will use different storage servers. + pass + elif (not limited) or (remaining >= max_space_per_bucket): + bw = shareset.make_bucket_writer(self, shnum, max_space_per_bucket, + lease_info, canary) + bucketwriters[shnum] = bw + self._active_writers[bw] = 1 + if limited: + remaining -= max_space_per_bucket + else: + # Bummer not enough space to accept this share. + pass hunk ./src/allmydata/storage/server.py 250 - self.add_latency("allocate", time.time() - start) - return alreadygot, bucketwriters + return alreadygot, bucketwriters + d.addCallback(_got_shares) + d.addBoth(self._add_latency, "allocate", start) + return d def remote_add_lease(self, storageindex, renew_secret, cancel_secret, owner_num=1): hunk ./src/allmydata/storage/server.py 306 bucket. Each lease is returned as a LeaseInfo instance. This method is not for client use. XXX do we need it at all? + For the time being this is synchronous. """ return self.backend.get_shareset(storageindex).get_leases() hunk ./src/allmydata/storage/server.py 319 si_s = si_b2a(storageindex) log.msg("storage: slot_writev %s" % si_s) - try: - shareset = self.backend.get_shareset(storageindex) - expiration_time = start + 31*24*60*60 # one month from now - return shareset.testv_and_readv_and_writev(self, secrets, test_and_write_vectors, - read_vector, expiration_time) - finally: - self.add_latency("writev", time.time() - start) + shareset = self.backend.get_shareset(storageindex) + expiration_time = start + 31*24*60*60 # one month from now + + d = shareset.testv_and_readv_and_writev(self, secrets, test_and_write_vectors, + read_vector, expiration_time) + d.addBoth(self._add_latency, "writev", start) + return d def remote_slot_readv(self, storageindex, shares, readv): start = time.time() hunk ./src/allmydata/storage/server.py 334 log.msg("storage: slot_readv %s %s" % (si_s, shares), facility="tahoe.storage", level=log.OPERATIONAL) - try: - shareset = self.backend.get_shareset(storageindex) - return shareset.readv(shares, readv) - finally: - self.add_latency("readv", time.time() - start) + shareset = self.backend.get_shareset(storageindex) + d = shareset.readv(shares, readv) + d.addBoth(self._add_latency, "readv", start) + return d def remote_advise_corrupt_share(self, share_type, storage_index, shnum, reason): self.backend.advise_corrupt_share(share_type, storage_index, shnum, reason) hunk ./src/allmydata/test/test_storage.py 3094 backend = DiskBackend(fp) ss = InstrumentedStorageServer("\x00" * 20, backend, fp) + # create a few shares, with some leases on them + d = self.make_shares(ss) + d.addCallback(self._do_test_basic, ss) + return d + + def _do_test_basic(self, ign, ss): # make it start sooner than usual. lc = ss.lease_checker lc.slow_start = 0 hunk ./src/allmydata/test/test_storage.py 3107 lc.stop_after_first_bucket = True webstatus = StorageStatus(ss) - # create a few shares, with some leases on them - self.make_shares(ss) + DAY = 24*60*60 + [immutable_si_0, immutable_si_1, mutable_si_2, mutable_si_3] = self.sis # add a non-sharefile to exercise another code path hunk ./src/allmydata/test/test_storage.py 3126 ss.setServiceParent(self.s) - DAY = 24*60*60 - d = fireEventually() hunk ./src/allmydata/test/test_storage.py 3127 - # now examine the state right after the first bucket has been # processed. def _after_first_bucket(ignored): hunk ./src/allmydata/test/test_storage.py 3287 } ss = InstrumentedStorageServer("\x00" * 20, backend, fp, expiration_policy=expiration_policy) + # create a few shares, with some leases on them + d = self.make_shares(ss) + d.addCallback(self._do_test_expire_cutoff_date, ss) + return d + + def _do_test_expire_age(self, ign, ss): # make it start sooner than usual. lc = ss.lease_checker lc.slow_start = 0 hunk ./src/allmydata/test/test_storage.py 3299 lc.stop_after_first_bucket = True webstatus = StorageStatus(ss) - # create a few shares, with some leases on them - self.make_shares(ss) [immutable_si_0, immutable_si_1, mutable_si_2, mutable_si_3] = self.sis def count_shares(si): hunk ./src/allmydata/test/test_storage.py 3437 } ss = InstrumentedStorageServer("\x00" * 20, backend, fp, expiration_policy=expiration_policy) + # create a few shares, with some leases on them + d = self.make_shares(ss) + d.addCallback(self._do_test_expire_cutoff_date, ss, now, then) + return d + + def _do_test_expire_cutoff_date(self, ign, ss, now, then): # make it start sooner than usual. lc = ss.lease_checker lc.slow_start = 0 hunk ./src/allmydata/test/test_storage.py 3449 lc.stop_after_first_bucket = True webstatus = StorageStatus(ss) - # create a few shares, with some leases on them - self.make_shares(ss) [immutable_si_0, immutable_si_1, mutable_si_2, mutable_si_3] = self.sis def count_shares(si): hunk ./src/allmydata/test/test_storage.py 3595 'sharetypes': ('immutable',), } ss = StorageServer("\x00" * 20, backend, fp, expiration_policy=expiration_policy) + + # create a few shares, with some leases on them + d = self.make_shares(ss) + d.addCallback(self._do_test_only_immutable, ss, now) + return d + + def _do_test_only_immutable(self, ign, ss, now): lc = ss.lease_checker lc.slow_start = 0 webstatus = StorageStatus(ss) hunk ./src/allmydata/test/test_storage.py 3606 - self.make_shares(ss) [immutable_si_0, immutable_si_1, mutable_si_2, mutable_si_3] = self.sis # set all leases to be expirable new_expiration_time = now - 3000 + 31*24*60*60 hunk ./src/allmydata/test/test_storage.py 3664 'sharetypes': ('mutable',), } ss = StorageServer("\x00" * 20, backend, fp, expiration_policy=expiration_policy) + + # create a few shares, with some leases on them + d = self.make_shares(ss) + d.addCallback(self._do_test_only_mutable, ss, now) + return d + + def _do_test_only_mutable(self, ign, ss, now): lc = ss.lease_checker lc.slow_start = 0 webstatus = StorageStatus(ss) hunk ./src/allmydata/test/test_storage.py 3675 - self.make_shares(ss) [immutable_si_0, immutable_si_1, mutable_si_2, mutable_si_3] = self.sis # set all leases to be expirable new_expiration_time = now - 3000 + 31*24*60*60 hunk ./src/allmydata/test/test_storage.py 3759 backend = DiskBackend(fp) ss = StorageServer("\x00" * 20, backend, fp) + # create a few shares, with some leases on them + d = self.make_shares(ss) + d.addCallback(self._do_test_limited_history, ss) + return d + + def _do_test_limited_history(self, ign, ss): # make it start sooner than usual. lc = ss.lease_checker lc.slow_start = 0 hunk ./src/allmydata/test/test_storage.py 3770 lc.cpu_slice = 500 - # create a few shares, with some leases on them - self.make_shares(ss) - ss.setServiceParent(self.s) def _wait_until_15_cycles_done(): hunk ./src/allmydata/test/test_storage.py 3796 backend = DiskBackend(fp) ss = StorageServer("\x00" * 20, backend, fp) + # create a few shares, with some leases on them + d = self.make_shares(ss) + d.addCallback(self._do_test_unpredictable_future, ss) + return d + + def _do_test_unpredictable_future(self, ign, ss): # make it start sooner than usual. lc = ss.lease_checker lc.slow_start = 0 hunk ./src/allmydata/test/test_storage.py 3807 lc.cpu_slice = -1.0 # stop quickly - self.make_shares(ss) - ss.setServiceParent(self.s) d = fireEventually() hunk ./src/allmydata/test/test_storage.py 3937 fp = FilePath(basedir) backend = DiskBackend(fp) ss = InstrumentedStorageServer("\x00" * 20, backend, fp) - w = StorageStatus(ss) hunk ./src/allmydata/test/test_storage.py 3938 + # create a few shares, with some leases on them + d = self.make_shares(ss) + d.addCallback(self._do_test_share_corruption, ss) + return d + + def _do_test_share_corruption(self, ign, ss): # make it start sooner than usual. lc = ss.lease_checker lc.stop_after_first_bucket = True hunk ./src/allmydata/test/test_storage.py 3949 lc.slow_start = 0 lc.cpu_slice = 500 - - # create a few shares, with some leases on them - self.make_shares(ss) + w = StorageStatus(ss) # now corrupt one, and make sure the lease-checker keeps going [immutable_si_0, immutable_si_1, mutable_si_2, mutable_si_3] = self.sis hunk ./src/allmydata/test/test_storage.py 4043 d = self.render1(page, args={"t": ["json"]}) return d + class WebStatus(unittest.TestCase, pollmixin.PollMixin, WebRenderingMixin): def setUp(self): } [Undo an incompatible change to RIStorageServer. refs #999 david-sarah@jacaranda.org**20110928013729 Ignore-this: bea4c0f6cb71202fab942cd846eab693 ] { hunk ./src/allmydata/interfaces.py 168 def slot_testv_and_readv_and_writev(storage_index=StorageIndex, secrets=TupleOf(WriteEnablerSecret, - LeaseRenewSecret), + LeaseRenewSecret, + LeaseCancelSecret), tw_vectors=TestAndWriteVectorsForShares, r_vector=ReadVector, ): hunk ./src/allmydata/interfaces.py 193 This secret is generated by the client and stored for later comparison by the server. Each server is given a different secret. - @param cancel_secret: ignored + @param cancel_secret: This no longer allows lease cancellation, but + must still be a unique value identifying the + lease. XXX stop relying on it to be unique. The 'secrets' argument is a tuple with (write_enabler, renew_secret). The write_enabler is required to perform any write. The renew_secret hunk ./src/allmydata/storage/backends/base.py 98 # def _create_mutable_share(self, storageserver, shnum, write_enabler): # """create a mutable share with the given shnum and write_enabler""" - write_enabler = secrets[0] - renew_secret = secrets[1] - if len(secrets) > 2: - cancel_secret = secrets[2] - else: - cancel_secret = renew_secret + (write_enabler, renew_secret, cancel_secret) = secrets sharemap = {} d = self.get_shares() } [test_system.py: incorrect arguments were being passed to the constructor for MutableDiskShare. refs #999 david-sarah@jacaranda.org**20110928013857 Ignore-this: e9719f74e7e073e37537f9a71614b8a0 ] { hunk ./src/allmydata/test/test_system.py 7 from twisted.trial import unittest from twisted.internet import defer from twisted.internet import threads # CLI tests use deferToThread +from twisted.python.filepath import FilePath import allmydata from allmydata import uri hunk ./src/allmydata/test/test_system.py 421 self.fail("unable to find any share files in %s" % basedir) return shares - def _corrupt_mutable_share(self, filename, which): - msf = MutableDiskShare(filename) + def _corrupt_mutable_share(self, what, which): + (storageindex, filename, shnum) = what + msf = MutableDiskShare(storageindex, shnum, FilePath(filename)) datav = msf.readv([ (0, 1000000) ]) final_share = datav[0] assert len(final_share) < 1000000 # ought to be truncated hunk ./src/allmydata/test/test_system.py 504 output = out.getvalue() self.failUnlessEqual(rc, 0) try: - self.failUnless("Mutable slot found:\n" in output) - self.failUnless("share_type: SDMF\n" in output) + self.failUnlessIn("Mutable slot found:\n", output) + self.failUnlessIn("share_type: SDMF\n", output) peerid = idlib.nodeid_b2a(self.clients[client_num].nodeid) hunk ./src/allmydata/test/test_system.py 507 - self.failUnless(" WE for nodeid: %s\n" % peerid in output) - self.failUnless(" num_extra_leases: 0\n" in output) - self.failUnless(" secrets are for nodeid: %s\n" % peerid - in output) - self.failUnless(" SDMF contents:\n" in output) - self.failUnless(" seqnum: 1\n" in output) - self.failUnless(" required_shares: 3\n" in output) - self.failUnless(" total_shares: 10\n" in output) - self.failUnless(" segsize: 27\n" in output, (output, filename)) - self.failUnless(" datalen: 25\n" in output) + self.failUnlessIn(" WE for nodeid: %s\n" % peerid, output) + self.failUnlessIn(" num_extra_leases: 0\n", output) + self.failUnlessIn(" secrets are for nodeid: %s\n" % peerid, output) + self.failUnlessIn(" SDMF contents:\n", output) + self.failUnlessIn(" seqnum: 1\n", output) + self.failUnlessIn(" required_shares: 3\n", output) + self.failUnlessIn(" total_shares: 10\n", output) + self.failUnlessIn(" segsize: 27\n", output) + self.failUnlessIn(" datalen: 25\n", output) # the exact share_hash_chain nodes depends upon the sharenum, # and is more of a hassle to compute than I want to deal with # now hunk ./src/allmydata/test/test_system.py 519 - self.failUnless(" share_hash_chain: " in output) - self.failUnless(" block_hash_tree: 1 nodes\n" in output) + self.failUnlessIn(" share_hash_chain: ", output) + self.failUnlessIn(" block_hash_tree: 1 nodes\n", output) expected = (" verify-cap: URI:SSK-Verifier:%s:" % base32.b2a(storage_index)) self.failUnless(expected in output) hunk ./src/allmydata/test/test_system.py 596 shares = self._find_all_shares(self.basedir) ## sort by share number #shares.sort( lambda a,b: cmp(a[3], b[3]) ) - where = dict([ (shnum, filename) - for (client_num, storage_index, filename, shnum) + where = dict([ (shnum, (storageindex, filename, shnum)) + for (client_num, storageindex, filename, shnum) in shares ]) assert len(where) == 10 # this test is designed for 3-of-10 hunk ./src/allmydata/test/test_system.py 600 - for shnum, filename in where.items(): + for shnum, what in where.items(): # shares 7,8,9 are left alone. read will check # (share_hash_chain, block_hash_tree, share_data). New # seqnum+R pairs will trigger a check of (seqnum, R, IV, hunk ./src/allmydata/test/test_system.py 608 if shnum == 0: # read: this will trigger "pubkey doesn't match # fingerprint". - self._corrupt_mutable_share(filename, "pubkey") - self._corrupt_mutable_share(filename, "encprivkey") + self._corrupt_mutable_share(what, "pubkey") + self._corrupt_mutable_share(what, "encprivkey") elif shnum == 1: # triggers "signature is invalid" hunk ./src/allmydata/test/test_system.py 612 - self._corrupt_mutable_share(filename, "seqnum") + self._corrupt_mutable_share(what, "seqnum") elif shnum == 2: # triggers "signature is invalid" hunk ./src/allmydata/test/test_system.py 615 - self._corrupt_mutable_share(filename, "R") + self._corrupt_mutable_share(what, "R") elif shnum == 3: # triggers "signature is invalid" hunk ./src/allmydata/test/test_system.py 618 - self._corrupt_mutable_share(filename, "segsize") + self._corrupt_mutable_share(what, "segsize") elif