2 from zope.interface import Interface
3 from foolscap.schema import StringConstraint, ListOf, TupleOf, SetOf, DictOf, \
4 ChoiceOf, IntegerConstraint, Any
5 from foolscap import RemoteInterface, Referenceable
9 Hash = StringConstraint(maxLength=HASH_SIZE,
10 minLength=HASH_SIZE)# binary format 32-byte SHA256 hash
11 Nodeid = StringConstraint(maxLength=20,
12 minLength=20) # binary format 20-byte SHA1 hash
13 FURL = StringConstraint(1000)
14 StorageIndex = StringConstraint(16)
15 URI = StringConstraint(300) # kind of arbitrary
17 MAX_BUCKETS = 256 # per peer -- zfec offers at most 256 shares per file
19 ShareData = StringConstraint(None)
20 URIExtensionData = StringConstraint(1000)
21 Number = IntegerConstraint(8) # 2**(8*8) == 16EiB ~= 18e18 ~= 18 exabytes
23 ReadSize = int # the 'int' constraint is 2**31 == 2Gib -- large files are processed in not-so-large increments
24 WriteEnablerSecret = Hash # used to protect mutable bucket modifications
25 LeaseRenewSecret = Hash # used to protect bucket lease renewal requests
26 LeaseCancelSecret = Hash # used to protect bucket lease cancellation requests
28 class RIStubClient(RemoteInterface):
29 """Each client publishes a service announcement for a dummy object called
30 the StubClient. This object doesn't actually offer any services, but the
31 announcement helps the Introducer keep track of which clients are
32 subscribed (so the grid admin can keep track of things like the size of
33 the grid and the client versions in use. This is the (empty)
34 RemoteInterface for the StubClient."""
36 class RIBucketWriter(RemoteInterface):
37 """ Objects of this kind live on the server side. """
38 def write(offset=Offset, data=ShareData):
43 If the data that has been written is incomplete or inconsistent then
44 the server will throw the data away, else it will store it for future
50 """Abandon all the data that has been written.
54 class RIBucketReader(RemoteInterface):
55 def read(offset=Offset, length=ReadSize):
58 def advise_corrupt_share(reason=str):
59 """Clients who discover hash failures in shares that they have
60 downloaded from me will use this method to inform me about the
61 failures. I will record their concern so that my operator can
62 manually inspect the shares in question. I return None.
64 This is a wrapper around RIStorageServer.advise_corrupt_share(),
65 which is tied to a specific share, and therefore does not need the
66 extra share-identifying arguments. Please see that method for full
70 TestVector = ListOf(TupleOf(Offset, ReadSize, str, str))
71 # elements are (offset, length, operator, specimen)
72 # operator is one of "lt, le, eq, ne, ge, gt"
73 # nop always passes and is used to fetch data while writing.
74 # you should use length==len(specimen) for everything except nop
75 DataVector = ListOf(TupleOf(Offset, ShareData))
76 # (offset, data). This limits us to 30 writes of 1MiB each per call
77 TestAndWriteVectorsForShares = DictOf(int,
80 ChoiceOf(None, Offset), # new_length
82 ReadVector = ListOf(TupleOf(Offset, ReadSize))
83 ReadData = ListOf(ShareData)
84 # returns data[offset:offset+length] for each element of TestVector
86 class RIStorageServer(RemoteInterface):
87 __remote_name__ = "RIStorageServer.tahoe.allmydata.com"
91 Return a dictionary of version information.
93 return DictOf(str, Any())
95 def allocate_buckets(storage_index=StorageIndex,
96 renew_secret=LeaseRenewSecret,
97 cancel_secret=LeaseCancelSecret,
98 sharenums=SetOf(int, maxLength=MAX_BUCKETS),
99 allocated_size=Offset, canary=Referenceable):
101 @param storage_index: the index of the bucket to be created or
103 @param sharenums: these are the share numbers (probably between 0 and
104 99) that the sender is proposing to store on this
106 @param renew_secret: This is the secret used to protect bucket refresh
107 This secret is generated by the client and
108 stored for later comparison by the server. Each
109 server is given a different secret.
110 @param cancel_secret: Like renew_secret, but protects bucket decref.
111 @param canary: If the canary is lost before close(), the bucket is
113 @return: tuple of (alreadygot, allocated), where alreadygot is what we
114 already have and allocated is what we hereby agree to accept.
115 New leases are added for shares in both lists.
117 return TupleOf(SetOf(int, maxLength=MAX_BUCKETS),
118 DictOf(int, RIBucketWriter, maxKeys=MAX_BUCKETS))
120 def add_lease(storage_index=StorageIndex,
121 renew_secret=LeaseRenewSecret,
122 cancel_secret=LeaseCancelSecret):
124 Add a new lease on the given bucket. If the renew_secret matches an
125 existing lease, that lease will be renewed instead. If there is no
126 bucket for the given storage_index, IndexError will be raised.
130 def renew_lease(storage_index=StorageIndex, renew_secret=LeaseRenewSecret):
132 Renew the lease on a given bucket, resetting the timer to 31 days.
133 Some networks will use this, some will not. If there is no bucket for
134 the given storage_index, IndexError will be raised.
136 For mutable shares, if the given renew_secret does not match an
137 existing lease, IndexError will be raised with a note listing the
138 server-nodeids on the existing leases, so leases on migrated shares
139 can be renewed or cancelled. For immutable shares, IndexError
140 (without the note) will be raised.
144 def cancel_lease(storage_index=StorageIndex,
145 cancel_secret=LeaseCancelSecret):
147 Cancel the lease on a given bucket. If this was the last lease on the
148 bucket, the bucket will be deleted. If there is no bucket for the
149 given storage_index, IndexError will be raised.
151 For mutable shares, if the given cancel_secret does not match an
152 existing lease, IndexError will be raised with a note listing the
153 server-nodeids on the existing leases, so leases on migrated shares
154 can be renewed or cancelled. For immutable shares, IndexError
155 (without the note) will be raised.
159 def get_buckets(storage_index=StorageIndex):
160 return DictOf(int, RIBucketReader, maxKeys=MAX_BUCKETS)
164 def slot_readv(storage_index=StorageIndex,
165 shares=ListOf(int), readv=ReadVector):
166 """Read a vector from the numbered shares associated with the given
167 storage index. An empty shares list means to return data from all
168 known shares. Returns a dictionary with one key per share."""
169 return DictOf(int, ReadData) # shnum -> results
171 def slot_testv_and_readv_and_writev(storage_index=StorageIndex,
172 secrets=TupleOf(WriteEnablerSecret,
175 tw_vectors=TestAndWriteVectorsForShares,
178 """General-purpose test-and-set operation for mutable slots. Perform
179 a bunch of comparisons against the existing shares. If they all pass,
180 then apply a bunch of write vectors to those shares. Then use the
181 read vectors to extract data from all the shares and return the data.
183 This method is, um, large. The goal is to allow clients to update all
184 the shares associated with a mutable file in a single round trip.
186 @param storage_index: the index of the bucket to be created or
188 @param write_enabler: a secret that is stored along with the slot.
189 Writes are accepted from any caller who can
190 present the matching secret. A different secret
191 should be used for each slot*server pair.
192 @param renew_secret: This is the secret used to protect bucket refresh
193 This secret is generated by the client and
194 stored for later comparison by the server. Each
195 server is given a different secret.
196 @param cancel_secret: Like renew_secret, but protects bucket decref.
198 The 'secrets' argument is a tuple of (write_enabler, renew_secret,
199 cancel_secret). The first is required to perform any write. The
200 latter two are used when allocating new shares. To simply acquire a
201 new lease on existing shares, use an empty testv and an empty writev.
203 Each share can have a separate test vector (i.e. a list of
204 comparisons to perform). If all vectors for all shares pass, then all
205 writes for all shares are recorded. Each comparison is a 4-tuple of
206 (offset, length, operator, specimen), which effectively does a bool(
207 (read(offset, length)) OPERATOR specimen ) and only performs the
208 write if all these evaluate to True. Basic test-and-set uses 'eq'.
209 Write-if-newer uses a seqnum and (offset, length, 'lt', specimen).
210 Write-if-same-or-newer uses 'le'.
212 Reads from the end of the container are truncated, and missing shares
213 behave like empty ones, so to assert that a share doesn't exist (for
214 use when creating a new share), use (0, 1, 'eq', '').
216 The write vector will be applied to the given share, expanding it if
217 necessary. A write vector applied to a share number that did not
218 exist previously will cause that share to be created.
220 Each write vector is accompanied by a 'new_length' argument. If
221 new_length is not None, use it to set the size of the container. This
222 can be used to pre-allocate space for a series of upcoming writes, or
223 truncate existing data. If the container is growing, new_length will
224 be applied before datav. If the container is shrinking, it will be
225 applied afterwards. If new_length==0, the share will be deleted.
227 The read vector is used to extract data from all known shares,
228 *before* any writes have been applied. The same vector is used for
229 all shares. This captures the state that was tested by the test
232 This method returns two values: a boolean and a dict. The boolean is
233 True if the write vectors were applied, False if not. The dict is
234 keyed by share number, and each value contains a list of strings, one
235 for each element of the read vector.
237 If the write_enabler is wrong, this will raise BadWriteEnablerError.
238 To enable share migration (using update_write_enabler), the exception
239 will have the nodeid used for the old write enabler embedded in it,
240 in the following string::
242 The write enabler was recorded by nodeid '%s'.
244 Note that the nodeid here is encoded using the same base32 encoding
245 used by Foolscap and allmydata.util.idlib.nodeid_b2a().
248 return TupleOf(bool, DictOf(int, ReadData))
250 def advise_corrupt_share(share_type=str, storage_index=StorageIndex,
251 shnum=int, reason=str):
252 """Clients who discover hash failures in shares that they have
253 downloaded from me will use this method to inform me about the
254 failures. I will record their concern so that my operator can
255 manually inspect the shares in question. I return None.