shnum == 4: hunk ./src/allmydata/test/test_system.py 620 - self._corrupt_mutable_share(filename, "share_hash_chain") + self._corrupt_mutable_share(what, "share_hash_chain") elif shnum == 5: hunk ./src/allmydata/test/test_system.py 622 - self._corrupt_mutable_share(filename, "block_hash_tree") + self._corrupt_mutable_share(what, "block_hash_tree") elif shnum == 6: hunk ./src/allmydata/test/test_system.py 624 - self._corrupt_mutable_share(filename, "share_data") + self._corrupt_mutable_share(what, "share_data") # other things to correct: IV, signature # 7,8,9 are left alone } [test_system.py: more debug output for a failing check in test_filesystem. refs #999 david-sarah@jacaranda.org**20110928014019 Ignore-this: e8bb77b8f7db12db7cd69efb6e0ed130 ] hunk ./src/allmydata/test/test_system.py 1371 self.failUnlessEqual(rc, 0) out.seek(0) descriptions = [sfn.strip() for sfn in out.readlines()] - self.failUnlessEqual(len(descriptions), 30) + self.failUnlessEqual(len(descriptions), 30, repr((cmd, descriptions))) matching = [line for line in descriptions if line.startswith("CHK %s " % storage_index_s)] [scripts/debug.py: fix incorrect arguments to dump_immutable_share. refs #999 david-sarah@jacaranda.org**20110928014049 Ignore-this: 1078ee3f06a2f36b29e0cf694d2851cd ] hunk ./src/allmydata/scripts/debug.py 52 return dump_mutable_share(options, share) else: assert share.sharetype == "immutable", share.sharetype - return dump_immutable_share(options) + return dump_immutable_share(options, share) def dump_immutable_share(options, share): out = options.stdout [mutable/publish.py: don't crash if there are no writers in _report_verinfo. refs #999 david-sarah@jacaranda.org**20110928014126 Ignore-this: 9999c82bb3057f755a6e86baeafb8a39 ] hunk ./src/allmydata/mutable/publish.py 885 def _record_verinfo(self): - self.versioninfo = self.writers.values()[0].get_verinfo() + writers = self.writers.values() + if len(writers) > 0: + self.versioninfo = writers[0].get_verinfo() def _connection_problem(self, f, writer): [Use factory functions to create share objects rather than their constructors, to allow the factory to return a Deferred. Also change some methods on IShareSet and IStoredShare to return Deferreds. Refactor some constants associated with mutable shares. refs #999 david-sarah@jacaranda.org**20110928052324 Ignore-this: bce0ac02f475bcf31b0e3b340cd91198 ] { hunk ./src/allmydata/interfaces.py 377 def make_bucket_writer(storageserver, shnum, max_space_per_bucket, lease_info, canary): """ Create a bucket writer that can be used to write data to a given share. + Returns a Deferred that fires with the bucket writer. @param storageserver=RIStorageServer @param shnum=int: A share number in this shareset hunk ./src/allmydata/interfaces.py 386 @param lease_info=LeaseInfo: The initial lease information @param canary=Referenceable: If the canary is lost before close(), the bucket is deleted. - @return an IStorageBucketWriter for the given share + @return a Deferred for an IStorageBucketWriter for the given share """ def make_bucket_reader(storageserver, share): hunk ./src/allmydata/interfaces.py 462 for lazy evaluation, such that in many use cases substantially less than all of the share data will be accessed. """ + def load(): + """ + Load header information for this share from disk, and return a Deferred that + fires when done. A user of this instance should wait until this Deferred has + fired before calling the get_data_length, get_size or get_used_space methods. + """ + def close(): """ Complete writing to this share. hunk ./src/allmydata/interfaces.py 510 Signal that this share can be removed from the backend storage. This does not guarantee that the share data will be immediately inaccessible, or that it will be securely erased. + Returns a Deferred that fires after the share has been removed. """ def readv(read_vector): hunk ./src/allmydata/interfaces.py 515 """ - XXX + Given a list of (offset, length) pairs, return a Deferred that fires with + a list of read results. """ hunk ./src/allmydata/interfaces.py 521 class IStoredMutableShare(IStoredShare): + def create(serverid, write_enabler): + """ + Create an empty mutable share with the given serverid and write enabler. + Return a Deferred that fires when the share has been created. + """ + def check_write_enabler(write_enabler): """ XXX hunk ./src/allmydata/mutable/layout.py 76 OFFSETS = ">LLLLQQ" OFFSETS_LENGTH = struct.calcsize(OFFSETS) +# our sharefiles share with a recognizable string, plus some random +# binary data to reduce the chance that a regular text file will look +# like a sharefile. +MUTABLE_MAGIC = "Tahoe mutable container v1\n" + "\x75\x09\x44\x03\x8e" + # These are still used for some tests. def unpack_header(data): o = {} hunk ./src/allmydata/scripts/debug.py 940 prefix = f.read(32) finally: f.close() + + # XXX this doesn't use the preferred load_[im]mutable_disk_share factory + # functions to load share objects, because they return Deferreds. Watch out + # for constructor argument changes. if prefix == MutableDiskShare.MAGIC: # mutable hunk ./src/allmydata/scripts/debug.py 946 - m = MutableDiskShare("", 0, fp) + m = MutableDiskShare(fp, "", 0) f = fp.open("rb") try: f.seek(m.DATA_OFFSET) hunk ./src/allmydata/scripts/debug.py 965 flip_bit(start, end) else: # otherwise assume it's immutable - f = ImmutableDiskShare("", 0, fp) + f = ImmutableDiskShare(fp, "", 0) bp = ReadBucketProxy(None, None, '') offsets = bp._parse_offsets(f.read_share_data(0, 0x24)) start = f._data_offset + offsets["data"] hunk ./src/allmydata/storage/backends/disk/disk_backend.py 13 from allmydata.storage.common import si_b2a, si_a2b from allmydata.storage.bucket import BucketWriter from allmydata.storage.backends.base import Backend, ShareSet -from allmydata.storage.backends.disk.immutable import ImmutableDiskShare -from allmydata.storage.backends.disk.mutable import MutableDiskShare, create_mutable_disk_share +from allmydata.storage.backends.disk.immutable import load_immutable_disk_share, create_immutable_disk_share +from allmydata.storage.backends.disk.mutable import load_mutable_disk_share, create_mutable_disk_share +from allmydata.mutable.layout import MUTABLE_MAGIC + # storage/ # storage/shares/incoming hunk ./src/allmydata/storage/backends/disk/disk_backend.py 37 return newfp.child(sia) -def get_share(storageindex, shnum, fp): - f = fp.open('rb') +def get_disk_share(home, storageindex, shnum): + f = home.open('rb') try: hunk ./src/allmydata/storage/backends/disk/disk_backend.py 40 - prefix = f.read(32) + prefix = f.read(len(MUTABLE_MAGIC)) finally: f.close() hunk ./src/allmydata/storage/backends/disk/disk_backend.py 44 - if prefix == MutableDiskShare.MAGIC: - return MutableDiskShare(storageindex, shnum, fp) + if prefix == MUTABLE_MAGIC: + return load_mutable_disk_share(home, storageindex, shnum) else: # assume it's immutable hunk ./src/allmydata/storage/backends/disk/disk_backend.py 48 - return ImmutableDiskShare(storageindex, shnum, fp) + return load_immutable_disk_share(home, storageindex, shnum) class DiskBackend(Backend): hunk ./src/allmydata/storage/backends/disk/disk_backend.py 159 if not NUM_RE.match(shnumstr): continue sharehome = self._sharehomedir.child(shnumstr) - yield get_share(self.get_storage_index(), int(shnumstr), sharehome) + yield get_disk_share(sharehome, self.get_storage_index(), int(shnumstr)) except UnlistableError: # There is no shares directory at all. pass hunk ./src/allmydata/storage/backends/disk/disk_backend.py 172 def make_bucket_writer(self, storageserver, shnum, max_space_per_bucket, lease_info, canary): finalhome = self._sharehomedir.child(str(shnum)) incominghome = self._incominghomedir.child(str(shnum)) - immsh = ImmutableDiskShare(self.get_storage_index(), shnum, incominghome, finalhome, - max_size=max_space_per_bucket) - bw = BucketWriter(storageserver, immsh, lease_info, canary) - if self._discard_storage: - bw.throw_out_all_data = True - return bw + d = create_immutable_disk_share(incominghome, finalhome, max_space_per_bucket, + self.get_storage_index(), shnum) + def _created(immsh): + bw = BucketWriter(storageserver, immsh, lease_info, canary) + if self._discard_storage: + bw.throw_out_all_data = True + return bw + d.addCallback(_created) + return d def _create_mutable_share(self, storageserver, shnum, write_enabler): fileutil.fp_make_dirs(self._sharehomedir) hunk ./src/allmydata/storage/backends/disk/disk_backend.py 186 sharehome = self._sharehomedir.child(str(shnum)) serverid = storageserver.get_serverid() - return create_mutable_disk_share(self.get_storage_index(), shnum, sharehome, serverid, write_enabler, storageserver) + return create_mutable_disk_share(sharehome, serverid, write_enabler, storageserver, + self.get_storage_index(), shnum) def _clean_up_after_unlink(self): fileutil.fp_rmdir_if_empty(self._sharehomedir) hunk ./src/allmydata/storage/backends/disk/immutable.py 51 HEADER = ">LLL" HEADER_SIZE = struct.calcsize(HEADER) - def __init__(self, storageindex, shnum, home, finalhome=None, max_size=None): + def __init__(self, home, storageindex, shnum, finalhome=None, max_size=None): """ If max_size is not None then I won't allow more than max_size to be written to me. If finalhome is not None (meaning that we are creating the share) then max_size hunk ./src/allmydata/storage/backends/disk/immutable.py 56 must not be None. + + Clients should use the load_immutable_disk_share and create_immutable_disk_share + factory functions rather than creating instances directly. """ precondition((max_size is not None) or (finalhome is None), max_size, finalhome) self._storageindex = storageindex hunk ./src/allmydata/storage/backends/disk/immutable.py 101 filesize = self._home.getsize() self._num_leases = num_leases self._lease_offset = filesize - (num_leases * self.LEASE_SIZE) - self._data_offset = 0xc + self._data_offset = self.HEADER_SIZE + self._loaded = False def __repr__(self): return ("" hunk ./src/allmydata/storage/backends/disk/immutable.py 108 % (si_b2a(self._storageindex), self._shnum, quote_filepath(self._home))) + def load(self): + self._loaded = True + return defer.succeed(self) + def close(self): fileutil.fp_make_dirs(self._finalhome.parent()) self._home.moveTo(self._finalhome) hunk ./src/allmydata/storage/backends/disk/immutable.py 145 return defer.succeed(None) def get_used_space(self): + assert self._loaded return defer.succeed(fileutil.get_used_space(self._finalhome) + fileutil.get_used_space(self._home)) hunk ./src/allmydata/storage/backends/disk/immutable.py 166 return self._max_size def get_size(self): + assert self._loaded return defer.succeed(self._home.getsize()) def get_data_length(self): hunk ./src/allmydata/storage/backends/disk/immutable.py 170 + assert self._loaded return defer.succeed(self._lease_offset - self._data_offset) def readv(self, readv): hunk ./src/allmydata/storage/backends/disk/immutable.py 325 space_freed = fileutil.get_used_space(self._home) self.unlink() return space_freed + + +def load_immutable_disk_share(home, storageindex=None, shnum=None): + imms = ImmutableDiskShare(home, storageindex=storageindex, shnum=shnum) + return imms.load() + +def create_immutable_disk_share(home, finalhome, max_size, storageindex=None, shnum=None): + imms = ImmutableDiskShare(home, finalhome=finalhome, max_size=max_size, + storageindex=storageindex, shnum=shnum) + return imms.load() hunk ./src/allmydata/storage/backends/disk/mutable.py 17 DataTooLargeError from allmydata.storage.lease import LeaseInfo from allmydata.storage.backends.base import testv_compare +from allmydata.mutable.layout import MUTABLE_MAGIC # The MutableDiskShare is like the ImmutableDiskShare, but used for mutable data. hunk ./src/allmydata/storage/backends/disk/mutable.py 58 DATA_OFFSET = HEADER_SIZE + 4*LEASE_SIZE assert DATA_OFFSET == 468, DATA_OFFSET - # our sharefiles share with a recognizable string, plus some random - # binary data to reduce the chance that a regular text file will look - # like a sharefile. - MAGIC = "Tahoe mutable container v1\n" + "\x75\x09\x44\x03\x8e" + MAGIC = MUTABLE_MAGIC assert len(MAGIC) == 32 MAX_SIZE = 2*1000*1000*1000 # 2GB, kind of arbitrary # TODO: decide upon a policy for max share size hunk ./src/allmydata/storage/backends/disk/mutable.py 63 - def __init__(self, storageindex, shnum, home, parent=None): + def __init__(self, home, storageindex, shnum, parent=None): + """ + Clients should use the load_mutable_disk_share and create_mutable_disk_share + factory functions rather than creating instances directly. + """ self._storageindex = storageindex self._shnum = shnum self._home = home hunk ./src/allmydata/storage/backends/disk/mutable.py 87 finally: f.close() self.parent = parent # for logging + self._loaded = False def log(self, *args, **kwargs): if self.parent: hunk ./src/allmydata/storage/backends/disk/mutable.py 93 return self.parent.log(*args, **kwargs) + def load(self): + self._loaded = True + return defer.succeed(self) + def create(self, serverid, write_enabler): assert not self._home.exists() data_length = 0 hunk ./src/allmydata/storage/backends/disk/mutable.py 118 # extra leases go here, none at creation finally: f.close() - return defer.succeed(None) + return defer.succeed(self) def __repr__(self): return ("" hunk ./src/allmydata/storage/backends/disk/mutable.py 125 % (si_b2a(self._storageindex), self._shnum, quote_filepath(self._home))) def get_used_space(self): - return defer.succeed(fileutil.get_used_space(self._home)) + assert self._loaded + return fileutil.get_used_space(self._home) def get_storage_index(self): return self._storageindex hunk ./src/allmydata/storage/backends/disk/mutable.py 442 return defer.succeed(datav) def get_size(self): - return defer.succeed(self._home.getsize()) + assert self._loaded + return self._home.getsize() def get_data_length(self): hunk ./src/allmydata/storage/backends/disk/mutable.py 446 + assert self._loaded f = self._home.open('rb') try: data_length = self._read_data_length(f) hunk ./src/allmydata/storage/backends/disk/mutable.py 452 finally: f.close() - return defer.succeed(data_length) + return data_length def check_write_enabler(self, write_enabler): f = self._home.open('rb+') hunk ./src/allmydata/storage/backends/disk/mutable.py 508 return defer.succeed(None) -def create_mutable_disk_share(storageindex, shnum, fp, serverid, write_enabler, parent): - ms = MutableDiskShare(storageindex, shnum, fp, parent) - ms.create(serverid, write_enabler) - del ms - return MutableDiskShare(storageindex, shnum, fp, parent) +def load_mutable_disk_share(home, storageindex=None, shnum=None, parent=None): + ms = MutableDiskShare(home, storageindex, shnum, parent) + return ms.load() + +def create_mutable_disk_share(home, serverid, write_enabler, storageindex=None, shnum=None, parent=None): + ms = MutableDiskShare(home, storageindex, shnum, parent) + return ms.create(serverid, write_enabler) hunk ./src/allmydata/storage/backends/null/null_backend.py 69 def get_shares(self): shares = [] for shnum in self._immutable_shnums: - shares.append(ImmutableNullShare(self, shnum)) + shares.append(load_immutable_null_share(self, shnum)) for shnum in self._mutable_shnums: hunk ./src/allmydata/storage/backends/null/null_backend.py 71 - shares.append(MutableNullShare(self, shnum)) + shares.append(load_mutable_null_share(self, shnum)) return defer.succeed(shares) def renew_lease(self, renew_secret, new_expiration_time): hunk ./src/allmydata/storage/backends/null/null_backend.py 94 def make_bucket_writer(self, storageserver, shnum, max_space_per_bucket, lease_info, canary): self._incoming_shnums.add(shnum) - immutableshare = ImmutableNullShare(self, shnum) + immutableshare = load_immutable_null_share(self, shnum) bw = BucketWriter(storageserver, immutableshare, lease_info, canary) bw.throw_out_all_data = True return bw hunk ./src/allmydata/storage/backends/null/null_backend.py 140 def __init__(self, shareset, shnum): self.shareset = shareset self.shnum = shnum + self._loaded = False + + def load(self): + self._loaded = True + return defer.succeed(self) def get_storage_index(self): return self.shareset.get_storage_index() hunk ./src/allmydata/storage/backends/null/null_backend.py 156 return self.shnum def get_data_length(self): - return defer.succeed(0) + assert self._loaded + return 0 def get_size(self): hunk ./src/allmydata/storage/backends/null/null_backend.py 160 - return defer.succeed(0) + assert self._loaded + return 0 def get_used_space(self): hunk ./src/allmydata/storage/backends/null/null_backend.py 164 - return defer.succeed(0) + assert self._loaded + return 0 def unlink(self): return defer.succeed(None) hunk ./src/allmydata/storage/backends/null/null_backend.py 208 implements(IStoredMutableShare) sharetype = "mutable" + def create(self, serverid, write_enabler): + return defer.succeed(self) + def check_write_enabler(self, write_enabler): # Null backend doesn't check write enablers. return defer.succeed(None) hunk ./src/allmydata/storage/backends/null/null_backend.py 223 def close(self): return defer.succeed(None) + + +def load_immutable_null_share(shareset, shnum): + return ImmutableNullShare(shareset, shnum).load() + +def create_immutable_null_share(shareset, shnum): + return ImmutableNullShare(shareset, shnum).load() + +def load_mutable_null_share(shareset, shnum): + return MutableNullShare(shareset, shnum).load() + +def create_mutable_null_share(shareset, shnum): + return MutableNullShare(shareset, shnum).load() hunk ./src/allmydata/storage/backends/s3/immutable.py 11 from allmydata.util.assertutil import precondition from allmydata.storage.common import si_b2a, UnknownImmutableContainerVersionError, DataTooLargeError +from allmydata.storage.backends.s3.s3_common import get_s3_share_key # Each share file (with key 'shares/$PREFIX/$STORAGEINDEX/$SHNUM') contains hunk ./src/allmydata/storage/backends/s3/immutable.py 34 HEADER = ">LLL" HEADER_SIZE = struct.calcsize(HEADER) - def __init__(self, storageindex, shnum, s3bucket, max_size=None, data=None): + def __init__(self, s3bucket, storageindex, shnum, max_size=None, data=None): """ If max_size is not None then I won't allow more than max_size to be written to me. hunk ./src/allmydata/storage/backends/s3/immutable.py 37 + + Clients should use the load_immutable_s3_share and create_immutable_s3_share + factory functions rather than creating instances directly. """ hunk ./src/allmydata/storage/backends/s3/immutable.py 41 - precondition((max_size is not None) or (data is not None), max_size, data) + self._s3bucket = s3bucket self._storageindex = storageindex self._shnum = shnum hunk ./src/allmydata/storage/backends/s3/immutable.py 44 - self._s3bucket = s3bucket self._max_size = max_size self._data = data hunk ./src/allmydata/storage/backends/s3/immutable.py 46 + self._key = get_s3_share_key(storageindex, shnum) + self._data_offset = self.HEADER_SIZE + self._loaded = False hunk ./src/allmydata/storage/backends/s3/immutable.py 50 - sistr = self.get_storage_index_string() - self._key = "shares/%s/%s/%d" % (sistr[:2], sistr, shnum) + def __repr__(self): + return ("" % (self._key,)) hunk ./src/allmydata/storage/backends/s3/immutable.py 53 - if data is None: # creating share + def load(self): + if self._max_size is not None: # creating share # The second field, which was the four-byte share data length in # Tahoe-LAFS versions prior to 1.3.0, is not used; we always write 0. # We also write 0 for the number of leases. hunk ./src/allmydata/storage/backends/s3/immutable.py 59 self._home.setContent(struct.pack(self.HEADER, 1, 0, 0) ) - self._end_offset = self.HEADER_SIZE + max_size + self._end_offset = self.HEADER_SIZE + self._max_size self._size = self.HEADER_SIZE self._writes = [] hunk ./src/allmydata/storage/backends/s3/immutable.py 62 + self._loaded = True + return defer.succeed(None) + + if self._data is None: + # If we don't already have the data, get it from S3. + d = self._s3bucket.get_object(self._key) else: hunk ./src/allmydata/storage/backends/s3/immutable.py 69 - (version, unused, num_leases) = struct.unpack(self.HEADER, data[:self.HEADER_SIZE]) + d = defer.succeed(self._data) + + def _got_data(data): + self._data = data + header = self._data[:self.HEADER_SIZE] + (version, unused, num_leases) = struct.unpack(self.HEADER, header) if version != 1: msg = "%r had version %d but we wanted 1" % (self, version) hunk ./src/allmydata/storage/backends/s3/immutable.py 83 # We cannot write leases in share files, but allow them to be present # in case a share file is copied from a disk backend, or in case we # need them in future. - self._size = len(data) + self._size = len(self._data) self._end_offset = self._size - (num_leases * self.LEASE_SIZE) hunk ./src/allmydata/storage/backends/s3/immutable.py 85 - self._data_offset = self.HEADER_SIZE - - def __repr__(self): - return ("" % (self._key,)) + self._loaded = True + d.addCallback(_got_data) + return d def close(self): # This will briefly use memory equal to double the share size. hunk ./src/allmydata/storage/backends/s3/immutable.py 92 # We really want to stream writes to S3, but I don't think txaws supports that yet - # (and neither does IS3Bucket, since that's a very thin wrapper over the txaws S3 API). + # (and neither does IS3Bucket, since that's a thin wrapper over the txaws S3 API). + self._data = "".join(self._writes) hunk ./src/allmydata/storage/backends/s3/immutable.py 95 - self._writes = None + del self._writes self._s3bucket.put_object(self._key, self._data) return defer.succeed(None) hunk ./src/allmydata/storage/backends/s3/immutable.py 100 def get_used_space(self): - return defer.succeed(self._size) + return self._size def get_storage_index(self): return self._storageindex hunk ./src/allmydata/storage/backends/s3/immutable.py 120 return self._max_size def get_size(self): - return defer.succeed(self._size) + return self._size def get_data_length(self): hunk ./src/allmydata/storage/backends/s3/immutable.py 123 - return defer.succeed(self._end_offset - self._data_offset) + return self._end_offset - self._data_offset def readv(self, readv): datav = [] hunk ./src/allmydata/storage/backends/s3/immutable.py 156 def add_lease(self, lease_info): pass + + +def load_immutable_s3_share(s3bucket, storageindex, shnum, data=None): + return ImmutableS3Share(s3bucket, storageindex, shnum, data=data).load() + +def create_immutable_s3_share(s3bucket, storageindex, shnum, max_size): + return ImmutableS3Share(s3bucket, storageindex, shnum, max_size=max_size).load() hunk ./src/allmydata/storage/backends/s3/mutable.py 4 import struct +from twisted.internet import defer + from zope.interface import implements from allmydata.interfaces import IStoredMutableShare, BadWriteEnablerError hunk ./src/allmydata/storage/backends/s3/mutable.py 17 DataTooLargeError from allmydata.storage.lease import LeaseInfo from allmydata.storage.backends.base import testv_compare +from allmydata.mutable.layout import MUTABLE_MAGIC # The MutableS3Share is like the ImmutableS3Share, but used for mutable data. hunk ./src/allmydata/storage/backends/s3/mutable.py 58 DATA_OFFSET = HEADER_SIZE + 4*LEASE_SIZE assert DATA_OFFSET == 468, DATA_OFFSET - # our sharefiles share with a recognizable string, plus some random - # binary data to reduce the chance that a regular text file will look - # like a sharefile. - MAGIC = "Tahoe mutable container v1\n" + "\x75\x09\x44\x03\x8e" + MAGIC = MUTABLE_MAGIC assert len(MAGIC) == 32 MAX_SIZE = 2*1000*1000*1000 # 2GB, kind of arbitrary # TODO: decide upon a policy for max share size hunk ./src/allmydata/storage/backends/s3/mutable.py 63 - def __init__(self, storageindex, shnum, home, parent=None): + def __init__(self, home, storageindex, shnum, parent=None): + """ + Clients should use the load_mutable_s3_share and create_mutable_s3_share + factory functions rather than creating instances directly. + """ self._storageindex = storageindex self._shnum = shnum self._home = home hunk ./src/allmydata/storage/backends/s3/mutable.py 87 finally: f.close() self.parent = parent # for logging + self._loaded = False def log(self, *args, **kwargs): if self.parent: hunk ./src/allmydata/storage/backends/s3/mutable.py 93 return self.parent.log(*args, **kwargs) + def load(self): + self._loaded = True + return defer.succeed(self) + def create(self, serverid, write_enabler): assert not self._home.exists() data_length = 0 hunk ./src/allmydata/storage/backends/s3/mutable.py 118 # extra leases go here, none at creation finally: f.close() + self._loaded = True + return defer.succeed(self) def __repr__(self): return ("" hunk ./src/allmydata/storage/backends/s3/mutable.py 126 % (si_b2a(self._storageindex), self._shnum, quote_filepath(self._home))) def get_used_space(self): + assert self._loaded return fileutil.get_used_space(self._home) def get_storage_index(self): hunk ./src/allmydata/storage/backends/s3/mutable.py 140 def unlink(self): self._home.remove() + return defer.succeed(None) def _read_data_length(self, f): f.seek(self.DATA_LENGTH_OFFSET) hunk ./src/allmydata/storage/backends/s3/mutable.py 342 datav.append(self._read_share_data(f, offset, length)) finally: f.close() - return datav + return defer.succeed(datav) def get_size(self): hunk ./src/allmydata/storage/backends/s3/mutable.py 345 + assert self._loaded return self._home.getsize() def get_data_length(self): hunk ./src/allmydata/storage/backends/s3/mutable.py 349 + assert self._loaded f = self._home.open('rb') try: data_length = self._read_data_length(f) hunk ./src/allmydata/storage/backends/s3/mutable.py 376 msg = "The write enabler was recorded by nodeid '%s'." % \ (idlib.nodeid_b2a(write_enabler_nodeid),) raise BadWriteEnablerError(msg) + return defer.succeed(None) def check_testv(self, testv): test_good = True hunk ./src/allmydata/storage/backends/s3/mutable.py 389 break finally: f.close() - return test_good + return defer.succeed(test_good) def writev(self, datav, new_length): f = self._home.open('rb+') hunk ./src/allmydata/storage/backends/s3/mutable.py 405 # self._change_container_size() here. finally: f.close() + return defer.succeed(None) def close(self): hunk ./src/allmydata/storage/backends/s3/mutable.py 408 - pass + return defer.succeed(None) + hunk ./src/allmydata/storage/backends/s3/mutable.py 411 +def load_mutable_s3_share(home, storageindex=None, shnum=None, parent=None): + return MutableS3Share(home, storageindex, shnum, parent).load() hunk ./src/allmydata/storage/backends/s3/mutable.py 414 -def create_mutable_s3_share(storageindex, shnum, fp, serverid, write_enabler, parent): - ms = MutableS3Share(storageindex, shnum, fp, parent) - ms.create(serverid, write_enabler) - del ms - return MutableS3Share(storageindex, shnum, fp, parent) +def create_mutable_s3_share(home, serverid, write_enabler, storageindex=None, shnum=None, parent=None): + return MutableS3Share(home, storageindex, shnum, parent).create(serverid, write_enabler) hunk ./src/allmydata/storage/backends/s3/s3_backend.py 2 -import re - -from zope.interface import implements, Interface +from zope.interface import implements from allmydata.interfaces import IStorageBackend, IShareSet hunk ./src/allmydata/storage/backends/s3/s3_backend.py 5 +from allmydata.util.deferredutil import gatherResults from allmydata.storage.common import si_a2b from allmydata.storage.bucket import BucketWriter from allmydata.storage.backends.base import Backend, ShareSet hunk ./src/allmydata/storage/backends/s3/s3_backend.py 9 -from allmydata.storage.backends.s3.immutable import ImmutableS3Share -from allmydata.storage.backends.s3.mutable import MutableS3Share - -# The S3 bucket has keys of the form shares/$PREFIX/$STORAGEINDEX/$SHNUM . - -NUM_RE=re.compile("^[0-9]+$") - - -class IS3Bucket(Interface): - """ - I represent an S3 bucket. - """ - def create(self): - """ - Create this bucket. - """ - - def delete(self): - """ - Delete this bucket. - The bucket must be empty before it can be deleted. - """ - - def list_objects(self, prefix=""): - """ - Get a list of all the objects in this bucket whose object names start with - the given prefix. - """ - - def put_object(self, object_name, data, content_type=None, metadata={}): - """ - Put an object in this bucket. - Any existing object of the same name will be replaced. - """ - - def get_object(self, object_name): - """ - Get an object from this bucket. - """ - - def head_object(self, object_name): - """ - Retrieve object metadata only. - """ - - def delete_object(self, object_name): - """ - Delete an object from this bucket. - Once deleted, there is no method to restore or undelete an object. - """ +from allmydata.storage.backends.s3.immutable import load_immutable_s3_share, create_immutable_s3_share +from allmydata.storage.backends.s3.mutable import load_mutable_s3_share, create_mutable_s3_share +from allmydata.storage.backends.s3.s3_common import get_s3_share_key, NUM_RE +from allmydata.mutable.layout import MUTABLE_MAGIC class S3Backend(Backend): hunk ./src/allmydata/storage/backends/s3/s3_backend.py 71 def __init__(self, storageindex, s3bucket): ShareSet.