257 'share_type' is either 'mutable' or 'immutable'. 'storage_index' is a
258 (binary) storage index string, and 'shnum' is the integer share
259 number. 'reason' is a human-readable explanation of the problem,
260 probably including some expected hash values and the computed ones
261 which did not match. Corruption advisories for mutable shares should
262 include a hash of the public key (the same value that appears in the
263 mutable-file verify-cap), since the current share format does not
267 class IStorageBucketWriter(Interface):
269 Objects of this kind live on the client side.
271 def put_block(segmentnum=int, data=ShareData):
272 """@param data: For most segments, this data will be 'blocksize'
273 bytes in length. The last segment might be shorter.
274 @return: a Deferred that fires (with None) when the operation completes
277 def put_plaintext_hashes(hashes=ListOf(Hash)):
279 @return: a Deferred that fires (with None) when the operation completes
282 def put_crypttext_hashes(hashes=ListOf(Hash)):
284 @return: a Deferred that fires (with None) when the operation completes
287 def put_block_hashes(blockhashes=ListOf(Hash)):
289 @return: a Deferred that fires (with None) when the operation completes
292 def put_share_hashes(sharehashes=ListOf(TupleOf(int, Hash))):
294 @return: a Deferred that fires (with None) when the operation completes
297 def put_uri_extension(data=URIExtensionData):
298 """This block of data contains integrity-checking information (hashes
299 of plaintext, crypttext, and shares), as well as encoding parameters
300 that are necessary to recover the data. This is a serialized dict
301 mapping strings to other strings. The hash of this data is kept in
302 the URI and verified before any of the data is used. All buckets for
303 a given file contain identical copies of this data.
305 The serialization format is specified with the following pseudocode:
306 for k in sorted(dict.keys()):
307 assert re.match(r'^[a-zA-Z_\-]+$', k)
308 write(k + ':' + netstring(dict[k]))
310 @return: a Deferred that fires (with None) when the operation completes
314 """Finish writing and close the bucket. The share is not finalized
315 until this method is called: if the uploading client disconnects
316 before calling close(), the partially-written share will be
319 @return: a Deferred that fires (with None) when the operation completes
322 class IStorageBucketReader(Interface):
324 def get_block_data(blocknum=int, blocksize=int, size=int):
325 """Most blocks will be the same size. The last block might be shorter
331 def get_crypttext_hashes():
333 @return: ListOf(Hash)
336 def get_block_hashes(at_least_these=SetOf(int)):
338 @return: ListOf(Hash)
341 def get_share_hashes(at_least_these=SetOf(int)):
343 @return: ListOf(TupleOf(int, Hash))
346 def get_uri_extension():
348 @return: URIExtensionData
353 # hm, we need a solution for forward references in schemas
354 FileNode_ = Any() # TODO: foolscap needs constraints on copyables
355 DirectoryNode_ = Any() # TODO: same
356 AnyNode_ = ChoiceOf(FileNode_, DirectoryNode_)
359 class IURI(Interface):
360 def init_from_string(uri):
361 """Accept a string (as created by my to_string() method) and populate
362 this instance with its data. I am not normally called directly,
363 please use the module-level uri.from_string() function to convert
364 arbitrary URI strings into IURI-providing instances."""
367 """Return False if this URI be used to modify the data. Return True
368 if this URI cannot be used to modify the data."""
371 """Return True if the data can be modified by *somebody* (perhaps
372 someone who has a more powerful URI than this one)."""
375 """Return another IURI instance, which represents a read-only form of
376 this one. If is_readonly() is True, this returns self."""
378 def get_verify_cap():
379 """Return an instance that provides IVerifierURI, which can be used
380 to check on the availability of the file or directory, without
381 providing enough capabilities to actually read or modify the
382 contents. This may return None if the file does not need checking or
383 verification (e.g. LIT URIs).
387 """Return a string of printable ASCII characters, suitable for
388 passing into init_from_string."""
390 class IVerifierURI(Interface, IURI):
391 def init_from_string(uri):
392 """Accept a string (as created by my to_string() method) and populate
393 this instance with its data. I am not normally called directly,
394 please use the module-level uri.from_string() function to convert
395 arbitrary URI strings into IURI-providing instances."""
398 """Return a string of printable ASCII characters, suitable for
399 passing into init_from_string."""
401 class IDirnodeURI(Interface):
402 """I am a URI which represents a dirnode."""
405 class IFileURI(Interface):
406 """I am a URI which represents a filenode."""
408 """Return the length (in bytes) of the file that I represent."""
410 class IImmutableFileURI(IFileURI):
413 class IMutableFileURI(Interface):
414 """I am a URI which represents a mutable filenode."""
415 class INewDirectoryURI(Interface):
417 class IReadonlyNewDirectoryURI(Interface):
421 class IFilesystemNode(Interface):
424 Return the URI that can be used by others to get access to this
425 node. If this node is read-only, the URI will only offer read-only
426 access. If this node is read-write, the URI will offer read-write
429 If you have read-write access to a node and wish to share merely
430 read-only access with others, use get_readonly_uri().
433 def get_readonly_uri():
434 """Return the directory URI that can be used by others to get
435 read-only access to this directory node. The result is a read-only
436 URI, regardless of whether this dirnode is read-only or read-write.
438 If you have merely read-only access to this dirnode,
439 get_readonly_uri() will return the same thing as get_uri().
442 def get_repair_cap():
443 """Return an IURI instance that can be used to repair the file, or
444 None if this node cannot be repaired (either because it is not
445 distributed, like a LIT file, or because the node does not represent
446 sufficient authority to create a repair-cap, like a read-only RSA
447 mutable file node [which cannot create the correct write-enablers]).
450 def get_verify_cap():
451 """Return an IVerifierURI instance that represents the
452 'verifiy/refresh capability' for this node. The holder of this
453 capability will be able to renew the lease for this node, protecting
454 it from garbage-collection. They will also be able to ask a server if
455 it holds a share for the file or directory.
458 def get_storage_index():
459 """Return a string with the (binary) storage index in use on this
460 download. This may be None if there is no storage index (i.e. LIT
464 """Return True if this reference provides mutable access to the given
465 file or directory (i.e. if you can modify it), or False if not. Note
466 that even if this reference is read-only, someone else may hold a
467 read-write reference to it."""
470 """Return True if this file or directory is mutable (by *somebody*,
471 not necessarily you), False if it is is immutable. Note that a file
472 might be mutable overall, but your reference to it might be
473 read-only. On the other hand, all references to an immutable file
474 will be read-only; there are no read-write references to an immutable
478 class IMutableFilesystemNode(IFilesystemNode):
481 class IFileNode(IFilesystemNode):
482 def download(target):
483 """Download the file's contents to a given IDownloadTarget"""
485 def download_to_data():
486 """Download the file's contents. Return a Deferred that fires
487 with those contents."""
490 """Return the length (in bytes) of the data this node represents."""
492 def read(consumer, offset=0, size=None):
493 """Download a portion (possibly all) of the file's contents, making
494 them available to the given IConsumer. Return a Deferred that fires
495 (with the consumer) when the consumer is unregistered (either because
496 the last byte has been given to it, or because the consumer threw an
497 exception during write(), possibly because it no longer wants to
498 receive data). The portion downloaded will start at 'offset' and
499 contain 'size' bytes (or the remainder of the file if size==None).
501 The consumer will be used in non-streaming mode: an IPullProducer
502 will be attached to it.
504 The consumer will not receive data right away: several network trips
505 must occur first. The order of events will be::
507 consumer.registerProducer(p, streaming)
508 (if streaming == False)::
509 consumer does p.resumeProducing()
511 consumer does p.resumeProducing()
512 consumer.write(data).. (repeat until all data is written)
513 consumer.unregisterProducer()
514 deferred.callback(consumer)
516 If a download error occurs, or an exception is raised by
517 consumer.registerProducer() or consumer.write(), I will call
518 consumer.unregisterProducer() and then deliver the exception via
519 deferred.errback(). To cancel the download, the consumer should call
520 p.stopProducing(), which will result in an exception being delivered
521 via deferred.errback().
523 A simple download-to-memory consumer example would look like this::
525 class MemoryConsumer:
526 implements(IConsumer)
530 def registerProducer(self, p, streaming):
531 assert streaming == False
534 def write(self, data):
535 self.chunks.append(data)
536 def unregisterProducer(self):
538 d = filenode.read(MemoryConsumer())
539 d.addCallback(lambda mc: "".join(mc.chunks))
544 class IMutableFileNode(IFileNode, IMutableFilesystemNode):
545 """I provide access to a 'mutable file', which retains its identity
546 regardless of what contents are put in it.
548 The consistency-vs-availability problem means that there might be
549 multiple versions of a file present in the grid, some of which might be
550 unrecoverable (i.e. have fewer than 'k' shares). These versions are
551 loosely ordered: each has a sequence number and a hash, and any version
552 with seqnum=N was uploaded by a node which has seen at least one version
555 The 'servermap' (an instance of IMutableFileServerMap) is used to
556 describe the versions that are known to be present in the grid, and which
557 servers are hosting their shares. It is used to represent the 'state of
558 the world', and is used for this purpose by my test-and-set operations.
559 Downloading the contents of the mutable file will also return a
560 servermap. Uploading a new version into the mutable file requires a
561 servermap as input, and the semantics of the replace operation is
562 'replace the file with my new version if it looks like nobody else has
563 changed the file since my previous download'. Because the file is
564 distributed, this is not a perfect test-and-set operation, but it will do
565 its best. If the replace process sees evidence of a simultaneous write,
566 it will signal an UncoordinatedWriteError, so that the caller can take
570 Most readers will want to use the 'best' current version of the file, and
571 should use my 'download_best_version()' method.
573 To unconditionally replace the file, callers should use overwrite(). This
574 is the mode that user-visible mutable files will probably use.