__init__(self, storageindex) self._s3bucket = s3bucket - sistr = self.get_storage_index_string() - self._key = 'shares/%s/%s/' % (sistr[:2], sistr) + self._key = get_s3_share_key(storageindex) def get_overhead(self): return 0 hunk ./src/allmydata/storage/backends/s3/s3_backend.py 87 # Is there a way to enumerate SIs more efficiently? shnums = [] for item in res.contents: - # XXX better error handling assert item.key.startswith(self._key), item.key path = item.key.split('/') hunk ./src/allmydata/storage/backends/s3/s3_backend.py 89 - assert len(path) == 4, path - shnumstr = path[3] - if NUM_RE.matches(shnumstr): - shnums.add(int(shnumstr)) + if len(path) == 4: + shnumstr = path[3] + if NUM_RE.match(shnumstr): + shnums.add(int(shnumstr)) hunk ./src/allmydata/storage/backends/s3/s3_backend.py 94 - return [self._get_share(shnum) for shnum in sorted(shnums)] + return gatherResults([self._load_share(shnum) for shnum in sorted(shnums)]) d.addCallback(_get_shares) return d hunk ./src/allmydata/storage/backends/s3/s3_backend.py 98 - def _get_share(self, shnum): - d = self._s3bucket.get_object("%s%d" % (self._key, shnum)) + def _load_share(self, shnum): + d = self._s3bucket.get_object(self._key + str(shnum)) def _make_share(data): hunk ./src/allmydata/storage/backends/s3/s3_backend.py 101 - if data.startswith(MutableS3Share.MAGIC): - return MutableS3Share(self._storageindex, shnum, self._s3bucket, data=data) + if data.startswith(MUTABLE_MAGIC): + return load_mutable_s3_share(self._s3bucket, self._storageindex, shnum, data=data) else: # assume it's immutable hunk ./src/allmydata/storage/backends/s3/s3_backend.py 105 - return ImmutableS3Share(self._storageindex, shnum, self._s3bucket, data=data) + return load_immutable_s3_share(self._s3bucket, self._storageindex, shnum, data=data) d.addCallback(_make_share) return d hunk ./src/allmydata/storage/backends/s3/s3_backend.py 114 return False def make_bucket_writer(self, storageserver, shnum, max_space_per_bucket, lease_info, canary): - immsh = ImmutableS3Share(self.get_storage_index(), shnum, self._s3bucket, - max_size=max_space_per_bucket) - bw = BucketWriter(storageserver, immsh, lease_info, canary) - return bw + d = create_immutable_s3_share(self._s3bucket, self.get_storage_index(), shnum, + max_size=max_space_per_bucket) + def _created(immsh): + return BucketWriter(storageserver, immsh, lease_info, canary) + d.addCallback(_created) + return d def _create_mutable_share(self, storageserver, shnum, write_enabler): hunk ./src/allmydata/storage/backends/s3/s3_backend.py 122 - # TODO serverid = storageserver.get_serverid() hunk ./src/allmydata/storage/backends/s3/s3_backend.py 123 - return MutableS3Share(self.get_storage_index(), shnum, self._s3bucket, serverid, - write_enabler, storageserver) + return create_mutable_s3_share(self._s3bucket, self.get_storage_index(), shnum, serverid, + write_enabler, storageserver) def _clean_up_after_unlink(self): pass addfile ./src/allmydata/storage/backends/s3/s3_common.py hunk ./src/allmydata/storage/backends/s3/s3_common.py 1 + +import re + +from zope.interface import Interface + +from allmydata.storage.common import si_b2a + + +# The S3 bucket has keys of the form shares/$PREFIX/$STORAGEINDEX/$SHNUM . + +def get_s3_share_key(si, shnum=None): + sistr = si_b2a(si) + if shnum is None: + return "shares/%s/%s/" % (sistr[:2], sistr) + else: + return "shares/%s/%s/%d" % (sistr[:2], sistr, shnum) + +NUM_RE=re.compile("^[0-9]+$") + + +class IS3Bucket(Interface): + """ + I represent an S3 bucket. + """ + def create(self): + """ + Create this bucket. + """ + + def delete(self): + """ + Delete this bucket. + The bucket must be empty before it can be deleted. + """ + + def list_objects(self, prefix=""): + """ + Get a list of all the objects in this bucket whose object names start with + the given prefix. + """ + + def put_object(self, object_name, data, content_type=None, metadata={}): + """ + Put an object in this bucket. + Any existing object of the same name will be replaced. + """ + + def get_object(self, object_name): + """ + Get an object from this bucket. + """ + + def head_object(self, object_name): + """ + Retrieve object metadata only. + """ + + def delete_object(self, object_name): + """ + Delete an object from this bucket. + Once deleted, there is no method to restore or undelete an object. + """ hunk ./src/allmydata/test/no_network.py 361 def find_uri_shares(self, uri): si = tahoe_uri.from_string(uri).get_storage_index() - shares = [] - for i,ss in self.g.servers_by_number.items(): - for share in ss.backend.get_shareset(si).get_shares(): - shares.append((share.get_shnum(), ss.get_serverid(), share._home)) - return sorted(shares) + sharelist = [] + d = defer.succeed(None) + for i, ss in self.g.servers_by_number.items(): + d.addCallback(lambda ign: ss.backend.get_shareset(si).get_shares()) + def _append_shares(shares_for_server): + for share in shares_for_server: + sharelist.append( (share.get_shnum(), ss.get_serverid(), share._home) ) + d.addCallback(_append_shares) + + d.addCallback(lambda ign: sorted(sharelist)) + return d def count_leases(self, uri): """Return (filename, leasecount) pairs in arbitrary order.""" hunk ./src/allmydata/test/no_network.py 377 si = tahoe_uri.from_string(uri).get_storage_index() lease_counts = [] - for i,ss in self.g.servers_by_number.items(): - for share in ss.backend.get_shareset(si).get_shares(): - num_leases = len(list(share.get_leases())) - lease_counts.append( (share._home.path, num_leases) ) - return lease_counts + d = defer.succeed(None) + for i, ss in self.g.servers_by_number.items(): + d.addCallback(lambda ign: ss.backend.get_shareset(si).get_shares()) + def _append_counts(shares_for_server): + for share in shares_for_server: + num_leases = len(list(share.get_leases())) + lease_counts.append( (share._home.path, num_leases) ) + d.addCallback(_append_counts) + + d.addCallback(lambda ign: lease_counts) + return d def copy_shares(self, uri): shares = {} hunk ./src/allmydata/test/no_network.py 391 - for (shnum, serverid, sharefp) in self.find_uri_shares(uri): - shares[sharefp.path] = sharefp.getContent() - return shares + d = self.find_uri_shares(uri) + def _got_shares(sharelist): + for (shnum, serverid, sharefp) in sharelist: + shares[sharefp.path] = sharefp.getContent() + + return shares + d.addCallback(_got_shares) + return d def copy_share(self, from_share, uri, to_server): si = tahoe_uri.from_string(uri).get_storage_index() hunk ./src/allmydata/test/test_backends.py 32 testnodeid = 'testnodeidxxxxxxxxxx' -class MockFileSystem(unittest.TestCase): - """ I simulate a filesystem that the code under test can use. I simulate - just the parts of the filesystem that the current implementation of Disk - backend needs. """ - def setUp(self): - # Make patcher, patch, and effects for disk-using functions. - msg( "%s.setUp()" % (self,)) - self.mockedfilepaths = {} - # keys are pathnames, values are MockFilePath objects. This is necessary because - # MockFilePath behavior sometimes depends on the filesystem. Where it does, - # self.mockedfilepaths has the relevant information. - self.storedir = MockFilePath('teststoredir', self.mockedfilepaths) - self.basedir = self.storedir.child('shares') - self.baseincdir = self.basedir.child('incoming') - self.sharedirfinalname = self.basedir.child('or').child('orsxg5dtorxxeylhmvpws3temv4a') - self.sharedirincomingname = self.baseincdir.child('or').child('orsxg5dtorxxeylhmvpws3temv4a') - self.shareincomingname = self.sharedirincomingname.child('0') - self.sharefinalname = self.sharedirfinalname.child('0') - - # FIXME: these patches won't work; disk_backend no longer imports FilePath, BucketCountingCrawler, - # or LeaseCheckingCrawler. - - self.FilePathFake = mock.patch('allmydata.storage.backends.disk.disk_backend.FilePath', new = MockFilePath) - self.FilePathFake.__enter__() - - self.BCountingCrawler = mock.patch('allmydata.storage.backends.disk.disk_backend.BucketCountingCrawler') - FakeBCC = self.BCountingCrawler.__enter__() - FakeBCC.side_effect = self.call_FakeBCC - - self.LeaseCheckingCrawler = mock.patch('allmydata.storage.backends.disk.disk_backend.LeaseCheckingCrawler') - FakeLCC = self.LeaseCheckingCrawler.__enter__() - FakeLCC.side_effect = self.call_FakeLCC - - self.get_available_space = mock.patch('allmydata.util.fileutil.get_available_space') - GetSpace = self.get_available_space.__enter__() - GetSpace.side_effect = self.call_get_available_space - - self.statforsize = mock.patch('allmydata.storage.backends.disk.core.filepath.stat') - getsize = self.statforsize.__enter__() - getsize.side_effect = self.call_statforsize - - def call_FakeBCC(self, StateFile): - return MockBCC() - - def call_FakeLCC(self, StateFile, HistoryFile, ExpirationPolicy): - return MockLCC() - - def call_get_available_space(self, storedir, reservedspace): - # The input vector has an input size of 85. - return 85 - reservedspace - - def call_statforsize(self, fakefpname): - return self.mockedfilepaths[fakefpname].fileobject.size() - - def tearDown(self): - msg( "%s.tearDown()" % (self,)) - self.FilePathFake.__exit__() - self.mockedfilepaths = {} - - -class MockFilePath: - def __init__(self, pathstring, ffpathsenvironment, existence=False): - # I can't just make the values MockFileObjects because they may be directories. - self.mockedfilepaths = ffpathsenvironment - self.path = pathstring - self.existence = existence - if not self.mockedfilepaths.has_key(self.path): - # The first MockFilePath object is special - self.mockedfilepaths[self.path] = self - self.fileobject = None - else: - self.fileobject = self.mockedfilepaths[self.path].fileobject - self.spawn = {} - self.antecedent = os.path.dirname(self.path) - - def setContent(self, contentstring): - # This method rewrites the data in the file that corresponds to its path - # name whether it preexisted or not. - self.fileobject = MockFileObject(contentstring) - self.existence = True - self.mockedfilepaths[self.path].fileobject = self.fileobject - self.mockedfilepaths[self.path].existence = self.existence - self.setparents() - - def create(self): - # This method chokes if there's a pre-existing file! - if self.mockedfilepaths[self.path].fileobject: - raise OSError - else: - self.existence = True - self.mockedfilepaths[self.path].fileobject = self.fileobject - self.mockedfilepaths[self.path].existence = self.existence - self.setparents() - - def open(self, mode='r'): - # XXX Makes no use of mode. - if not self.mockedfilepaths[self.path].fileobject: - # If there's no fileobject there already then make one and put it there. - self.fileobject = MockFileObject() - self.existence = True - self.mockedfilepaths[self.path].fileobject = self.fileobject - self.mockedfilepaths[self.path].existence = self.existence - else: - # Otherwise get a ref to it. - self.fileobject = self.mockedfilepaths[self.path].fileobject - self.existence = self.mockedfilepaths[self.path].existence - return self.fileobject.open(mode) - - def child(self, childstring): - arg2child = os.path.join(self.path, childstring) - child = MockFilePath(arg2child, self.mockedfilepaths) - return child - - def children(self): - childrenfromffs = [ffp for ffp in self.mockedfilepaths.values() if ffp.path.startswith(self.path)] - childrenfromffs = [ffp for ffp in childrenfromffs if not ffp.path.endswith(self.path)] - childrenfromffs = [ffp for ffp in childrenfromffs if ffp.exists()] - self.spawn = frozenset(childrenfromffs) - return self.spawn - - def parent(self): - if self.mockedfilepaths.has_key(self.antecedent): - parent = self.mockedfilepaths[self.antecedent] - else: - parent = MockFilePath(self.antecedent, self.mockedfilepaths) - return parent - - def parents(self): - antecedents = [] - def f(fps, antecedents): - newfps = os.path.split(fps)[0] - if newfps: - antecedents.append(newfps) - f(newfps, antecedents) - f(self.path, antecedents) - return antecedents - - def setparents(self): - for fps in self.parents(): - if not self.mockedfilepaths.has_key(fps): - self.mockedfilepaths[fps] = MockFilePath(fps, self.mockedfilepaths, exists=True) - - def basename(self): - return os.path.split(self.path)[1] - - def moveTo(self, newffp): - # XXX Makes no distinction between file and directory arguments, this is deviation from filepath.moveTo - if self.mockedfilepaths[newffp.path].exists(): - raise OSError - else: - self.mockedfilepaths[newffp.path] = self - self.path = newffp.path - - def getsize(self): - return self.fileobject.getsize() - - def exists(self): - return self.existence - - def isdir(self): - return True - - def makedirs(self): - # XXX These methods assume that fp_ functions in fileutil will be tested elsewhere! - pass - - def remove(self): - pass - - -class MockFileObject: - def __init__(self, contentstring=''): - self.buffer = contentstring - self.pos = 0 - def open(self, mode='r'): - return self - def write(self, instring): - begin = self.pos - padlen = begin - len(self.buffer) - if padlen > 0: - self.buffer += '\x00' * padlen - end = self.pos + len(instring) - self.buffer = self.buffer[:begin]+instring+self.buffer[end:] - self.pos = end - def close(self): - self.pos = 0 - def seek(self, pos): - self.pos = pos - def read(self, numberbytes): - return self.buffer[self.pos:self.pos+numberbytes] - def tell(self): - return self.pos - def size(self): - # XXX This method A: Is not to be found in a real file B: Is part of a wild-mung-up of filepath.stat! - # XXX Finally we shall hopefully use a getsize method soon, must consult first though. - # Hmmm... perhaps we need to sometimes stat the address when there's not a mockfileobject present? - return {stat.ST_SIZE:len(self.buffer)} - def getsize(self): - return len(self.buffer) - -class MockBCC: - def setServiceParent(self, Parent): - pass - - -class MockLCC: - def setServiceParent(self, Parent): - pass - - class TestServerWithNullBackend(unittest.TestCase, ReallyEqualMixin): """ NullBackend is just for testing and executable documentation, so this test is actually a test of StorageServer in which we're using hunk ./src/allmydata/test/test_storage.py 15 from allmydata.util import fileutil, hashutil, base32, pollmixin, time_format from allmydata.storage.server import StorageServer from allmydata.storage.backends.disk.disk_backend import DiskBackend -from allmydata.storage.backends.disk.immutable import ImmutableDiskShare -from allmydata.storage.backends.disk.mutable import MutableDiskShare +from allmydata.storage.backends.disk.immutable import load_immutable_disk_share, create_immutable_disk_share +from allmydata.storage.backends.disk.mutable import load_mutable_disk_share, MutableDiskShare +from allmydata.storage.backends.s3.s3_backend import S3Backend from allmydata.storage.bucket import BucketWriter, BucketReader from allmydata.storage.common import DataTooLargeError, UnknownContainerVersionError, \ UnknownMutableContainerVersionError, UnknownImmutableContainerVersionError hunk ./src/allmydata/test/test_storage.py 38 from allmydata.test.common import LoggingServiceParent, ShouldFailMixin from allmydata.test.common_web import WebRenderingMixin from allmydata.test.no_network import NoNetworkServer +from allmydata.test.mock_s3 import MockS3Bucket from allmydata.web.storage import StorageStatus, remove_prefix hunk ./src/allmydata/test/test_storage.py 95 def test_create(self): incoming, final = self.make_workdir("test_create") - share = ImmutableDiskShare("", 0, incoming, final, max_size=200) - bw = BucketWriter(self, share, self.make_lease(), FakeCanary()) - bw.remote_write(0, "a"*25) - bw.remote_write(25, "b"*25) - bw.remote_write(50, "c"*25) - bw.remote_write(75, "d"*7) - bw.remote_close() + d = create_immutable_disk_share(incoming, final, max_size=200) + def _got_share(share): + bw = BucketWriter(self, share, self.make_lease(), FakeCanary()) + d2 = defer.succeed(None) + d2.addCallback(lambda ign: bw.remote_write(0, "a"*25)) + d2.addCallback(lambda ign: bw.remote_write(25, "b"*25)) + d2.addCallback(lambda ign: bw.remote_write(50, "c"*25)) + d2.addCallback(lambda ign: bw.remote_write(75, "d"*7)) + d2.addCallback(lambda ign: bw.remote_close()) + return d2 + d.addCallback(_got_share) + return d def test_readwrite(self): incoming, final = self.make_workdir("test_readwrite") hunk ./src/allmydata/test/test_storage.py 110 - share = ImmutableDiskShare("", 0, incoming, final, max_size=200) - bw = BucketWriter(self, share, self.make_lease(), FakeCanary()) - bw.remote_write(0, "a"*25) - bw.remote_write(25, "b"*25) - bw.remote_write(50, "c"*7) # last block may be short - bw.remote_close() + d = create_immutable_disk_share(incoming, final, max_size=200) + def _got_share(share): + bw = BucketWriter(self, share, self.make_lease(), FakeCanary()) + d2 = defer.succeed(None) + d2.addCallback(lambda ign: bw.remote_write(0, "a"*25)) + d2.addCallback(lambda ign: bw.remote_write(25, "b"*25)) + d2.addCallback(lambda ign: bw.remote_write(50, "c"*7)) # last block may be short + d2.addCallback(lambda ign: bw.remote_close()) hunk ./src/allmydata/test/test_storage.py 119 - # now read from it - br = BucketReader(self, share) - self.failUnlessEqual(br.remote_read(0, 25), "a"*25) - self.failUnlessEqual(br.remote_read(25, 25), "b"*25) - self.failUnlessEqual(br.remote_read(50, 7), "c"*7) + # now read from it + def _read(ign): + br = BucketReader(self, share) + d3 = defer.succeed(None) + d3.addCallback(lambda ign: br.remote_read(0, 25)) + d3.addCallback(lambda res: self.failUnlessEqual(res), "a"*25)) + d3.addCallback(lambda ign: br.remote_read(25, 25)) + d3.addCallback(lambda res: self.failUnlessEqual(res), "b"*25)) + d3.addCallback(lambda ign: br.remote_read(50, 7)) + d3.addCallback(lambda res: self.failUnlessEqual(res), "c"*7)) + return d3 + d2.addCallback(_read) + return d2 + d.addCallback(_got_share) + return d def test_read_past_end_of_share_data(self): # test vector for immutable files (hard-coded contents of an immutable share hunk ./src/allmydata/test/test_storage.py 166 incoming, final = self.make_workdir("test_read_past_end_of_share_data") final.setContent(share_file_data) - share = ImmutableDiskShare("", 0, final) + d = load_immutable_disk_share(final) + def _got_share(share): + mockstorageserver = mock.Mock() hunk ./src/allmydata/test/test_storage.py 170 - mockstorageserver = mock.Mock() + # Now read from it. + br = BucketReader(mockstorageserver, share) hunk ./src/allmydata/test/test_storage.py 173 - # Now read from it. - br = BucketReader(mockstorageserver, share) + d2 = br.remote_read(0, len(share_data)) + d2.addCallback(lambda res: self.failUnlessEqual(res, share_data)) hunk ./src/allmydata/test/test_storage.py 176 - self.failUnlessEqual(br.remote_read(0, len(share_data)), share_data) + # Read past the end of share data to get the cancel secret. + read_length = len(share_data) + len(ownernumber) + len(renewsecret) + len(cancelsecret) hunk ./src/allmydata/test/test_storage.py 179 - # Read past the end of share data to get the cancel secret. - read_length = len(share_data) + len(ownernumber) + len(renewsecret) + len(cancelsecret) + d2.addCallback(lambda ign: br.remote_read(0, read_length)) + d2.addCallback(lambda res: self.failUnlessEqual(res, share_data)) hunk ./src/allmydata/test/test_storage.py 182 - result_of_read = br.remote_read(0, read_length) - self.failUnlessEqual(result_of_read, share_data) - - result_of_read = br.remote_read(0, len(share_data)+1) - self.failUnlessEqual(result_of_read, share_data) + d2.addCallback(lambda ign: br.remote_read(0, len(share_data)+1)) + d2.addCallback(lambda res: self.failUnlessEqual(res, share_data)) + return d2 + d.addCallback(_got_share) + return d class RemoteBucket: hunk ./src/allmydata/test/test_storage.py 215 tmpdir.makedirs() incoming = tmpdir.child("bucket") final = basedir.child("bucket") - share = ImmutableDiskShare("", 0, incoming, final, size) - bw = BucketWriter(self, share, self.make_lease(), FakeCanary()) - rb = RemoteBucket() - rb.target = bw - return bw, rb, final + d = create_immutable_disk_share(incoming, final, size) + def _got_share(share): + bw = BucketWriter(self, share, self.make_lease(), FakeCanary()) + rb = RemoteBucket() + rb.target = bw + return bw, rb, final + d.addCallback(_got_share) + return d def make_lease(self): owner_num = 0 hunk ./src/allmydata/test/test_storage.py 240 pass def test_create(self): - bw, rb, sharefp = self.make_bucket("test_create", 500) - bp = WriteBucketProxy(rb, None, - data_size=300, - block_size=10, - num_segments=5, - num_share_hashes=3, - uri_extension_size_max=500) - self.failUnless(interfaces.IStorageBucketWriter.providedBy(bp), bp) + d = self.make_bucket("test_create", 500) + def _made_bucket( (bw, rb, sharefp) ): + bp = WriteBucketProxy(rb, None, + data_size=300, + block_size=10, + num_segments=5, + num_share_hashes=3, + uri_extension_size_max=500) + self.failUnless(interfaces.IStorageBucketWriter.providedBy(bp), bp) + d.addCallback(_made_bucket) + return d def _do_test_readwrite(self, name, header_size, wbp_class, rbp_class): # Let's pretend each share has 100 bytes of data, and that there are hunk ./src/allmydata/test/test_storage.py 274 for i in (1,9,13)] uri_extension = "s" + "E"*498 + "e" - bw, rb, sharefp = self.make_bucket(name, sharesize) - bp = wbp_class(rb, None, - data_size=95, - block_size=25, - num_segments=4, - num_share_hashes=3, - uri_extension_size_max=len(uri_extension)) + d = self.make_bucket(name, sharesize) + def _made_bucket( (bw, rb, sharefp) ): + bp = wbp_class(rb, None, + data_size=95, + block_size=25, + num_segments=4, + num_share_hashes=3, + uri_extension_size_max=len(uri_extension)) + + d2 = bp.put_header() + d2.addCallback(lambda ign: bp.put_block(0, "a"*25)) + d2.addCallback(lambda ign: bp.put_block(1, "b"*25)) + d2.addCallback(lambda ign: bp.put_block(2, "c"*25)) + d2.addCallback(lambda ign: bp.put_block(3, "d"*20)) + d2.addCallback(lambda ign: bp.put_crypttext_hashes(crypttext_hashes)) + d2.addCallback(lambda ign: bp.put_block_hashes(block_hashes)) + d2.addCallback(lambda ign: bp.put_share_hashes(share_hashes)) + d2.addCallback(lambda ign: bp.put_uri_extension(uri_extension)) + d2.addCallback(lambda ign: bp.close()) hunk ./src/allmydata/test/test_storage.py 294 - d = bp.put_header() - d.addCallback(lambda res: bp.put_block(0, "a"*25)) - d.addCallback(lambda res: bp.put_block(1, "b"*25)) - d.addCallback(lambda res: bp.put_block(2, "c"*25)) - d.addCallback(lambda res: bp.put_block(3, "d"*20)) - d.addCallback(lambda res: bp.put_crypttext_hashes(crypttext_hashes)) - d.addCallback(lambda res: bp.put_block_hashes(block_hashes)) - d.addCallback(lambda res: bp.put_share_hashes(share_hashes)) - d.addCallback(lambda res: bp.put_uri_extension(uri_extension)) - d.addCallback(lambda res: bp.close()) + d2.addCallback(lambda ign: load_immutable_disk_share(sharefp)) + return d2 + d.addCallback(_made_bucket) # now read everything back hunk ./src/allmydata/test/test_storage.py 299 - def _start_reading(res): - share = ImmutableDiskShare("", 0, sharefp) + def _start_reading(share): br = BucketReader(self, share) rb = RemoteBucket() rb.target = br hunk ./src/allmydata/test/test_storage.py 308 self.failUnlessIn("to peer", repr(rbp)) self.failUnless(interfaces.IStorageBucketReader.providedBy(rbp), rbp) - d1 = rbp.get_block_data(0, 25, 25) - d1.addCallback(lambda res: self.failUnlessEqual(res, "a"*25)) - d1.addCallback(lambda res: rbp.get_block_data(1, 25, 25)) - d1.addCallback(lambda res: self.failUnlessEqual(res, "b"*25)) - d1.addCallback(lambda res: rbp.get_block_data(2, 25, 25)) - d1.addCallback(lambda res: self.failUnlessEqual(res, "c"*25)) - d1.addCallback(lambda res: rbp.get_block_data(3, 25, 20)) - d1.addCallback(lambda res: self.failUnlessEqual(res, "d"*20)) - - d1.addCallback(lambda res: rbp.get_crypttext_hashes()) - d1.addCallback(lambda res: - self.failUnlessEqual(res, crypttext_hashes)) - d1.addCallback(lambda res: rbp.get_block_hashes(set(range(4)))) - d1.addCallback(lambda res: self.failUnlessEqual(res, block_hashes)) - d1.addCallback(lambda res: rbp.get_share_hashes()) - d1.addCallback(lambda res: self.failUnlessEqual(res, share_hashes)) - d1.addCallback(lambda res: rbp.get_uri_extension()) - d1.addCallback(lambda res: - self.failUnlessEqual(res, uri_extension)) - - return d1 + d2 = defer.succeed(None) + d2.addCallback(lambda ign: rbp.get_block_data(0, 25, 25)) + d2.addCallback(lambda res: self.failUnlessEqual(res, "a"*25)) + d2.addCallback(lambda ign: rbp.get_block_data(1, 25, 25)) + d2.addCallback(lambda res: self.failUnlessEqual(res, "b"*25)) + d2.addCallback(lambda ign: rbp.get_block_data(2, 25, 25)) + d2.addCallback(lambda res: self.failUnlessEqual(res, "c"*25)) + d2.addCallback(lambda ign: rbp.get_block_data(3, 25, 20)) + d2.addCallback(lambda res: self.failUnlessEqual(res, "d"*20)) hunk ./src/allmydata/test/test_storage.py 318 + d2.addCallback(lambda ign: rbp.get_crypttext_hashes()) + d2.addCallback(lambda res: self.failUnlessEqual(res, crypttext_hashes)) + d2.addCallback(lambda ign: rbp.get_block_hashes(set(range(4)))) + d2.addCallback(lambda res: self.failUnlessEqual(res, block_hashes)) + d2.addCallback(lambda ign: rbp.get_share_hashes()) + d2.addCallback(lambda res: self.failUnlessEqual(res, share_hashes)) + d2.addCallback(lambda ign: rbp.get_uri_extension()) + d2.addCallback(lambda res: self.failUnlessEqual(res, uri_extension)) + return d2 d.addCallback(_start_reading) hunk ./src/allmydata/test/test_storage.py 328 - return d def test_readwrite_v1(self): hunk ./src/allmydata/test/test_storage.py 351 def workdir(self, name): return FilePath("storage").child("Server").child(name) - def create(self, name, reserved_space=0, klass=StorageServer): - workdir = self.workdir(name) - backend = DiskBackend(workdir, readonly=False, reserved_space=reserved_space) - ss = klass("\x00" * 20, backend, workdir, - stats_provider=FakeStatsProvider()) - ss.setServiceParent(self.sparent) - return ss - def test_create(self): self.create("test_create") hunk ./src/allmydata/test/test_storage.py 1059 write = ss.remote_slot_testv_and_readv_and_writev read = ss.remote_slot_readv - def reset(): - write("si1", secrets, - {0: ([], [(0,data)], None)}, - []) + def _reset(ign): + return write("si1", secrets, + {0: ([], [(0,data)], None)}, + []) hunk ./src/allmydata/test/test_storage.py 1064 - reset() + d = defer.succeed(None) + d.addCallback(_reset) # lt hunk ./src/allmydata/test/test_storage.py 1068 - answer = write("si1", secrets, {0: ([(10, 5, "lt", "11110"), - ], - [(0, "x"*100)], - None, - )}, [(10,5)]) - self.failUnlessEqual(answer, (False, {0: ["11111"]})) - self.failUnlessEqual(read("si1", [0], [(0,100)]), {0: [data]}) - self.failUnlessEqual(read("si1", [], [(0,100)]), {0: [data]}) - reset() + d.addCallback(lambda ign: write("si1", secrets, {0: ([(10, 5, "lt", "11110"),], + [(0, "x"*100)], + None, + )}, [(10,5)]) + d.addCallback(lambda res: self.failUnlessEqual(res, (False, {0: ["11111"]}))) + d.addCallback(lambda ign: read("si1", [0], [(0,100)])) + d.addCallback(lambda res: self.failUnlessEqual(res, {0: [data]})) + d.addCallback(lambda ign: read("si1", [], [(0,100)])) + d.addCallback(lambda res: self.failUnlessEqual(res, {0: [data]})) + d.addCallback(_reset) answer = write("si1", secrets, {0: ([(10, 5, "lt", "11111"), ], hunk ./src/allmydata/test/test_storage.py 1238 write = ss.remote_slot_testv_and_readv_and_writev read = ss.remote_slot_readv data = [("%d" % i) * 100 for i in range(3)] - rc = write("si1", secrets, - {0: ([], [(0,data[0])], None), - 1: ([], [(0,data[1])], None), - 2: ([], [(0,data[2])], None), - }, []) - self.failUnlessEqual(rc, (True, {})) hunk ./src/allmydata/test/test_storage.py 1239 - answer = read("si1", [], [(0, 10)]) - self.failUnlessEqual(answer, {0: ["0"*10], - 1: ["1"*10], - 2: ["2"*10]}) + d = defer.succeed(None) + d.addCallback(lambda ign: write("si1", secrets, + {0: ([], [(0,data[0])], None), + 1: ([], [(0,data[1])], None), + 2: ([], [(0,data[2])], None), + }, []) + d.addCallback(lambda res: self.failUnlessEqual(res, (True, {}))) + + d.addCallback(lambda ign: read("si1", [], [(0, 10)])) + d.addCallback(lambda res: self.failUnlessEqual(res, {0: ["0"*10], + 1: ["1"*10], + 2: ["2"*10]})) + return d def compare_leases_without_timestamps(self, leases_a, leases_b): self.failUnlessEqual(len(leases_a), len(leases_b)) hunk ./src/allmydata/test/test_storage.py 1291 bucket_dir = ss.backend.get_shareset("si1")._sharehomedir bucket_dir.child("ignore_me.txt").setContent("you ought to be ignoring me\n") - s0 = MutableDiskShare("", 0, bucket_dir.child("0")) - self.failUnlessEqual(len(list(s0.get_leases())), 1) + d = defer.succeed(None) + d.addCallback(lambda ign: load_mutable_disk_share(bucket_dir.child("0"))) + def _got_s0(s0): + self.failUnlessEqual(len(list(s0.get_leases())), 1) hunk ./src/allmydata/test/test_storage.py 1296 - # add-lease on a missing storage index is silently ignored - self.failUnlessEqual(ss.remote_add_lease("si18", "", ""), None) + d2 = defer.succeed(None) + d2.addCallback(lambda ign: ss.remote_add_lease("si18", "", "")) + # add-lease on a missing storage index is silently ignored + d2.addCallback(lambda res: self.failUnlessEqual(res, None)) + + # re-allocate the slots and use the same secrets, that should update + # the lease + d2.addCallback(lambda ign: write("si1", secrets(0), {0: ([], [(0,data)], None)}, [])) + d2.addCallback(lambda ign: self.failUnlessEqual(len(list(s0.get_leases())), 1)) hunk ./src/allmydata/test/test_storage.py 1306 - # re-allocate the slots and use the same secrets, that should update - # the lease - write("si1", secrets(0), {0: ([], [(0,data)], None)}, []) - self.failUnlessEqual(len(list(s0.get_leases())), 1) + # renew it directly + d2.addCallback(lambda ign: ss.remote_renew_lease("si1", secrets(0)[1])) + d2.addCallback(lambda ign: self.failUnlessEqual(len(list(s0.get_leases())), 1)) hunk ./src/allmydata/test/test_storage.py 1310 - # renew it directly - ss.remote_renew_lease("si1", secrets(0)[1]) - self.failUnlessEqual(len(list(s0.get_leases())), 1) + # now allocate them with a bunch of different secrets, to trigger the + # extended lease code. Use add_lease for one of them. + d2.addCallback(lambda ign: write("si1", secrets(1), {0: ([], [(0,data)], None)}, [])) + d2.addCallback(lambda ign: self.failUnlessEqual(len(list(s0.get_leases())), 2)) + secrets2 = secrets(2) + d2.addCallback(lambda ign: ss.remote_add_lease("si1", secrets2[1], secrets2[2])) + d2.addCallback(lambda ign: self.failUnlessEqual(len(list(s0.get_leases())), 3)) + d2.addCallback(lambda ign: write("si1", secrets(3), {0: ([], [(0,data)], None)}, [])) + d2.addCallback(lambda ign: write("si1", secrets(4), {0: ([], [(0,data)], None)}, [])) + d2.addCallback(lambda ign: write("si1", secrets(5), {0: ([], [(0,data)], None)}, [])) hunk ./src/allmydata/test/test_storage.py 1321 - # now allocate them with a bunch of different secrets, to trigger the - # extended lease code. Use add_lease for one of them. - write("si1", secrets(1), {0: ([], [(0,data)], None)}, []) - self.failUnlessEqual(len(list(s0.get_leases())), 2) - secrets2 = secrets(2) - ss.remote_add_lease("si1", secrets2[1], secrets2[2]) - self.failUnlessEqual(len(list(s0.get_leases())), 3) - write("si1", secrets(3), {0: ([], [(0,data)], None)}, []) - write("si1", secrets(4), {0: ([], [(0,data)], None)}, []) - write("si1", secrets(5), {0: ([], [(0,data)], None)}, []) + d2.addCallback(lambda ign: self.failUnlessEqual(len(list(s0.get_leases())), 6)) hunk ./src/allmydata/test/test_storage.py 1323 - self.failUnlessEqual(len(list(s0.get_leases())), 6) + def _check_all_leases(ign): + all_leases = list(s0.get_leases()) hunk ./src/allmydata/test/test_storage.py 1326 - all_leases = list(s0.get_leases()) - # and write enough data to expand the container, forcing the server - # to move the leases - write("si1", secrets(0), - {0: ([], [(0,data)], 200), }, - []) + # and write enough data to expand the container, forcing the server + # to move the leases + d3 = defer.succeed(None) + d3.addCallback(lambda ign: write("si1", secrets(0), + {0: ([], [(0,data)], 200), }, + [])) hunk ./src/allmydata/test/test_storage.py 1333 - # read back the leases, make sure they're still intact. - self.compare_leases_without_timestamps(all_leases, list(s0.get_leases())) + # read back the leases, make sure they're still intact. + d3.addCallback(lambda ign: self.compare_leases_without_timestamps(all_leases, + list(s0.get_leases()))) hunk ./src/allmydata/test/test_storage.py 1337 - ss.remote_renew_lease("si1", secrets(0)[1]) - ss.remote_renew_lease("si1", secrets(1)[1]) - ss.remote_renew_lease("si1", secrets(2)[1]) - ss.remote_renew_lease("si1", secrets(3)[1]) - ss.remote_renew_lease("si1", secrets(4)[1]) - self.compare_leases_without_timestamps(all_leases, list(s0.get_leases())) - # get a new copy of the leases, with the current timestamps. Reading - # data and failing to renew/cancel leases should leave the timestamps - # alone. - all_leases = list(s0.get_leases()) - # renewing with a bogus token should prompt an error message + d3.addCallback(lambda ign: ss.remote_renew_lease("si1", secrets(0)[1])) + d3.addCallback(lambda ign: ss.remote_renew_lease("si1", secrets(1)[1])) + d3.addCallback(lambda ign: ss.remote_renew_lease("si1", secrets(2)[1])) + d3.addCallback(lambda ign: ss.remote_renew_lease("si1", secrets(3)[1])) + d3.addCallback(lambda ign: ss.remote_renew_lease("si1", secrets(4)[1])) + d3.addCallback(lambda ign: self.compare_leases_without_timestamps(all_leases, + list(s0.get_leases()))) + d2.addCallback(_check_all_leases) hunk ./src/allmydata/test/test_storage.py 1346 - # examine the exception thus raised, make sure the old nodeid is - # present, to provide for share migration - e = self.failUnlessRaises(IndexError, - ss.remote_renew_lease, "si1", - secrets(20)[1]) - e_s = str(e) - self.failUnlessIn("Unable to renew non-existent lease", e_s) - self.failUnlessIn("I have leases accepted by nodeids:", e_s) - self.failUnlessIn("nodeids: 'aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa' .", e_s) + def _check_all_leases_again(ign): + # get a new copy of the leases, with the current timestamps. Reading + # data and failing to renew/cancel leases should leave the timestamps + # alone. + all_leases = list(s0.get_leases()) + # renewing with a bogus token should prompt an error message hunk ./src/allmydata/test/test_storage.py 1353 - self.compare_leases(all_leases, list(s0.get_leases())) + # examine the exception thus raised, make sure the old nodeid is + # present, to provide for share migration + d3 = self.shouldFail(IndexError, 'old nodeid present', + "Unable to renew non-existent lease\n" + "I have leases accepted by nodeids:\n" + "nodeids: 'aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa' .", + ss.remote_renew_lease, "si1", secrets(20)[1]) hunk ./src/allmydata/test/test_storage.py 1361 - # reading shares should not modify the timestamp - read("si1", [], [(0,200)]) - self.compare_leases(all_leases, list(s0.get_leases())) + d3.addCallback(lambda ign: self.compare_leases(all_leases, list(s0.get_leases()))) hunk ./src/allmydata/test/test_storage.py 1363 - write("si1", secrets(0), - {0: ([], [(200, "make me bigger")], None)}, []) - self.compare_leases_without_timestamps(all_leases, list(s0.get_leases())) + # reading shares should not modify the timestamp + d3.addCallback(lambda ign: read("si1", [], [(0,200)])) + d3.addCallback(lambda ign: self.compare_leases(all_leases, list(s0.get_leases()))) hunk ./src/allmydata/test/test_storage.py 1367 - write("si1", secrets(0), - {0: ([], [(500, "make me really bigger")], None)}, []) - self.compare_leases_without_timestamps(all_leases, list(s0.get_leases())) + d3.addCallback(lambda ign: write("si1", secrets(0), + {0: ([], [(200, "make me bigger")], None)}, [])) + d3.addCallback(lambda ign: self.compare_leases_without_timestamps(all_leases, list(s0.get_leases()))) + + d3.addCallback(lambda ign: write("si1", secrets(0), + {0: ([], [(500, "make me really bigger")], None)}, [])) + d3.addCallback(lambda ign: self.compare_leases_without_timestamps(all_leases, list(s0.get_leases()))) + d2.addCallback(_check_all_leases_again) + return d2 + d.addCallback(_got_s0) + return d def test_remove(self): ss = self.create("test_remove") hunk ./src/allmydata/test/test_storage.py 1381 - self.allocate(ss, "si1", "we1", self._lease_secret.next(), - set([0,1,2]), 100) readv = ss.remote_slot_readv writev = ss.remote_slot_testv_and_readv_and_writev secrets = ( self.write_enabler("we1"), hunk ./src/allmydata/test/test_storage.py 1386 self.renew_secret("we1"), self.cancel_secret("we1") ) + + d = defer.succeed(None) + d.addCallback(lambda ign: self.allocate(ss, "si1", "we1", self._lease_secret.next(), + set([0,1,2]), 100) # delete sh0 by setting its size to zero hunk ./src/allmydata/test/test_storage.py 1391 - answer = writev("si1", secrets, - {0: ([], [], 0)}, - []) + d.addCallback(lambda ign: writev("si1", secrets, + {0: ([], [], 0)}, + [])) # the answer should mention all the shares that existed before the # write hunk ./src/allmydata/test/test_storage.py 1396 - self.failUnlessEqual(answer, (True, {0:[],1:[],2:[]}) ) + d.addCallback(lambda answer: self.failUnlessEqual(answer, (True, {0:[],1:[],2:[]}) )) # but a new read should show only sh1 and sh2 hunk ./src/allmydata/test/test_storage.py 1398 - self.failUnlessEqual(readv("si1", [], [(0,10)]), - {1: [""], 2: [""]}) + d.addCallback(lambda ign: readv("si1", [], [(0,10)])) + d.addCallback(lambda answer: self.failUnlessEqual(answer, {1: [""], 2: [""]})) # delete sh1 by setting its size to zero hunk ./src/allmydata/test/test_storage.py 1402 - answer = writev("si1", secrets, - {1: ([], [], 0)}, - []) - self.failUnlessEqual(answer, (True, {1:[],2:[]}) ) - self.failUnlessEqual(readv("si1", [], [(0,10)]), - {2: [""]}) + d.addCallback(lambda ign: writev("si1", secrets, + {1: ([], [], 0)}, + [])) + d.addCallback(lambda answer: self.failUnlessEqual(answer, (True, {1:[],2:[]}) )) + d.addCallback(lambda ign: readv("si1", [], [(0,10)])) + d.addCallback(lambda answer: self.failUnlessEqual(answer, {2: [""]})) # delete sh2 by setting its size to zero hunk ./src/allmydata/test/test_storage.py 1410 - answer = writev("si1", secrets, - {2: ([], [], 0)}, - []) - self.failUnlessEqual(answer, (True, {2:[]}) ) - self.failUnlessEqual(readv("si1", [], [(0,10)]), - {}) + d.addCallback(lambda ign: writev("si1", secrets, + {2: ([], [], 0)}, + [])) + d.addCallback(lambda answer: self.failUnlessEqual(answer, (True, {2:[]}) )) + d.addCallback(lambda ign: readv("si1", [], [(0,10)])) + d.addCallback(lambda answer: self.failUnlessEqual(answer, {})) # and the bucket directory should now be gone hunk ./src/allmydata/test/test_storage.py 1417 - si = base32.b2a("si1") - # note: this is a detail of the storage server implementation, and - # may change in the future - prefix = si[:2] - prefixdir = self.workdir("test_remove").child("shares").child(prefix) - bucketdir = prefixdir.child(si) - self.failUnless(prefixdir.exists(), prefixdir) - self.failIf(bucketdir.exists(), bucketdir) + def _check_gone(ign): + si = base32.b2a("si1") + # note: this is a detail of the storage server implementation, and + # may change in the future + prefix = si[:2] + prefixdir = self.workdir("test_remove").child("shares").child(prefix) + bucketdir = prefixdir.child(si) + self.failUnless(prefixdir.exists(), prefixdir) + self.failIf(bucketdir.exists(), bucketdir) + d.addCallback(_check_gone) + return d + + +class ServerWithS3Backend(Server): + def create(self, name, reserved_space=0, klass=StorageServer): + workdir = self.workdir(name) + s3bucket = MockS3Bucket(workdir) + backend = S3Backend(s3bucket, readonly=False, reserved_space=reserved_space) + ss = klass("\x00" * 20, backend, workdir, + stats_provider=FakeStatsProvider()) + ss.setServiceParent(self.sparent) + return ss + + +class ServerWithDiskBackend(Server): + def create(self, name, reserved_space=0, klass=StorageServer): + workdir = self.workdir(name) + backend = DiskBackend(workdir, readonly=False, reserved_space=reserved_space) + ss = klass("\x00" * 20, backend, workdir, + stats_provider=FakeStatsProvider()) + ss.setServiceParent(self.sparent) + return ss class MDMFProxies(unittest.TestCase, ShouldFailMixin): hunk ./src/allmydata/test/test_storage.py 4028 f.write("BAD MAGIC") finally: f.close() - # if get_share_file() doesn't see the correct mutable magic, it - # assumes the file is an immutable share, and then - # immutable.ShareFile sees a bad version. So regardless of which kind + + # If the backend doesn't see the correct mutable magic, it + # assumes the file is an immutable share, and then the immutable + # share