576 To apply some delta to the file, call modify() with a callable modifier
577 function that can apply the modification that you want to make. This is
578 the mode that dirnodes will use, since most directory modification
579 operations can be expressed in terms of deltas to the directory state.
582 Three methods are available for users who need to perform more complex
583 operations. The first is get_servermap(), which returns an up-to-date
584 servermap using a specified mode. The second is download_version(), which
585 downloads a specific version (not necessarily the 'best' one). The third
586 is 'upload', which accepts new contents and a servermap (which must have
587 been updated with MODE_WRITE). The upload method will attempt to apply
588 the new contents as long as no other node has modified the file since the
589 servermap was updated. This might be useful to a caller who wants to
590 merge multiple versions into a single new one.
592 Note that each time the servermap is updated, a specific 'mode' is used,
593 which determines how many peers are queried. To use a servermap for my
594 replace() method, that servermap must have been updated in MODE_WRITE.
595 These modes are defined in allmydata.mutable.common, and consist of
596 MODE_READ, MODE_WRITE, MODE_ANYTHING, and MODE_CHECK. Please look in
597 allmydata/mutable/servermap.py for details about the differences.
599 Mutable files are currently limited in size (about 3.5MB max) and can
600 only be retrieved and updated all-at-once, as a single big string. Future
601 versions of our mutable files will remove this restriction.
604 def download_best_version():
605 """Download the 'best' available version of the file, meaning one of
606 the recoverable versions with the highest sequence number. If no
607 uncoordinated writes have occurred, and if enough shares are
608 available, then this will be the most recent version that has been
611 I update an internal servermap with MODE_READ, determine which
612 version of the file is indicated by
613 servermap.best_recoverable_version(), and return a Deferred that
614 fires with its contents. If no version is recoverable, the Deferred
615 will errback with UnrecoverableFileError.
618 def get_size_of_best_version():
619 """Find the size of the version that would be downloaded with
620 download_best_version(), without actually downloading the whole file.
622 I return a Deferred that fires with an integer.
625 def overwrite(new_contents):
626 """Unconditionally replace the contents of the mutable file with new
627 ones. This simply chains get_servermap(MODE_WRITE) and upload(). This
628 is only appropriate to use when the new contents of the file are
629 completely unrelated to the old ones, and you do not care about other
632 I return a Deferred that fires (with a PublishStatus object) when the
633 update has completed.
636 def modify(modifier_cb):
637 """Modify the contents of the file, by downloading the current
638 version, applying the modifier function (or bound method), then
639 uploading the new version. I return a Deferred that fires (with a
640 PublishStatus object) when the update is complete.
642 The modifier callable will be given three arguments: a string (with
643 the old contents), a 'first_time' boolean, and a servermap. As with
644 download_best_version(), the old contents will be from the best
645 recoverable version, but the modifier can use the servermap to make
646 other decisions (such as refusing to apply the delta if there are
647 multiple parallel versions, or if there is evidence of a newer
648 unrecoverable version). 'first_time' will be True the first time the
649 modifier is called, and False on any subsequent calls.
651 The callable should return a string with the new contents. The
652 callable must be prepared to be called multiple times, and must
653 examine the input string to see if the change that it wants to make
654 is already present in the old version. If it does not need to make
655 any changes, it can either return None, or return its input string.
657 If the modifier raises an exception, it will be returned in the
662 def get_servermap(mode):
663 """Return a Deferred that fires with an IMutableFileServerMap
664 instance, updated using the given mode.
667 def download_version(servermap, version):
668 """Download a specific version of the file, using the servermap
669 as a guide to where the shares are located.
671 I return a Deferred that fires with the requested contents, or
672 errbacks with UnrecoverableFileError. Note that a servermap which was
673 updated with MODE_ANYTHING or MODE_READ may not know about shares for
674 all versions (those modes stop querying servers as soon as they can
675 fulfil their goals), so you may want to use MODE_CHECK (which checks
676 everything) to get increased visibility.
679 def upload(new_contents, servermap):
680 """Replace the contents of the file with new ones. This requires a
681 servermap that was previously updated with MODE_WRITE.
683 I attempt to provide test-and-set semantics, in that I will avoid
684 modifying any share that is different than the version I saw in the
685 servermap. However, if another node is writing to the file at the
686 same time as me, I may manage to update some shares while they update
687 others. If I see any evidence of this, I will signal
688 UncoordinatedWriteError, and the file will be left in an inconsistent
689 state (possibly the version you provided, possibly the old version,
690 possibly somebody else's version, and possibly a mix of shares from
693 The recommended response to UncoordinatedWriteError is to either
694 return it to the caller (since they failed to coordinate their
695 writes), or to attempt some sort of recovery. It may be sufficient to
696 wait a random interval (with exponential backoff) and repeat your
697 operation. If I do not signal UncoordinatedWriteError, then I was
698 able to write the new version without incident.
700 I return a Deferred that fires (with a PublishStatus object) when the
701 publish has completed. I will update the servermap in-place with the
702 location of all new shares.
706 """Return this filenode's writekey, or None if the node does not have
707 write-capability. This may be used to assist with data structures
708 that need to make certain data available only to writers, such as the
709 read-write child caps in dirnodes. The recommended process is to have
710 reader-visible data be submitted to the filenode in the clear (where
711 it will be encrypted by the filenode using the readkey), but encrypt
712 writer-visible data using this writekey.
715 class NotEnoughSharesError(Exception):
718 class ExistingChildError(Exception):
719 """A directory node was asked to add or replace a child that already
720 exists, and overwrite= was set to False."""
722 class NoSuchChildError(Exception):
723 """A directory node was asked to fetch a child which does not exist."""
725 class IDirectoryNode(IMutableFilesystemNode):
726 """I represent a name-to-child mapping, holding the tahoe equivalent of a
727 directory. All child names are unicode strings, and all children are some
728 sort of IFilesystemNode (either files or subdirectories).
733 The dirnode ('1') URI returned by this method can be used in
734 set_uri() on a different directory ('2') to 'mount' a reference to
735 this directory ('1') under the other ('2'). This URI is just a
736 string, so it can be passed around through email or other out-of-band
740 def get_readonly_uri():
742 The dirnode ('1') URI returned by this method can be used in
743 set_uri() on a different directory ('2') to 'mount' a reference to
744 this directory ('1') under the other ('2'). This URI is just a
745 string, so it can be passed around through email or other out-of-band
750 """I return a Deferred that fires with a dictionary mapping child
751 name (a unicode string) to (node, metadata_dict) tuples, in which
752 'node' is either an IFileNode or IDirectoryNode, and 'metadata_dict'
753 is a dictionary of metadata."""
756 """I return a Deferred that fires with a boolean, True if there
757 exists a child of the given name, False if not. The child name must
758 be a unicode string."""
761 """I return a Deferred that fires with a specific named child node,
762 either an IFileNode or an IDirectoryNode. The child name must be a
763 unicode string. I raise NoSuchChildError if I do not have a child by
766 def get_metadata_for(name):
767 """I return a Deferred that fires with the metadata dictionary for a
768 specific named child node. This metadata is stored in the *edge*, not
769 in the child, so it is attached to the parent dirnode rather than the
770 child dir-or-file-node. The child name must be a unicode string. I
771 raise NoSuchChildError if I do not have a child by that name."""
773 def set_metadata_for(name, metadata):
774 """I replace any existing metadata for the named child with the new
775 metadata. The child name must be a unicode string. This metadata is
776 stored in the *edge*, not in the child, so it is attached to the
777 parent dirnode rather than the child dir-or-file-node. I return a
778 Deferred (that fires with this dirnode) when the operation is
779 complete. I raise NoSuchChildError if I do not have a child by that
782 def get_child_at_path(path):
783 """Transform a child path into an IDirectoryNode or IFileNode.
785 I perform a recursive series of 'get' operations to find the named
786 descendant node. I return a Deferred that fires with the node, or
787 errbacks with NoSuchChildError if the node could not be found.
789 The path can be either a single string (slash-separated) or a list of
790 path-name elements. All elements must be unicode strings.
793 def get_child_and_metadata_at_path(path):
794 """Transform a child path into an IDirectoryNode/IFileNode and
797 I am like get_child_at_path(), but my Deferred fires with a tuple of
798 (node, metadata). The metadata comes from the last edge. If the path
799 is empty, the metadata will be an empty dictionary.
802 def set_uri(name, child_uri, metadata=None, overwrite=True):
803 """I add a child (by URI) at the specific name. I return a Deferred
804 that fires when the operation finishes. If overwrite= is True, I will
805 replace any existing child of the same name, otherwise an existing
806 child will cause me to return ExistingChildError. The child name must
809 The child_uri could be for a file, or for a directory (either
810 read-write or read-only, using a URI that came from get_uri() ).
812 If metadata= is provided, I will use it as the metadata for the named
813 edge. This will replace any existing metadata. If metadata= is left
814 as the default value of None, I will set ['mtime'] to the current
815 time, and I will set ['ctime'] to the current time if there was not
816 already a child by this name present. This roughly matches the
817 ctime/mtime semantics of traditional filesystems.
819 If this directory node is read-only, the Deferred will errback with a
822 def set_children(entries, overwrite=True):
823 """Add multiple (name, child_uri) pairs (or (name, child_uri,
824 metadata) triples) to a directory node. Returns a Deferred that fires
825 (with None) when the operation finishes. This is equivalent to
826 calling set_uri() multiple times, but is much more efficient. All
827 child names must be unicode strings.
830 def set_node(name, child, metadata=None, overwrite=True):
831 """I add a child at the specific name. I return a Deferred that fires
832 when the operation finishes. This Deferred will fire with the child
833 node that was just added. I will replace any existing child of the
834 same name. The child name must be a unicode string. The 'child'
835 instance must be an instance providing IDirectoryNode or IFileNode.
837 If metadata= is provided, I will use it as the metadata for the named
838 edge. This will replace any existing metadata. If metadata= is left
839 as the default value of None, I will set ['mtime'] to the current
840 time, and I will set ['ctime'] to the current time if there was not
841 already a child by this name present. This roughly matches the
842 ctime/mtime semantics of traditional filesystems.
844 If this directory node is read-only, the Deferred will errback with a
847 def set_nodes(entries, overwrite=True):
848 """Add multiple (name, child_node) pairs (or (name, child_node,
849 metadata) triples) to a directory node. Returns a Deferred that fires
850 (with None) when the operation finishes. This is equivalent to
851 calling set_node() multiple times, but is much more efficient. All
852 child names must be unicode strings."""
855 def add_file(name, uploadable, metadata=None, overwrite=True):
856 """I upload a file (using the given IUploadable), then attach the
857 resulting FileNode to the directory at the given name. I set metadata
858 the same way as set_uri and set_node. The child name must be a
861 I return a Deferred that fires (with the IFileNode of the uploaded
862 file) when the operation completes."""
865 """I remove the child at the specific name. I return a Deferred that
866 fires when the operation finishes. The child name must be a unicode
867 string. I raise NoSuchChildError if I do not have a child by that
870 def create_empty_directory(name, overwrite=True):
871 """I create and attach an empty directory at the given name. The
872 child name must be a unicode string. I return a Deferred that fires
873 when the operation finishes."""
875 def move_child_to(current_child_name, new_parent, new_child_name=None,
877 """I take one of my children and move them to a new parent. The child
878 is referenced by name. On the new parent, the child will live under
879 'new_child_name', which defaults to 'current_child_name'. TODO: what
880 should we do about metadata? I return a Deferred that fires when the
881 operation finishes. The child name must be a unicode string. I raise
882 NoSuchChildError if I do not have a child by that name."""
884 def build_manifest():
885 """I generate a table of everything reachable from this directory.
886 I also compute deep-stats as described below.
888 I return a Monitor. The Monitor's results will be a dictionary with
891 res['manifest']: a list of (path, cap) tuples for all nodes
892 (directories and files) reachable from this one.
893 'path' will be a tuple of unicode strings. The
894 origin dirnode will be represented by an empty path
896 res['verifycaps']: a list of (printable) verifycap strings, one for
897 each reachable non-LIT node. This is a set:
898 it will contain no duplicates.
899 res['storage-index']: a list of (base32) storage index strings,
900 one for each reachable non-LIT node. This is
901 a set: it will contain no duplicates.
902 res['stats']: a dictionary, the same that is generated by
903 start_deep_stats() below.
905 The Monitor will also have an .origin_si attribute with the (binary)
906 storage index of the starting point.
909 def start_deep_stats():
910 """Return a Monitor, examining all nodes (directories and files)
911 reachable from this one. The Monitor's results will be a dictionary
912 with the following keys::
914 count-immutable-files: count of how many CHK files are in the set
915 count-mutable-files: same, for mutable files (does not include
917 count-literal-files: same, for LIT files
918 count-files: sum of the above three
920 count-directories: count of directories
922 size-immutable-files: total bytes for all CHK files in the set
923 size-mutable-files (TODO): same, for current version of all mutable
924 files, does not include directories
925 size-literal-files: same, for LIT files
926 size-directories: size of mutable files used by directories
928 largest-directory: number of bytes in the largest directory
929 largest-directory-children: number of children in the largest
931 largest-immutable-file: number of bytes in the largest CHK file
933 size-mutable-files is not yet implemented, because it would involve
934 even more queries than deep_stats does.
936 The Monitor will also have an .origin_si attribute with the (binary)
937 storage index of the starting point.
939 This operation will visit every directory node underneath this one,
940 and can take a long time to run. On a typical workstation with good
941 bandwidth, this can examine roughly 15 directories per second (and
942 takes several minutes of 100% CPU for ~1700 directories).
945 class ICodecEncoder(Interface):
946 def set_params(data_size, required_shares, max_shares):
947 """Set up the parameters of this encoder.
949 This prepares the encoder to perform an operation that converts a
950 single block of data into a number of shares, such that a future
951 ICodecDecoder can use a subset of these shares to recover the
952 original data. This operation is invoked by calling encode(). Once
953 the encoding parameters are set up, the encode operation can be
954 invoked multiple times.
956 set_params() prepares the encoder to accept blocks of input data that
957 are exactly 'data_size' bytes in length. The encoder will be prepared
958 to produce 'max_shares' shares for each encode() operation (although
959 see the 'desired_share_ids' to use less CPU). The encoding math will
960 be chosen such that the decoder can get by with as few as
961 'required_shares' of these shares and still reproduce the original
962 data. For example, set_params(1000, 5, 5) offers no redundancy at
963 all, whereas set_params(1000, 1, 10) provides 10x redundancy.
965 Numerical Restrictions: 'data_size' is required to be an integral
966 multiple of 'required_shares'. In general, the caller should choose
967 required_shares and max_shares based upon their reliability
968 requirements and the number of peers available (the total storage
969 space used is roughly equal to max_shares*data_size/required_shares),
970 then choose data_size to achieve the memory footprint desired (larger
971 data_size means more efficient operation, smaller data_size means
972 smaller memory footprint).
974 In addition, 'max_shares' must be equal to or greater than
975 'required_shares'. Of course, setting them to be equal causes
976 encode() to degenerate into a particularly slow form of the 'split'
979 See encode() for more details about how these parameters are used.
981 set_params() must be called before any other ICodecEncoder methods
986 """Return the 3-tuple of data_size, required_shares, max_shares"""
988 def get_encoder_type():
989 """Return a short string that describes the type of this encoder.
991 There is required to be a global table of encoder classes. This method
992 returns an index into this table; the value at this index is an
993 encoder class, and this encoder is an instance of that class.
996 def get_block_size():
997 """Return the length of the shares that encode() will produce.
1000 def encode_proposal(data, desired_share_ids=None):
1001 """Encode some data.
1003 'data' must be a string (or other buffer object), and len(data) must
1004 be equal to the 'data_size' value passed earlier to set_params().
1006 This will return a Deferred that will fire with two lists. The first
1007 is a list of shares, each of which is a string (or other buffer
1008 object) such that len(share) is the same as what get_share_size()
1009 returned earlier. The second is a list of shareids, in which each is
1010 an integer. The lengths of the two lists will always be equal to each
1011 other. The user should take care to keep each share closely
1012 associated with its shareid, as one is useless without the other.
1014 The length of this output list will normally be the same as the value
1015 provided to the 'max_shares' parameter of set_params(). This may be
1016 different if 'desired_share_ids' is provided.
1018 'desired_share_ids', if provided, is required to be a sequence of
1019 ints, each of which is required to be >= 0 and < max_shares. If not
1020 provided, encode() will produce 'max_shares' shares, as if
1021 'desired_share_ids' were set to range(max_shares). You might use this
1022 if you initially thought you were going to use 10 peers, started
1023 encoding, and then two of the peers dropped out: you could use
1024 desired_share_ids= to skip the work (both memory and CPU) of
1025 producing shares for the peers which are no longer available.
1029 def encode(inshares, desired_share_ids=None):
1030 """Encode some data. This may be called multiple times. Each call is
1033 inshares is a sequence of length required_shares, containing buffers
1034 (i.e. strings), where each buffer contains the next contiguous
1035 non-overlapping segment of the input data. Each buffer is required to
1036 be the same length, and the sum of the lengths of the buffers is
1037 required to be exactly the data_size promised by set_params(). (This
1038 implies that the data has to be padded before being passed to
1039 encode(), unless of course it already happens to be an even multiple
1040 of required_shares in length.)
1042 ALSO: the requirement to break up your data into 'required_shares'
1043 chunks before calling encode() feels a bit surprising, at least from
1044 the point of view of a user who doesn't know how FEC works. It feels
1045 like an implementation detail that has leaked outside the
1046 abstraction barrier. Can you imagine a use case in which the data to
1047 be encoded might already be available in pre-segmented chunks, such
1048 that it is faster or less work to make encode() take a list rather
1049 than splitting a single string?
1051 ALSO ALSO: I think 'inshares' is a misleading term, since encode()
1052 is supposed to *produce* shares, so what it *accepts* should be
1053 something other than shares. Other places in this interface use the
1054 word 'data' for that-which-is-not-shares.. maybe we should use that
1057 'desired_share_ids', if provided, is required to be a sequence of
1058 ints, each of which is required to be >= 0 and < max_shares. If not
1059 provided, encode() will produce 'max_shares' shares, as if
1060 'desired_share_ids' were set to range(max_shares). You might use this
1061 if you initially thought you were going to use 10 peers, started
1062 encoding, and then two of the peers dropped out: you could use
1063 desired_share_ids= to skip the work (both memory and CPU) of
1064 producing shares for the peers which are no longer available.
1066 For each call, encode() will return a Deferred that fires with two
1067 lists, one containing shares and the other containing the shareids.
1068 The get_share_size() method can be used to determine the length of
1069 the share strings returned by encode(). Each shareid is a small
1070 integer, exactly as passed into 'desired_share_ids' (or
1071 range(max_shares), if desired_share_ids was not provided).
1073 The shares and their corresponding shareids are required to be kept
1074 together during storage and retrieval. Specifically, the share data is
1075 useless by itself: the decoder needs to be told which share is which
1076 by providing it with both the shareid and the actual share data.
1078 This function will allocate an amount of memory roughly equal to::
1080 (max_shares - required_shares) * get_share_size()
1082 When combined with the memory that the caller must allocate to
1083 provide the input data, this leads to a memory footprint roughly
1084 equal to the size of the resulting encoded shares (i.e. the expansion
1085 factor times the size of the input segment).