class will see a bad version. So regardless of which kind # of share we corrupted, this will trigger an # UnknownImmutableContainerVersionError. hunk ./src/allmydata/test/test_system.py 11 import allmydata from allmydata import uri -from allmydata.storage.backends.disk.mutable import MutableDiskShare +from allmydata.storage.backends.disk.mutable import load_mutable_disk_share from allmydata.storage.server import si_a2b from allmydata.immutable import offloaded, upload from allmydata.immutable.literal import LiteralFileNode hunk ./src/allmydata/test/test_system.py 421 self.fail("unable to find any share files in %s" % basedir) return shares - def _corrupt_mutable_share(self, what, which): + def _corrupt_mutable_share(self, ign, what, which): (storageindex, filename, shnum) = what hunk ./src/allmydata/test/test_system.py 423 - msf = MutableDiskShare(storageindex, shnum, FilePath(filename)) - datav = msf.readv([ (0, 1000000) ]) - final_share = datav[0] - assert len(final_share) < 1000000 # ought to be truncated - pieces = mutable_layout.unpack_share(final_share) - (seqnum, root_hash, IV, k, N, segsize, datalen, - verification_key, signature, share_hash_chain, block_hash_tree, - share_data, enc_privkey) = pieces + d = load_mutable_disk_share(FilePath(filename), storageindex, shnum) + def _got_share(msf): + d2 = msf.readv([ (0, 1000000) ]) + def _got_data(datav): + final_share = datav[0] + assert len(final_share) < 1000000 # ought to be truncated + pieces = mutable_layout.unpack_share(final_share) + (seqnum, root_hash, IV, k, N, segsize, datalen, + verification_key, signature, share_hash_chain, block_hash_tree, + share_data, enc_privkey) = pieces hunk ./src/allmydata/test/test_system.py 434 - if which == "seqnum": - seqnum = seqnum + 15 - elif which == "R": - root_hash = self.flip_bit(root_hash) - elif which == "IV": - IV = self.flip_bit(IV) - elif which == "segsize": - segsize = segsize + 15 - elif which == "pubkey": - verification_key = self.flip_bit(verification_key) - elif which == "signature": - signature = self.flip_bit(signature) - elif which == "share_hash_chain": - nodenum = share_hash_chain.keys()[0] - share_hash_chain[nodenum] = self.flip_bit(share_hash_chain[nodenum]) - elif which == "block_hash_tree": - block_hash_tree[-1] = self.flip_bit(block_hash_tree[-1]) - elif which == "share_data": - share_data = self.flip_bit(share_data) - elif which == "encprivkey": - enc_privkey = self.flip_bit(enc_privkey) + if which == "seqnum": + seqnum = seqnum + 15 + elif which == "R": + root_hash = self.flip_bit(root_hash) + elif which == "IV": + IV = self.flip_bit(IV) + elif which == "segsize": + segsize = segsize + 15 + elif which == "pubkey": + verification_key = self.flip_bit(verification_key) + elif which == "signature": + signature = self.flip_bit(signature) + elif which == "share_hash_chain": + nodenum = share_hash_chain.keys()[0] + share_hash_chain[nodenum] = self.flip_bit(share_hash_chain[nodenum]) + elif which == "block_hash_tree": + block_hash_tree[-1] = self.flip_bit(block_hash_tree[-1]) + elif which == "share_data": + share_data = self.flip_bit(share_data) + elif which == "encprivkey": + enc_privkey = self.flip_bit(enc_privkey) hunk ./src/allmydata/test/test_system.py 456 - prefix = mutable_layout.pack_prefix(seqnum, root_hash, IV, k, N, - segsize, datalen) - final_share = mutable_layout.pack_share(prefix, - verification_key, - signature, - share_hash_chain, - block_hash_tree, - share_data, - enc_privkey) - msf.writev( [(0, final_share)], None) + prefix = mutable_layout.pack_prefix(seqnum, root_hash, IV, k, N, + segsize, datalen) + final_share = mutable_layout.pack_share(prefix, + verification_key, + signature, + share_hash_chain, + block_hash_tree, + share_data, + enc_privkey) hunk ./src/allmydata/test/test_system.py 466 + return msf.writev( [(0, final_share)], None) + d2.addCallback(_got_data) + return d2 + d.addCallback(_got_share) + return d def test_mutable(self): self.basedir = "system/SystemTest/test_mutable" hunk ./src/allmydata/test/test_system.py 606 for (client_num, storageindex, filename, shnum) in shares ]) assert len(where) == 10 # this test is designed for 3-of-10 + + d2 = defer.succeed(None) for shnum, what in where.items(): # shares 7,8,9 are left alone. read will check # (share_hash_chain, block_hash_tree, share_data). New hunk ./src/allmydata/test/test_system.py 616 if shnum == 0: # read: this will trigger "pubkey doesn't match # fingerprint". - self._corrupt_mutable_share(what, "pubkey") - self._corrupt_mutable_share(what, "encprivkey") + d2.addCallback(self._corrupt_mutable_share, what, "pubkey") + d2.addCallback(self._corrupt_mutable_share, what, "encprivkey") elif shnum == 1: # triggers "signature is invalid" hunk ./src/allmydata/test/test_system.py 620 - self._corrupt_mutable_share(what, "seqnum") + d2.addCallback(self._corrupt_mutable_share, what, "seqnum") elif shnum == 2: # triggers "signature is invalid" hunk ./src/allmydata/test/test_system.py 623 - self._corrupt_mutable_share(what, "R") + d2.addCallback(self._corrupt_mutable_share, what, "R") elif shnum == 3: # triggers "signature is invalid" hunk ./src/allmydata/test/test_system.py 626 - self._corrupt_mutable_share(what, "segsize") + d2.addCallback(self._corrupt_mutable_share, what, "segsize") elif shnum == 4: hunk ./src/allmydata/test/test_system.py 628 - self._corrupt_mutable_share(what, "share_hash_chain") + d2.addCallback(self._corrupt_mutable_share, what, "share_hash_chain") elif shnum == 5: hunk ./src/allmydata/test/test_system.py 630 - self._corrupt_mutable_share(what, "block_hash_tree") + d2.addCallback(self._corrupt_mutable_share, what, "block_hash_tree") elif shnum == 6: hunk ./src/allmydata/test/test_system.py 632 - self._corrupt_mutable_share(what, "share_data") + d2.addCallback(self._corrupt_mutable_share, what, "share_data") # other things to correct: IV, signature # 7,8,9 are left alone hunk ./src/allmydata/test/test_system.py 648 # for one failure mode at a time. # when we retrieve this, we should get three signature - # failures (where we've mangled seqnum, R, and segsize). The - # pubkey mangling + # failures (where we've mangled seqnum, R, and segsize). + return d2 d.addCallback(_corrupt_shares) d.addCallback(lambda res: self._newnode3.download_best_version()) } Context: [test/test_runner.py: BinTahoe.test_path has rare nondeterministic failures; this patch probably fixes a problem where the actual cause of failure is masked by a string conversion error. david-sarah@jacaranda.org**20110927225336 Ignore-this: 6f1ad68004194cc9cea55ace3745e4af ] [docs/configuration.rst: add section about the types of node, and clarify when setting web.port enables web-API service. fixes #1444 zooko@zooko.com**20110926203801 Ignore-this: ab94d470c68e720101a7ff3c207a719e ] [TAG allmydata-tahoe-1.9.0a2 warner@lothar.com**20110925234811 Ignore-this: e9649c58f9c9017a7d55008938dba64f ] [NEWS: tidy up a little bit, reprioritize some items, hide some non-user-visible items warner@lothar.com**20110925233529 Ignore-this: 61f334cc3fa2539742c3e5d2801aee81 ] [docs: fix some broken .rst links. refs #1542 david-sarah@jacaranda.org**20110925051001 Ignore-this: 5714ee650abfcaab0914537e1f206972 ] [mutable/publish.py: fix an unused import. refs #1542 david-sarah@jacaranda.org**20110925052206 Ignore-this: 2d69ac9e605e789c0aedfecb8877b7d7 ] [NEWS: fix .rst formatting. david-sarah@jacaranda.org**20110925050119 Ignore-this: aa1d20acd23bdb8f8f6d0fa048ea0277 ] [NEWS: updates for 1.9alpha2. david-sarah@jacaranda.org**20110925045343 Ignore-this: d2c44e4e05d2ed662b7adfd2e43928bc ] [mutable/layout.py: make unpack_sdmf_checkstring and unpack_mdmf_checkstring more similar, and change an assert to give a more useful message if it fails. refs #1540 david-sarah@jacaranda.org**20110925023651 Ignore-this: 977aaa8cb16e06a6dcc3e27cb6e23956 ] [mutable/publish: handle unknown mutable share formats when handling errors kevan@isnotajoke.com**20110925004305 Ignore-this: 4d5fa44ef7d777c432eb10c9584ad51f ] [mutable/layout: break unpack_checkstring into unpack_mdmf_checkstring and unpack_sdmf_checkstring, add distinguisher function for checkstrings kevan@isnotajoke.com**20110925004134 Ignore-this: 57f49ed5a72e418a69c7286a225cc8fb ] [test/test_mutable: reenable mdmf publish surprise test kevan@isnotajoke.com**20110924235415 Ignore-this: f752e47a703684491305cc83d16248fb ] [mutable/publish: use unpack_mdmf_checkstring and unpack_sdmf_checkstring instead of unpack_checkstring. fixes #1540 kevan@isnotajoke.com**20110924235137 Ignore-this: 52ca3d9627b8b0ba758367b2bd6c7085 ] [mutable/publish.py: copy the self.writers dict before iterating over it, since we remove elements from it during the iteration. refs #393 david-sarah@jacaranda.org**20110924211208 Ignore-this: 76d4066b55d50ace2a34b87443b39094 ] [mutable/publish.py: simplify by refactoring self.outstanding to self.num_outstanding. refs #393 david-sarah@jacaranda.org**20110924205004 Ignore-this: 902768cfc529ae13ae0b7f67768a3643 ] [test_mutable.py: update SkipTest message for test_publish_surprise_mdmf to reference the right ticket number. refs #1540. david-sarah@jacaranda.org**20110923211622 Ignore-this: 44f16a6817a6b75930bbba18b0a516be ] [control.py: unbreak speed-test: overwrite() wants a MutableData, not str Brian Warner **20110923073748 Ignore-this: 7dad7aff3d66165868a64ae22d225fa3 Really, all the upload/modify APIs should take a string or a filehandle, and internally wrap it as needed. Callers should not need to be aware of Uploadable() or MutableData() classes. ] [test_mutable.py: skip test_publish_surprise_mdmf, which is causing an error. refs #1534, #393 david-sarah@jacaranda.org**20110920183319 Ignore-this: 6fb020e09e8de437cbcc2c9f57835b31 ] [test/test_mutable: write publish surprise test for MDMF, rename existing test_publish_surprise to clarify that it is for SDMF kevan@isnotajoke.com**20110918003657 Ignore-this: 722c507e8f5b537ff920e0555951059a ] [test/test_mutable: refactor publish surprise test into common test fixture, rewrite test_publish_surprise to use test fixture kevan@isnotajoke.com**20110918003533 Ignore-this: 6f135888d400a99a09b5f9a4be443b6e ] [mutable/publish: add errback immediately after write, don't consume errors from other parts of the publisher kevan@isnotajoke.com**20110917234708 Ignore-this: 12bf6b0918a5dc5ffc30ece669fad51d ] [.darcs-boringfile: minor cleanups. david-sarah@jacaranda.org**20110920154918 Ignore-this: cab78e30d293da7e2832207dbee2ffeb ] [uri.py: fix two interface violations in verifier URI classes. refs #1474 david-sarah@jacaranda.org**20110920030156 Ignore-this: 454ddd1419556cb1d7576d914cb19598 ] [misc/coding_tools/check_interfaces.py: report all violations rather than only one for a given class, by including a forked version of verifyClass. refs #1474 david-sarah@jacaranda.org**20110916223450 Ignore-this: 927efeecf4d12588316826a4b3479aa9 ] [misc/coding_tools/check_interfaces.py: use os.walk instead of FilePath, since this script shouldn't really depend on Twisted. refs #1474 david-sarah@jacaranda.org**20110916212633 Ignore-this: 46eeb4236b34375227dac71ef53f5428 ] [misc/coding_tools/check-interfaces.py: reduce false-positives by adding Dummy* to the set of excluded classnames, and bench-* to the set of excluded basenames. refs #1474 david-sarah@jacaranda.org**20110916212624 Ignore-this: 4e78f6e6fe6c0e9be9df826a0e206804 ] [Add a script 'misc/coding_tools/check-interfaces.py' that checks whether zope interfaces are enforced. Also add 'check-interfaces', 'version-and-path', and 'code-checks' targets to the Makefile. fixes #1474 david-sarah@jacaranda.org**20110915161532 Ignore-this: 32d9bdc5bc4a86d21e927724560ad4b4 ] [Make platform-detection code tolerate linux-3.0, patch by zooko. Brian Warner **20110915202620 Ignore-this: af63cf9177ae531984dea7a1cad03762 Otherwise address-autodetection can't find ifconfig. refs #1536 ] [test_web.py: fix a bug in _count_leases that was causing us to check only the lease count of one share file, not of all share files as intended. david-sarah@jacaranda.org**20110915185126 Ignore-this: d96632bc48d770b9b577cda1bbd8ff94 ] [docs: insert a newline at the beginning of known_issues.rst to see if this makes it render more nicely in trac zooko@zooko.com**20110914064728 Ignore-this: aca15190fa22083c5d4114d3965f5d65 ] [docs: remove the coding: utf-8 declaration at the to of known_issues.rst, since the trac rendering doesn't hide it zooko@zooko.com**20110914055713 Ignore-this: 941ed32f83ead377171aa7a6bd198fcf ] [docs: more cleanup of known_issues.rst -- now it passes "rst2html --verbose" without comment zooko@zooko.com**20110914055419 Ignore-this: 5505b3d76934bd97d0312cc59ed53879 ] [docs: more formatting improvements to known_issues.rst zooko@zooko.com**20110914051639 Ignore-this: 9ae9230ec9a38a312cbacaf370826691 ] [docs: reformatting of known_issues.rst zooko@zooko.com**20110914050240 Ignore-this: b8be0375079fb478be9d07500f9aaa87 ] [docs: fix formatting error in docs/known_issues.rst zooko@zooko.com**20110914045909 Ignore-this: f73fe74ad2b9e655aa0c6075acced15a ] [merge Tahoe-LAFS v1.8.3 release announcement with trunk zooko@zooko.com**20110913210544 Ignore-this: 163f2c3ddacca387d7308e4b9332516e ] [docs: release notes for Tahoe-LAFS v1.8.3 zooko@zooko.com**20110913165826 Ignore-this: 84223604985b14733a956d2fbaeb4e9f ] [tests: bump up the timeout in this test that fails on FreeStorm's CentOS in order to see if it is just very slow zooko@zooko.com**20110913024255 Ignore-this: 6a86d691e878cec583722faad06fb8e4 ] [interfaces: document that the 'fills-holes-with-zero-bytes' key should be used to detect whether a storage server has that behavior. refs #1528 david-sarah@jacaranda.org**20110913002843 Ignore-this: 1a00a6029d40f6792af48c5578c1fd69 ] [CREDITS: more CREDITS for Kevan and David-Sarah zooko@zooko.com**20110912223357 Ignore-this: 4ea8f0d6f2918171d2f5359c25ad1ada ] [merge NEWS about the mutable file bounds fixes with