1090 # returning a list of (shareidN,shareN) tuples instead of a pair of
1091 # lists (shareids..,shares..). Brian thought the tuples would
1092 # encourage users to keep the share and shareid together throughout
1093 # later processing, Zooko pointed out that the code to iterate
1094 # through two lists is not really more complicated than using a list
1095 # of tuples and there's also a performance improvement
1097 # having 'data_size' not required to be an integral multiple of
1098 # 'required_shares'. Doing this would require encode() to perform
1099 # padding internally, and we'd prefer to have any padding be done
1100 # explicitly by the caller. Yes, it is an abstraction leak, but
1101 # hopefully not an onerous one.
1104 class ICodecDecoder(Interface):
1105 def set_params(data_size, required_shares, max_shares):
1106 """Set the params. They have to be exactly the same ones that were used for encoding. """
1108 def get_needed_shares():
1109 """Return the number of shares needed to reconstruct the data.
1110 set_params() is required to be called before this."""
1112 def decode(some_shares, their_shareids):
1113 """Decode a partial list of shares into data.
1115 'some_shares' is required to be a sequence of buffers of sharedata, a
1116 subset of the shares returned by ICodecEncode.encode(). Each share is
1117 required to be of the same length. The i'th element of their_shareids
1118 is required to be the shareid of the i'th buffer in some_shares.
1120 This returns a Deferred which fires with a sequence of buffers. This
1121 sequence will contain all of the segments of the original data, in
1122 order. The sum of the lengths of all of the buffers will be the
1123 'data_size' value passed into the original ICodecEncode.set_params()
1124 call. To get back the single original input block of data, use
1125 ''.join(output_buffers), or you may wish to simply write them in
1126 order to an output file.
1128 Note that some of the elements in the result sequence may be
1129 references to the elements of the some_shares input sequence. In
1130 particular, this means that if those share objects are mutable (e.g.
1131 arrays) and if they are changed, then both the input (the
1132 'some_shares' parameter) and the output (the value given when the
1133 deferred is triggered) will change.
1135 The length of 'some_shares' is required to be exactly the value of
1136 'required_shares' passed into the original ICodecEncode.set_params()
1140 class IEncoder(Interface):
1141 """I take an object that provides IEncryptedUploadable, which provides
1142 encrypted data, and a list of shareholders. I then encode, hash, and
1143 deliver shares to those shareholders. I will compute all the necessary
1144 Merkle hash trees that are necessary to validate the crypttext that
1145 eventually comes back from the shareholders. I provide the URI Extension
1146 Block Hash, and the encoding parameters, both of which must be included
1149 I do not choose shareholders, that is left to the IUploader. I must be
1150 given a dict of RemoteReferences to storage buckets that are ready and
1151 willing to receive data.
1155 """Specify the number of bytes that will be encoded. This must be
1156 peformed before get_serialized_params() can be called.
1158 def set_params(params):
1159 """Override the default encoding parameters. 'params' is a tuple of
1160 (k,d,n), where 'k' is the number of required shares, 'd' is the
1161 shares_of_happiness, and 'n' is the total number of shares that will
1164 Encoding parameters can be set in three ways. 1: The Encoder class
1165 provides defaults (3/7/10). 2: the Encoder can be constructed with
1166 an 'options' dictionary, in which the
1167 needed_and_happy_and_total_shares' key can be a (k,d,n) tuple. 3:
1168 set_params((k,d,n)) can be called.
1170 If you intend to use set_params(), you must call it before
1171 get_share_size or get_param are called.
1174 def set_encrypted_uploadable(u):
1175 """Provide a source of encrypted upload data. 'u' must implement
1176 IEncryptedUploadable.
1178 When this is called, the IEncryptedUploadable will be queried for its
1179 length and the storage_index that should be used.
1181 This returns a Deferred that fires with this Encoder instance.
1183 This must be performed before start() can be called.
1186 def get_param(name):
1187 """Return an encoding parameter, by name.
1189 'storage_index': return a string with the (16-byte truncated SHA-256
1190 hash) storage index to which these shares should be
1193 'share_counts': return a tuple describing how many shares are used:
1194 (needed_shares, shares_of_happiness, total_shares)
1196 'num_segments': return an int with the number of segments that
1199 'segment_size': return an int with the size of each segment.
1201 'block_size': return the size of the individual blocks that will
1202 be delivered to a shareholder's put_block() method. By
1203 knowing this, the shareholder will be able to keep all
1204 blocks in a single file and still provide random access
1205 when reading them. # TODO: can we avoid exposing this?
1207 'share_size': an int with the size of the data that will be stored
1208 on each shareholder. This is aggregate amount of data
1209 that will be sent to the shareholder, summed over all
1210 the put_block() calls I will ever make. It is useful to
1211 determine this size before asking potential
1212 shareholders whether they will grant a lease or not,
1213 since their answers will depend upon how much space we
1214 need. TODO: this might also include some amount of
1215 overhead, like the size of all the hashes. We need to
1216 decide whether this is useful or not.
1218 'serialized_params': a string with a concise description of the
1219 codec name and its parameters. This may be passed
1220 into the IUploadable to let it make sure that
1221 the same file encoded with different parameters
1222 will result in different storage indexes.
1224 Once this is called, set_size() and set_params() may not be called.
1227 def set_shareholders(shareholders):
1228 """Tell the encoder where to put the encoded shares. 'shareholders'
1229 must be a dictionary that maps share number (an integer ranging from
1230 0 to n-1) to an instance that provides IStorageBucketWriter. This
1231 must be performed before start() can be called."""
1234 """Begin the encode/upload process. This involves reading encrypted
1235 data from the IEncryptedUploadable, encoding it, uploading the shares
1236 to the shareholders, then sending the hash trees.
1238 set_encrypted_uploadable() and set_shareholders() must be called
1239 before this can be invoked.
1241 This returns a Deferred that fires with a verify cap when the upload process is
1242 complete. The verifycap, plus the encryption key, is sufficient to construct the read
1246 class IDecoder(Interface):
1247 """I take a list of shareholders and some setup information, then
1248 download, validate, decode, and decrypt data from them, writing the
1249 results to an output file.
1251 I do not locate the shareholders, that is left to the IDownloader. I must
1252 be given a dict of RemoteReferences to storage buckets that are ready to
1257 """I take a file-like object (providing write and close) to which all
1258 the plaintext data will be written.
1260 TODO: producer/consumer . Maybe write() should return a Deferred that
1261 indicates when it will accept more data? But probably having the
1262 IDecoder be a producer is easier to glue to IConsumer pieces.
1265 def set_shareholders(shareholders):
1266 """I take a dictionary that maps share identifiers (small integers)
1267 to RemoteReferences that provide RIBucketReader. This must be called
1271 """I start the download. This process involves retrieving data and
1272 hash chains from the shareholders, using the hashes to validate the
1273 data, decoding the shares into segments, decrypting the segments,
1274 then writing the resulting plaintext to the output file.
1276 I return a Deferred that will fire (with self) when the download is
1280 class IDownloadTarget(Interface):
1281 # Note that if the IDownloadTarget is also an IConsumer, the downloader
1282 # will register itself as a producer. This allows the target to invoke
1283 # downloader.pauseProducing, resumeProducing, and stopProducing.
1285 """Called before any calls to write() or close(). If an error
1286 occurs before any data is available, fail() may be called without
1287 a previous call to open().
1289 'size' is the length of the file being downloaded, in bytes."""
1292 """Output some data to the target."""
1294 """Inform the target that there is no more data to be written."""
1296 """fail() is called to indicate that the download has failed. 'why'
1297 is a Failure object indicating what went wrong. No further methods
1298 will be invoked on the IDownloadTarget after fail()."""
1299 def register_canceller(cb):
1300 """The CiphertextDownloader uses this to register a no-argument function
1301 that the target can call to cancel the download. Once this canceller
1302 is invoked, no further calls to write() or close() will be made."""
1304 """When the CiphertextDownloader is done, this finish() function will be
1305 called. Whatever it returns will be returned to the invoker of
1306 Downloader.download.
1308 # The following methods are just because that target might be a repairer.DownUpConnector,
1309 # and just because the current CHKUpload object expects to find the storage index and
1310 # encoding parameters in its Uploadable.
1311 def set_storageindex(storageindex):
1312 """ Set the storage index. """
1313 def set_encodingparams(encodingparams):
1314 """ Set the encoding parameters. """
1316 class IDownloader(Interface):
1317 def download(uri, target):
1318 """Perform a CHK download, sending the data to the given target.
1319 'target' must provide IDownloadTarget.
1321 Returns a Deferred that fires (with the results of target.finish)
1322 when the download is finished, or errbacks if something went wrong."""
1324 class IEncryptedUploadable(Interface):
1325 def set_upload_status(upload_status):
1326 """Provide an IUploadStatus object that should be filled with status
1327 information. The IEncryptedUploadable is responsible for setting
1328 key-determination progress ('chk'), size, storage_index, and
1329 ciphertext-fetch progress. It may delegate some of this
1330 responsibility to others, in particular to the IUploadable."""
1333 """This behaves just like IUploadable.get_size()."""
1335 def get_all_encoding_parameters():
1336 """Return a Deferred that fires with a tuple of
1337 (k,happy,n,segment_size). The segment_size will be used as-is, and
1338 must match the following constraints: it must be a multiple of k, and
1339 it shouldn't be unreasonably larger than the file size (if
1340 segment_size is larger than filesize, the difference must be stored
1343 This usually passes through to the IUploadable method of the same
1346 The encoder strictly obeys the values returned by this method. To
1347 make an upload use non-default encoding parameters, you must arrange
1348 to control the values that this method returns.
1351 def get_storage_index():
1352 """Return a Deferred that fires with a 16-byte storage index.
1355 def read_encrypted(length, hash_only):
1356 """This behaves just like IUploadable.read(), but returns crypttext
1357 instead of plaintext. If hash_only is True, then this discards the
1358 data (and returns an empty list); this improves efficiency when
1359 resuming an interrupted upload (where we need to compute the
1360 plaintext hashes, but don't need the redundant encrypted data)."""
1362 def get_plaintext_hashtree_leaves(first, last, num_segments):
1363 """OBSOLETE; Get the leaf nodes of a merkle hash tree over the plaintext
1364 segments, i.e. get the tagged hashes of the given segments. The
1365 segment size is expected to be generated by the IEncryptedUploadable
1366 before any plaintext is read or ciphertext produced, so that the
1367 segment hashes can be generated with only a single pass.
1369 This returns a Deferred which fires with a sequence of hashes, using:
1371 tuple(segment_hashes[first:last])
1373 'num_segments' is used to assert that the number of segments that the
1374 IEncryptedUploadable handled matches the number of segments that the
1375 encoder was expecting.
1377 This method must not be called until the final byte has been read
1378 from read_encrypted(). Once this method is called, read_encrypted()
1379 can never be called again.
1382 def get_plaintext_hash():
1383 """OBSOLETE; Get the hash of the whole plaintext.
1385 This returns a Deferred which fires with a tagged SHA-256 hash of the
1386 whole plaintext, obtained from hashutil.plaintext_hash(data).
1390 """Just like IUploadable.close()."""
1392 class IUploadable(Interface):
1393 def set_upload_status(upload_status):
1394 """Provide an IUploadStatus object that should be filled with status
1395 information. The IUploadable is responsible for setting
1396 key-determination progress ('chk')."""
1398 def set_default_encoding_parameters(params):
1399 """Set the default encoding parameters, which must be a dict mapping
1400 strings to ints. The meaningful keys are 'k', 'happy', 'n', and
1401 'max_segment_size'. These might have an influence on the final
1402 encoding parameters returned by get_all_encoding_parameters(), if the
1403 Uploadable doesn't have more specific preferences.
1405 This call is optional: if it is not used, the Uploadable will use
1406 some built-in defaults. If used, this method must be called before
1407 any other IUploadable methods to have any effect.
1411 """Return a Deferred that will fire with the length of the data to be
1412 uploaded, in bytes. This will be called before the data is actually
1413 used, to compute encoding parameters.
1416 def get_all_encoding_parameters():
1417 """Return a Deferred that fires with a tuple of
1418 (k,happy,n,segment_size). The segment_size will be used as-is, and
1419 must match the following constraints: it must be a multiple of k, and
1420 it shouldn't be unreasonably larger than the file size (if
1421 segment_size is larger than filesize, the difference must be stored
1424 The relative values of k and n allow some IUploadables to request
1425 better redundancy than others (in exchange for consuming more space
1428 Larger values of segment_size reduce hash overhead, while smaller
1429 values reduce memory footprint and cause data to be delivered in
1430 smaller pieces (which may provide a smoother and more predictable
1431 download experience).
1433 The encoder strictly obeys the values returned by this method. To
1434 make an upload use non-default encoding parameters, you must arrange
1435 to control the values that this method returns. One way to influence
1436 them may be to call set_encoding_parameters() before calling
1437 get_all_encoding_parameters().
1440 def get_encryption_key():
1441 """Return a Deferred that fires with a 16-byte AES key. This key will
1442 be used to encrypt the data. The key will also be hashed to derive
1445 Uploadables which want to achieve convergence should hash their file
1446 contents and the serialized_encoding_parameters to form the key
1447 (which of course requires a full pass over the data). Uploadables can
1448 use the upload.ConvergentUploadMixin class to achieve this
1451 Uploadables which do not care about convergence (or do not wish to
1452 make multiple passes over the data) can simply return a
1453 strongly-random 16 byte string.
1455 get_encryption_key() may be called multiple times: the IUploadable is
1456 required to return the same value each time.
1460 """Return a Deferred that fires with a list of strings (perhaps with
1461 only a single element) which, when concatenated together, contain the
1462 next 'length' bytes of data. If EOF is near, this may provide fewer
1463 than 'length' bytes. The total number of bytes provided by read()
1464 before it signals EOF must equal the size provided by get_size().
1466 If the data must be acquired through multiple internal read
1467 operations, returning a list instead of a single string may help to
1468 reduce string copies.
1470 'length' will typically be equal to (min(get_size(),1MB)/req_shares),
1471 so a 10kB file means length=3kB, 100kB file means length=30kB,
1472 and >=1MB file means length=300kB.
1474 This method provides for a single full pass through the data. Later
1475 use cases may desire multiple passes or access to only parts of the
1476 data (such as a mutable file making small edits-in-place). This API
1477 will be expanded once those use cases are better understood.
1481 """The upload is finished, and whatever filehandle was in use may be
1484 class IUploadResults(Interface):
1485 """I am returned by upload() methods. I contain a number of public
1486 attributes which can be read to determine the results of the upload. Some
1487 of these are functional, some are timing information. All of these may be
1490 .file_size : the size of the file, in bytes
1491 .uri : the CHK read-cap for the file
1492 .ciphertext_fetched : how many bytes were fetched by the helper
1493 .sharemap: dict mapping share identifier to set of serverids
1494 (binary strings). This indicates which servers were given
1495 which shares. For immutable files, the shareid is an
1496 integer (the share number, from 0 to N-1). For mutable
1497 files, it is a string of the form 'seq%d-%s-sh%d',
1498 containing the sequence number, the roothash, and the
1500 .servermap : dict mapping server peerid to a set of share numbers
1501 .timings : dict of timing information, mapping name to seconds (float)
1502 total : total upload time, start to finish
1503 storage_index : time to compute the storage index
1504 peer_selection : time to decide which peers will be used
1505 contacting_helper : initial helper query to upload/no-upload decision
1506 existence_check : helper pre-upload existence check
1507 helper_total : initial helper query to helper finished pushing
1508 cumulative_fetch : helper waiting for ciphertext requests
1509 total_fetch : helper start to last ciphertext response
1510 cumulative_encoding : just time spent in zfec
1511 cumulative_sending : just time spent waiting for storage servers
1512 hashes_and_close : last segment push to shareholder close
1513 total_encode_and_push : first encode to shareholder close
1517 class IDownloadResults(Interface):
1518 """I am created internally by download() methods. I contain a number of
1519 public attributes which contain details about the download process.::
1521 .file_size : the size of the file, in bytes
1522 .servers_used : set of server peerids that were used during download
1523 .server_problems : dict mapping server peerid to a problem string. Only
1524 servers that had problems (bad hashes, disconnects) are
1526 .servermap : dict mapping server peerid to a set of share numbers. Only
1527 servers that had any shares are listed here.
1528 .timings : dict of timing information, mapping name to seconds (float)
1529 peer_selection : time to ask servers about shares
1530 servers_peer_selection : dict of peerid to DYHB-query time
1531 uri_extension : time to fetch a copy of the URI extension block
1532 hashtrees : time to fetch the hash trees
1533 segments : time to fetch, decode, and deliver segments
1534 cumulative_fetch : time spent waiting for storage servers
1535 cumulative_decode : just time spent in zfec
1536 cumulative_decrypt : just time spent in decryption
1537 total : total download time, start to finish
1538 fetch_per_server : dict of peerid to list of per-segment fetch times
1542 class IUploader(Interface):
1543 def upload(uploadable):
1544 """Upload the file. 'uploadable' must impement IUploadable. This
1545 returns a Deferred which fires with an UploadResults instance, from
1546 which the URI of the file can be obtained as results.uri ."""
1548 def upload_ssk(write_capability, new_version, uploadable):
1549 """TODO: how should this work?"""
1551 class ICheckable(Interface):
1552 def check(monitor, verify=False):
1553 """Check upon my health, optionally repairing any problems.
1555 This returns a Deferred that fires with an instance that provides
1556 ICheckResults, or None if the object is non-distributed (i.e. LIT
1559 The monitor will be checked periodically to see if the operation has
1560 been cancelled. If so, no new queries will be sent, and the Deferred
1561 will fire (with a OperationCancelledError) immediately.
1563 Filenodes and dirnodes (which provide IFilesystemNode) are also
1564 checkable. Instances that represent verifier-caps will be checkable
1565 but not downloadable. Some objects (like LIT files) do not actually
1566 live in the grid, and their checkers return None (non-distributed
1567 files are always healthy).
1569 If verify=False, a relatively lightweight check will be performed: I
1570 will ask all servers if they have a share for me, and I will believe
1571 whatever they say. If there are at least N distinct shares on the
1572 grid, my results will indicate r.is_healthy()==True. This requires a
1573 roundtrip to each server, but does not transfer very much data, so
1574 the network bandwidth is fairly low.
1576 If verify=True, a more resource-intensive check will be performed:
1577 every share will be downloaded, and the hashes will be validated on
1578 every bit. I will ignore any shares that failed their hash checks. If
1579 there are at least N distinct valid shares on the grid, my results
1580 will indicate r.is_healthy()==True. This requires N/k times as much
1581 download bandwidth (and server disk IO) as a regular download. If a
1582 storage server is holding a corrupt share, or is experiencing memory
1583 failures during retrieval, or is malicious or buggy, then
1584 verification will detect the problem, but checking will not.
1586 TODO: any problems seen during checking will be reported to the
1587 health-manager.furl, a centralized object which is responsible for
1588 figuring out why files are unhealthy so corrective action can be
1592 def check_and_repair(monitor, verify=False):
1593 """Like check(), but if the file/directory is not healthy, attempt to
1596 Any non-healthy result will cause an immediate repair operation, to
1597 generate and upload new shares. After repair, the file will be as
1598 healthy as we can make it. Details about what sort of repair is done
1599 will be put in the check-and-repair results. The Deferred will not
1600 fire until the repair is complete.
1602 This returns a Deferred which fires with an instance of
1603 ICheckAndRepairResults."""
1605 class IDeepCheckable(Interface):
1606 def start_deep_check(verify=False):
1607 """Check upon the health of me and everything I can reach.