NEWS about work-in-progress zooko@zooko.com**20110913205521 Ignore-this: 4289a4225f848d6ae6860dd39bc92fa8 ] [doc: add NEWS item about fixes to potential palimpsest issues in mutable files zooko@zooko.com**20110912223329 Ignore-this: 9d63c95ddf95c7d5453c94a1ba4d406a ref. #1528 ] [merge the NEWS about the security fix (#1528) with the work-in-progress NEWS zooko@zooko.com**20110913205153 Ignore-this: 88e88a2ad140238c62010cf7c66953fc ] [doc: add NEWS entry about the issue which allows unauthorized deletion of shares zooko@zooko.com**20110912223246 Ignore-this: 77e06d09103d2ef6bb51ea3e5d6e80b0 ref. #1528 ] [doc: add entry in known_issues.rst about the issue which allows unauthorized deletion of shares zooko@zooko.com**20110912223135 Ignore-this: b26c6ea96b6c8740b93da1f602b5a4cd ref. #1528 ] [storage: more paranoid handling of bounds and palimpsests in mutable share files zooko@zooko.com**20110912222655 Ignore-this: a20782fa423779ee851ea086901e1507 * storage server ignores requests to extend shares by sending a new_length * storage server fills exposed holes (created by sending a write vector whose offset begins after the end of the current data) with 0 to avoid "palimpsest" exposure of previous contents * storage server zeroes out lease info at the old location when moving it to a new location ref. #1528 ] [storage: test that the storage server ignores requests to extend shares by sending a new_length, and that the storage server fills exposed holes with 0 to avoid "palimpsest" exposure of previous contents zooko@zooko.com**20110912222554 Ignore-this: 61ebd7b11250963efdf5b1734a35271 ref. #1528 ] [immutable: prevent clients from reading past the end of share data, which would allow them to learn the cancellation secret zooko@zooko.com**20110912222458 Ignore-this: da1ebd31433ea052087b75b2e3480c25 Declare explicitly that we prevent this problem in the server's version dict. fixes #1528 (there are two patches that are each a sufficient fix to #1528 and this is one of them) ] [storage: remove the storage server's "remote_cancel_lease" function zooko@zooko.com**20110912222331 Ignore-this: 1c32dee50e0981408576daffad648c50 We're removing this function because it is currently unused, because it is dangerous, and because the bug described in #1528 leaks the cancellation secret, which allows anyone who knows a file's storage index to abuse this function to delete shares of that file. fixes #1528 (there are two patches that are each a sufficient fix to #1528 and this is one of them) ] [storage: test that the storage server does *not* have a "remote_cancel_lease" function zooko@zooko.com**20110912222324 Ignore-this: 21c652009704652d35f34651f98dd403 We're removing this function because it is currently unused, because it is dangerous, and because the bug described in #1528 leaks the cancellation secret, which allows anyone who knows a file's storage index to abuse this function to delete shares of that file. ref. #1528 ] [immutable: test whether the server allows clients to read past the end of share data, which would allow them to learn the cancellation secret zooko@zooko.com**20110912221201 Ignore-this: 376e47b346c713d37096531491176349 Also test whether the server explicitly declares that it prevents this problem. ref #1528 ] [Retrieve._activate_enough_peers: rewrite Verify logic Brian Warner **20110909181150 Ignore-this: 9367c11e1eacbf025f75ce034030d717 ] [Retrieve: implement/test stopProducing Brian Warner **20110909181150 Ignore-this: 47b2c3df7dc69835e0a066ca12e3c178 ] [move DownloadStopped from download.common to interfaces Brian Warner **20110909181150 Ignore-this: 8572acd3bb16e50341dbed8eb1d90a50 ] [retrieve.py: remove vestigal self._validated_readers Brian Warner **20110909181150 Ignore-this: faab2ec14e314a53a2ffb714de626e2d ] [Retrieve: rewrite flow-control: use a top-level loop() to catch all errors Brian Warner **20110909181150 Ignore-this: e162d2cd53b3d3144fc6bc757e2c7714 This ought to close the potential for dropped errors and hanging downloads. Verify needs to be examined, I may have broken it, although all tests pass. ] [Retrieve: merge _validate_active_prefixes into _add_active_peers Brian Warner **20110909181150 Ignore-this: d3ead31e17e69394ae7058eeb5beaf4c ] [Retrieve: remove the initial prefix-is-still-good check Brian Warner **20110909181150 Ignore-this: da66ee51c894eaa4e862e2dffb458acc This check needs to be done with each fetch from the storage server, to detect when someone has changed the share (i.e. our servermap goes stale). Doing it just once at the beginning of retrieve isn't enough: a write might occur after the first segment but before the second, etc. _try_to_validate_prefix() was not removed: it will be used by the future check-with-each-fetch code. test_mutable.Roundtrip.test_corrupt_all_seqnum_late was disabled, since it fails until this check is brought back. (the corruption it applies only touches the prefix, not the block data, so the check-less retrieve actually tolerates it). Don't forget to re-enable it once the check is brought back. ] [MDMFSlotReadProxy: remove the queue Brian Warner **20110909181150 Ignore-this: 96673cb8dda7a87a423de2f4897d66d2 This is a neat trick to reduce Foolscap overhead, but the need for an explicit flush() complicates the Retrieve path and makes it prone to lost-progress bugs. Also change test_mutable.FakeStorageServer to tolerate multiple reads of the same share in a row, a limitation exposed by turning off the queue. ] [rearrange Retrieve: first step, shouldn't change order of execution Brian Warner **20110909181149 Ignore-this: e3006368bfd2802b82ea45c52409e8d6 ] [CLI: test_cli.py -- remove an unnecessary call in test_mkdir_mutable_type. refs #1527 david-sarah@jacaranda.org**20110906183730 Ignore-this: 122e2ffbee84861c32eda766a57759cf ] [CLI: improve test for 'tahoe mkdir --mutable-type='. refs #1527 david-sarah@jacaranda.org**20110906183020 Ignore-this: f1d4598e6c536f0a2b15050b3bc0ef9d ] [CLI: make the --mutable-type option value for 'tahoe put' and 'tahoe mkdir' case-insensitive, and change --help for these commands accordingly. fixes #1527 david-sarah@jacaranda.org**20110905020922 Ignore-this: 75a6df0a2df9c467d8c010579e9a024e ] [cli: make --mutable-type imply --mutable in 'tahoe put' Kevan Carstensen **20110903190920 Ignore-this: 23336d3c43b2a9554e40c2a11c675e93 ] [SFTP: add a comment about a subtle interaction between OverwriteableFileConsumer and GeneralSFTPFile, and test the case it is commenting on. david-sarah@jacaranda.org**20110903222304 Ignore-this: 980c61d4dd0119337f1463a69aeebaf0 ] [improve the storage/mutable.py asserts even more warner@lothar.com**20110901160543 Ignore-this: 5b2b13c49bc4034f96e6e3aaaa9a9946 ] [storage/mutable.py: special characters in struct.foo arguments indicate standard as opposed to native sizes, we should be using these characters in these asserts wilcoxjg@gmail.com**20110901084144 Ignore-this: 28ace2b2678642e4d7269ddab8c67f30 ] [docs/write_coordination.rst: fix formatting and add more specific warning about access via sshfs. david-sarah@jacaranda.org**20110831232148 Ignore-this: cd9c851d3eb4e0a1e088f337c291586c ] [test_mutable.Version: consolidate some tests, reduce runtime from 19s to 15s warner@lothar.com**20110831050451 Ignore-this: 64815284d9e536f8f3798b5f44cf580c ] [mutable/retrieve: handle the case where self._read_length is 0. Kevan Carstensen **20110830210141 Ignore-this: fceafbe485851ca53f2774e5a4fd8d30 Note that the downloader will still fetch a segment for a zero-length read, which is wasteful. Fixing that isn't specifically required to fix #1512, but it should probably be fixed before 1.9. ] [NEWS: added summary of all changes since 1.8.2. Needs editing. Brian Warner **20110830163205 Ignore-this: 273899b37a899fc6919b74572454b8b2 ] [test_mutable.Update: only upload the files needed for each test. refs #1500 Brian Warner **20110829072717 Ignore-this: 4d2ab4c7523af9054af7ecca9c3d9dc7 This first step shaves 15% off the runtime: from 139s to 119s on my laptop. It also fixes a couple of places where a Deferred was being dropped, which would cause two tests to run in parallel and also confuse error reporting. ] [Let Uploader retain History instead of passing it into upload(). Fixes #1079. Brian Warner **20110829063246 Ignore-this: 3902c58ec12bd4b2d876806248e19f17 This consistently records all immutable uploads in the Recent Uploads And Downloads page, regardless of code path. Previously, certain webapi upload operations (like PUT /uri/$DIRCAP/newchildname) failed to pass the History object and were left out. ] [Fix mutable publish/retrieve timing status displays. Fixes #1505. Brian Warner **20110828232221 Ignore-this: 4080ce065cf481b2180fd711c9772dd6 publish: * encrypt and encode times are cumulative, not just current-segment retrieve: * same for decrypt and decode times * update "current status" to include segment number * set status to Finished/Failed when download is complete * set progress to 1.0 when complete More improvements to consider: * progress is currently 0% or 100%: should calculate how many segments are involved (remembering retrieve can be less than the whole file) and set it to a fraction * "fetch" time is fuzzy: what we want is to know how much of the delay is not our own fault, but since we do decode/decrypt work while waiting for more shares, it's not straightforward ] [Teach 'tahoe debug catalog-shares about MDMF. Closes #1507. Brian Warner **20110828080931 Ignore-this: 56ef2951db1a648353d7daac6a04c7d1 ] [debug.py: remove some dead comments Brian Warner **20110828074556 Ignore-this: 40e74040dd4d14fd2f4e4baaae506b31 ] [hush pyflakes Brian Warner **20110828074254 Ignore-this: bef9d537a969fa82fe4decc4ba2acb09 ] [MutableFileNode.set_downloader_hints: never depend upon order of dict.values() Brian Warner **20110828074103 Ignore-this: caaf1aa518dbdde4d797b7f335230faa The old code was calculating the "extension parameters" (a list) from the downloader hints (a dictionary) with hints.values(), which is not stable, and would result in corrupted filecaps (with the 'k' and 'segsize' hints occasionally swapped). The new code always uses [k,segsize]. ] [layout.py: fix MDMF share layout documentation Brian Warner **20110828073921 Ignore-this: 3f13366fed75b5e31b51ae895450a225 ] [teach 'tahoe debug dump-share' about MDMF and offsets. refs #1507 Brian Warner **20110828073834 Ignore-this: 3a9d2ef9c47a72bf1506ba41199a1dea ] [test_mutable.Version.test_debug: use splitlines() to fix buildslaves Brian Warner **20110828064728 Ignore-this: c7f6245426fc80b9d1ae901d5218246a Any slave running in a directory with spaces in the name was miscounting shares, causing the test to fail. ] [test_mutable.Version: exercise 'tahoe debug find-shares' on MDMF. refs #1507 Brian Warner **20110828005542 Ignore-this: cb20bea1c28bfa50a72317d70e109672 Also changes NoNetworkGrid to put shares in storage/shares/ . ] [test_mutable.py: oops, missed a .todo Brian Warner **20110828002118 Ignore-this: fda09ae86481352b7a627c278d2a3940 ] [test_mutable: merge davidsarah's patch with my Version refactorings warner@lothar.com**20110827235707 Ignore-this: b5aaf481c90d99e33827273b5d118fd0 ] [Make the immutable/read-only constraint checking for MDMF URIs identical to that for SSK URIs. refs #393 david-sarah@jacaranda.org**20110823012720 Ignore-this: e1f59d7ff2007c81dbef2aeb14abd721 ] [Additional tests for MDMF URIs and for zero-length files. refs #393 david-sarah@jacaranda.org**20110823011532 Ignore-this: a7cc0c09d1d2d72413f9cd227c47a9d5 ] [Additional tests for zero-length partial reads and updates to mutable versions. refs #393 david-sarah@jacaranda.org**20110822014111 Ignore-this: 5fc6f4d06e11910124e4a277ec8a43ea ] [test_mutable.Version: factor out some expensive uploads, save 25% runtime Brian Warner **20110827232737 Ignore-this: ea37383eb85ea0894b254fe4dfb45544 ] [SDMF: update filenode with correct k/N after Retrieve. Fixes #1510. Brian Warner **20110827225031 Ignore-this: b50ae6e1045818c400079f118b4ef48 Without this, we get a regression when modifying a mutable file that was created with more shares (larger N) than our current tahoe.cfg . The modification attempt creates new versions of the (0,1,..,newN-1) shares, but leaves the old versions of the (newN,..,oldN-1) shares alone (and throws a assertion error in SDMFSlotWriteProxy.finish_publishing in the process). The mixed versions that result (some shares with e.g. N=10, some with N=20, such that both versions are recoverable) cause problems for the Publish code, even before MDMF landed. Might be related to refs #1390 and refs #1042. ] [layout.py: annotate assertion to figure out 'tahoe backup' failure Brian Warner **20110827195253 Ignore-this: 9b92b954e3ed0d0f80154fff1ff674e5 ] [Add 'tahoe debug dump-cap' support for MDMF, DIR2-CHK, DIR2-MDMF. refs #1507. Brian Warner **20110827195048 Ignore-this: 61c6af5e33fc88e0251e697a50addb2c This also adds tests for all those cases, and fixes an omission in uri.py that broke parsing of DIR2-MDMF-Verifier and DIR2-CHK-Verifier. ] [MDMF: more writable/writeable consistentifications warner@lothar.com**20110827190602 Ignore-this: 22492a9e20c1819ddb12091062888b55 ] [MDMF: s/Writable/Writeable/g, for consistency with existing SDMF code warner@lothar.com**20110827183357 Ignore-this: 9dd312acedbdb2fc2f7bef0d0fb17c0b ] [setup.cfg: remove no-longer-supported test_mac_diskimage alias. refs #1479 david-sarah@jacaranda.org**20110826230345 Ignore-this: 40e908b8937322a290fb8012bfcad02a ] [test_mutable.Update: increase timeout from 120s to 400s, slaves are failing Brian Warner **20110825230140 Ignore-this: 101b1924a30cdbda9b2e419e95ca15ec ] [tests: fix check_memory test zooko@zooko.com**20110825201116 Ignore-this: 4d66299fa8cb61d2ca04b3f45344d835 fixes #1503 ] [TAG allmydata-tahoe-1.9.0a1 warner@lothar.com**20110825161122 Ignore-this: 3cbf49f00dbda58189f893c427f65605 ] Patch bundle hash: 50386a0e43498fa1d630cfce6557749c2c12121a