1609 This is a recursive form of check(), useable only on dirnodes.
1611 I return a Monitor, with results that are an IDeepCheckResults
1615 def start_deep_check_and_repair(verify=False):
1616 """Check upon the health of me and everything I can reach. Repair
1617 anything that isn't healthy.
1619 This is a recursive form of check_and_repair(), useable only on
1622 I return a Monitor, with results that are an
1623 IDeepCheckAndRepairResults object.
1626 class ICheckResults(Interface):
1627 """I contain the detailed results of a check/verify operation.
1630 def get_storage_index():
1631 """Return a string with the (binary) storage index."""
1632 def get_storage_index_string():
1633 """Return a string with the (printable) abbreviated storage index."""
1635 """Return the (string) URI of the object that was checked."""
1638 """Return a boolean, True if the file/dir is fully healthy, False if
1639 it is damaged in any way. Non-distributed LIT files always return
1642 def is_recoverable():
1643 """Return a boolean, True if the file/dir can be recovered, False if
1644 not. Unrecoverable files are obviously unhealthy. Non-distributed LIT
1645 files always return True."""
1647 def needs_rebalancing():
1648 """Return a boolean, True if the file/dir's reliability could be
1649 improved by moving shares to new servers. Non-distributed LIT files
1650 always return False."""
1654 """Return a dictionary that describes the state of the file/dir. LIT
1655 files always return an empty dictionary. Normal files and directories return a
1656 dictionary with the following keys (note that these use binary strings rather than
1657 base32-encoded ones) (also note that for mutable files, these counts are for the 'best'
1660 count-shares-good: the number of distinct good shares that were found
1661 count-shares-needed: 'k', the number of shares required for recovery
1662 count-shares-expected: 'N', the number of total shares generated
1663 count-good-share-hosts: the number of distinct storage servers with
1664 good shares. If this number is less than
1665 count-shares-good, then some shares are
1666 doubled up, increasing the correlation of
1667 failures. This indicates that one or more
1668 shares should be moved to an otherwise unused
1669 server, if one is available.
1670 count-corrupt-shares: the number of shares with integrity failures
1671 list-corrupt-shares: a list of 'share locators', one for each share
1672 that was found to be corrupt. Each share
1673 locator is a list of (serverid, storage_index,
1675 count-incompatible-shares: the number of shares which are of a share format unknown to
1677 list-incompatible-shares: a list of 'share locators', one for each share that was found
1678 to be of an unknown format. Each share locator is a list of
1679 (serverid, storage_index, sharenum).
1680 servers-responding: list of (binary) storage server identifiers,
1681 one for each server which responded to the share
1682 query (even if they said they didn't have shares,
1683 and even if they said they did have shares but then
1684 didn't send them when asked, or dropped the
1685 connection, or returned a Failure, and even if they
1686 said they did have shares and sent incorrect ones
1688 sharemap: dict mapping share identifier to list of serverids
1689 (binary strings). This indicates which servers are holding
1690 which shares. For immutable files, the shareid is an
1691 integer (the share number, from 0 to N-1). For mutable
1692 files, it is a string of the form 'seq%d-%s-sh%d',
1693 containing the sequence number, the roothash, and the
1696 The following keys are most relevant for mutable files, but immutable
1697 files will provide sensible values too::
1699 count-wrong-shares: the number of shares for versions other than the
1700 'best' one (which is defined as being the
1701 recoverable version with the highest sequence
1702 number, then the highest roothash). These are
1703 either leftover shares from an older version
1704 (perhaps on a server that was offline when an
1705 update occurred), shares from an unrecoverable
1706 newer version, or shares from an alternate
1707 current version that results from an
1708 uncoordinated write collision. For a healthy
1709 file, this will equal 0.
1711 count-recoverable-versions: the number of recoverable versions of
1712 the file. For a healthy file, this will
1715 count-unrecoverable-versions: the number of unrecoverable versions
1716 of the file. For a healthy file, this
1722 """Return a string with a brief (one-line) summary of the results."""
1725 """Return a list of strings with more detailed results."""
1727 class ICheckAndRepairResults(Interface):
1728 """I contain the detailed results of a check/verify/repair operation.
1730 The IFilesystemNode.check()/verify()/repair() methods all return
1731 instances that provide ICheckAndRepairResults.
1734 def get_storage_index():
1735 """Return a string with the (binary) storage index."""
1736 def get_storage_index_string():
1737 """Return a string with the (printable) abbreviated storage index."""
1738 def get_repair_attempted():
1739 """Return a boolean, True if a repair was attempted."""
1740 def get_repair_successful():
1741 """Return a boolean, True if repair was attempted and the file/dir
1742 was fully healthy afterwards. False if no repair was attempted or if
1743 a repair attempt failed."""
1744 def get_pre_repair_results():
1745 """Return an ICheckResults instance that describes the state of the
1746 file/dir before any repair was attempted."""
1747 def get_post_repair_results():
1748 """Return an ICheckResults instance that describes the state of the
1749 file/dir after any repair was attempted. If no repair was attempted,
1750 the pre-repair and post-repair results will be identical."""
1753 class IDeepCheckResults(Interface):
1754 """I contain the results of a deep-check operation.
1756 This is returned by a call to ICheckable.deep_check().
1759 def get_root_storage_index_string():
1760 """Return the storage index (abbreviated human-readable string) of
1761 the first object checked."""
1763 """Return a dictionary with the following keys::
1765 count-objects-checked: count of how many objects were checked
1766 count-objects-healthy: how many of those objects were completely
1768 count-objects-unhealthy: how many were damaged in some way
1769 count-objects-unrecoverable: how many were unrecoverable
1770 count-corrupt-shares: how many shares were found to have
1771 corruption, summed over all objects
1775 def get_corrupt_shares():
1776 """Return a set of (serverid, storage_index, sharenum) for all shares
1777 that were found to be corrupt. Both serverid and storage_index are
1780 def get_all_results():
1781 """Return a dictionary mapping pathname (a tuple of strings, ready to
1782 be slash-joined) to an ICheckResults instance, one for each object
1783 that was checked."""
1785 def get_results_for_storage_index(storage_index):
1786 """Retrive the ICheckResults instance for the given (binary)
1787 storage index. Raises KeyError if there are no results for that
1791 """Return a dictionary with the same keys as
1792 IDirectoryNode.deep_stats()."""
1794 class IDeepCheckAndRepairResults(Interface):
1795 """I contain the results of a deep-check-and-repair operation.
1797 This is returned by a call to ICheckable.deep_check_and_repair().
1800 def get_root_storage_index_string():
1801 """Return the storage index (abbreviated human-readable string) of
1802 the first object checked."""
1804 """Return a dictionary with the following keys::
1806 count-objects-checked: count of how many objects were checked
1807 count-objects-healthy-pre-repair: how many of those objects were
1808 completely healthy (before any
1810 count-objects-unhealthy-pre-repair: how many were damaged in
1812 count-objects-unrecoverable-pre-repair: how many were unrecoverable
1813 count-objects-healthy-post-repair: how many of those objects were
1814 completely healthy (after any
1816 count-objects-unhealthy-post-repair: how many were damaged in
1818 count-objects-unrecoverable-post-repair: how many were
1820 count-repairs-attempted: repairs were attempted on this many
1821 objects. The count-repairs- keys will
1822 always be provided, however unless
1823 repair=true is present, they will all
1825 count-repairs-successful: how many repairs resulted in healthy
1827 count-repairs-unsuccessful: how many repairs resulted did not
1828 results in completely healthy objects
1829 count-corrupt-shares-pre-repair: how many shares were found to
1830 have corruption, summed over all
1831 objects examined (before any
1833 count-corrupt-shares-post-repair: how many shares were found to
1834 have corruption, summed over all
1835 objects examined (after any
1840 """Return a dictionary with the same keys as
1841 IDirectoryNode.deep_stats()."""
1843 def get_corrupt_shares():
1844 """Return a set of (serverid, storage_index, sharenum) for all shares
1845 that were found to be corrupt before any repair was attempted. Both
1846 serverid and storage_index are binary.
1848 def get_remaining_corrupt_shares():
1849 """Return a set of (serverid, storage_index, sharenum) for all shares
1850 that were found to be corrupt after any repair was completed. Both
1851 serverid and storage_index are binary. These are shares that need
1852 manual inspection and probably deletion.
1854 def get_all_results():
1855 """Return a dictionary mapping pathname (a tuple of strings, ready to
1856 be slash-joined) to an ICheckAndRepairResults instance, one for each
1857 object that was checked."""
1859 def get_results_for_storage_index(storage_index):
1860 """Retrive the ICheckAndRepairResults instance for the given (binary)
1861 storage index. Raises KeyError if there are no results for that
1865 class IRepairable(Interface):
1866 def repair(check_results):
1867 """Attempt to repair the given object. Returns a Deferred that fires
1868 with a IRepairResults object.
1870 I must be called with an object that implements ICheckResults, as
1871 proof that you have actually discovered a problem with this file. I
1872 will use the data in the checker results to guide the repair process,
1873 such as which servers provided bad data and should therefore be
1874 avoided. The ICheckResults object is inside the
1875 ICheckAndRepairResults object, which is returned by the
1876 ICheckable.check() method::
1878 d = filenode.check(repair=False)
1879 def _got_results(check_and_repair_results):
1880 check_results = check_and_repair_results.get_pre_repair_results()
1881 return filenode.repair(check_results)
1882 d.addCallback(_got_results)
1886 class IRepairResults(Interface):
1887 """I contain the results of a repair operation."""
1890 class IClient(Interface):
1891 def upload(uploadable):
1892 """Upload some data into a CHK, get back the UploadResults for it.
1893 @param uploadable: something that implements IUploadable
1894 @return: a Deferred that fires with the UploadResults instance.
1895 To get the URI for this file, use results.uri .
1898 def create_mutable_file(contents=""):
1899 """Create a new mutable file with contents, get back the URI string.
1900 @param contents: the initial contents to place in the file.
1901 @return: a Deferred that fires with tne (string) SSK URI for the new
1905 def create_empty_dirnode():
1906 """Create a new dirnode, empty and unattached.
1907 @return: a Deferred that fires with the new IDirectoryNode instance.
1910 def create_node_from_uri(uri):
1911 """Create a new IFilesystemNode instance from the uri, synchronously.
1912 @param uri: a string or IURI-providing instance. This could be for a
1913 LiteralFileNode, a CHK file node, a mutable file node, or
1915 @return: an instance that provides IFilesystemNode (or more usefully one
1916 of its subclasses). File-specifying URIs will result in
1917 IFileNode or IMutableFileNode -providing instances, like
1918 FileNode, LiteralFileNode, or MutableFileNode.
1919 Directory-specifying URIs will result in
1920 IDirectoryNode-providing instances, like NewDirectoryNode.
1923 class IClientStatus(Interface):
1924 def list_all_uploads():
1925 """Return a list of uploader objects, one for each upload which
1926 currently has an object available (tracked with weakrefs). This is
1927 intended for debugging purposes."""
1928 def list_active_uploads():
1929 """Return a list of active IUploadStatus objects."""
1930 def list_recent_uploads():
1931 """Return a list of IUploadStatus objects for the most recently
1934 def list_all_downloads():
1935 """Return a list of downloader objects, one for each download which
1936 currently has an object available (tracked with weakrefs). This is
1937 intended for debugging purposes."""
1938 def list_active_downloads():
1939 """Return a list of active IDownloadStatus objects."""
1940 def list_recent_downloads():
1941 """Return a list of IDownloadStatus objects for the most recently
1942 started downloads."""
1944 class IUploadStatus(Interface):
1946 """Return a timestamp (float with seconds since epoch) indicating
1947 when the operation was started."""
1948 def get_storage_index():
1949 """Return a string with the (binary) storage index in use on this
1950 upload. Returns None if the storage index has not yet been
1953 """Return an integer with the number of bytes that will eventually
1954 be uploaded for this file. Returns None if the size is not yet known.
1957 """Return True if this upload is using a Helper, False if not."""
1959 """Return a string describing the current state of the upload
1962 """Returns a tuple of floats, (chk, ciphertext, encode_and_push),
1963 each from 0.0 to 1.0 . 'chk' describes how much progress has been
1964 made towards hashing the file to determine a CHK encryption key: if
1965 non-convergent encryption is in use, this will be trivial, otherwise
1966 the whole file must be hashed. 'ciphertext' describes how much of the
1967 ciphertext has been pushed to the helper, and is '1.0' for non-helper
1968 uploads. 'encode_and_push' describes how much of the encode-and-push
1969 process has finished: for helper uploads this is dependent upon the
1970 helper providing progress reports. It might be reasonable to add all
1971 three numbers and report the sum to the user."""
1973 """Return True if the upload is currently active, False if not."""
1975 """Return an instance of UploadResults (which contains timing and
1976 sharemap information). Might return None if the upload is not yet
1979 """Each upload status gets a unique number: this method returns that
1980 number. This provides a handle to this particular upload, so a web
1981 page can generate a suitable hyperlink."""
1983 class IDownloadStatus(Interface):
1985 """Return a timestamp (float with seconds since epoch) indicating
1986 when the operation was started."""
1987 def get_storage_index():
1988 """Return a string with the (binary) storage index in use on this
1989 download. This may be None if there is no storage index (i.e. LIT
1992 """Return an integer with the number of bytes that will eventually be
1993 retrieved for this file. Returns None if the size is not yet known.
1996 """Return True if this download is using a Helper, False if not."""
1998 """Return a string describing the current state of the download
2001 """Returns a float (from 0.0 to 1.0) describing the amount of the
2002 download that has completed. This value will remain at 0.0 until the
2003 first byte of plaintext is pushed to the download target."""
2005 """Return True if the download is currently active, False if not."""
2007 """Each download status gets a unique number: this method returns
2008 that number. This provides a handle to this particular download, so a
2009 web page can generate a suitable hyperlink."""
2011 class IServermapUpdaterStatus(Interface):
2013 class IPublishStatus(Interface):
2015 class IRetrieveStatus(Interface):
2018 class NotCapableError(Exception):
2019 """You have tried to write to a read-only node."""
2021 class BadWriteEnablerError(Exception):
2024 class RIControlClient(RemoteInterface):
2026 def wait_for_client_connections(num_clients=int):
2027 """Do not return until we have connections to at least NUM_CLIENTS
2031 def upload_from_file_to_uri(filename=str, convergence=ChoiceOf(None, StringConstraint(2**20))):
2032 """Upload a file to the grid. This accepts a filename (which must be
2033 absolute) that points to a file on the node's local disk. The node will
2034 read the contents of this file, upload it to the grid, then return the
2035 URI at which it was uploaded. If convergence is None then a random
2036 encryption key will be used, else the plaintext will be hashed, then
2037 that hash will be mixed together with the "convergence" string to form
2042 def download_from_uri_to_file(uri=URI, filename=str):
2043 """Download a file from the grid, placing it on the node's local disk
2044 at the given filename (which must be absolute[?]). Returns the
2045 absolute filename where the file was written."""
2050 def get_memory_usage():
2051 """Return a dict describes the amount of memory currently in use. The
2052 keys are 'VmPeak', 'VmSize', and 'VmData'. The values are integers,
2053 measuring memory consupmtion in bytes."""
2054 return DictOf(str, int)
2056 def speed_test(count=int, size=int, mutable=Any()):
2057 """Write 'count' tempfiles to disk, all of the given size. Measure
2058 how long (in seconds) it takes to upload them all to the servers.
2059 Then measure how long it takes to download all of them. If 'mutable'
2060 is 'create', time creation of mutable files. If 'mutable' is
2061 'upload', then time access to the same mutable file instead of
2064 Returns a tuple of (upload_time, download_time).
2066 return (float, float)
2068 def measure_peer_response_time():
2069 """Send a short message to each connected peer, and measure the time
2070 it takes for them to respond to it. This is a rough measure of the
2071 application-level round trip time.
2073 @return: a dictionary mapping peerid to a float (RTT time in seconds)
2076 return DictOf(Nodeid, float)
2078 UploadResults = Any() #DictOf(str, str)
2080 class RIEncryptedUploadable(RemoteInterface):
2081 __remote_name__ = "RIEncryptedUploadable.tahoe.allmydata.com"
2086 def get_all_encoding_parameters():
2087 return (int, int, int, long)
2089 def read_encrypted(offset=Offset, length=ReadSize):
2092 def get_plaintext_hashtree_leaves(first=int, last=int, num_segments=int):
2095 def get_plaintext_hash():
2102 class RICHKUploadHelper(RemoteInterface):
2103 __remote_name__ = "RIUploadHelper.tahoe.allmydata.com"
2107 Return a dictionary of version information.
2109 return DictOf(str, Any())
2111 def upload(reader=RIEncryptedUploadable):
2112 return UploadResults
2115 class RIHelper(RemoteInterface):
2116 __remote_name__ = "RIHelper.tahoe.allmydata.com"
2120 Return a dictionary of version information.
2122 return DictOf(str, Any())
2124 def upload_chk(si=StorageIndex):
2125 """See if a file with a given storage index needs uploading. The
2126 helper will ask the appropriate storage servers to see if the file
2127 has already been uploaded. If so, the helper will return a set of
2128 'upload results' that includes whatever hashes are needed to build
2129 the read-cap, and perhaps a truncated sharemap.
2131 If the file has not yet been uploaded (or if it was only partially
2132 uploaded), the helper will return an empty upload-results dictionary
2133 and also an RICHKUploadHelper object that will take care of the
2134 upload process. The client should call upload() on this object and
2135 pass it a reference to an RIEncryptedUploadable object that will
2136 provide ciphertext. When the upload is finished, the upload() method
2137 will finish and return the upload results.
2139 return (UploadResults, ChoiceOf(RICHKUploadHelper, None))
2142 class RIStatsProvider(RemoteInterface):
2143 __remote_name__ = "RIStatsProvider.tahoe.allmydata.com"
2145 Provides access to statistics and monitoring information.
2150 returns a dictionary containing 'counters' and 'stats', each a dictionary
2151 with string counter/stat name keys, and numeric values. counters are
2152 monotonically increasing measures of work done, and stats are instantaneous
2153 measures (potentially time averaged internally)
2155 return DictOf(str, DictOf(str, ChoiceOf(float, int, long)))
2157 class RIStatsGatherer(RemoteInterface):
2158 __remote_name__ = "RIStatsGatherer.tahoe.allmydata.com"
2160 Provides a monitoring service for centralised collection of stats
2163 def provide(provider=RIStatsProvider, nickname=str):
2165 @param provider: a stats collector instance which should be polled
2166 periodically by the gatherer to collect stats.
2167 @param nickname: a name useful to identify the provided client
2172 class IStatsProducer(Interface):
2175 returns a dictionary, with str keys representing the names of stats
2176 to be monitored, and numeric values.
2179 class RIKeyGenerator(RemoteInterface):
2180 __remote_name__ = "RIKeyGenerator.tahoe.allmydata.com"
2182 Provides a service offering to make RSA key pairs.
2185 def get_rsa_key_pair(key_size=int):
2187 @param key_size: the size of the signature key.
2188 @return: tuple(verifying_key, signing_key)
2190 return TupleOf(str, str)
2193 class FileTooLargeError(Exception):
2196 class IValidatedThingProxy(Interface):
2198 """ Acquire a thing and validate it. Return a deferred which is eventually fired with
2199 self if the thing is valid or errbacked if it can't be acquired or validated. """
2201 class InsufficientVersionError(Exception):
2202 def __init__(self, needed, got):
2203 self.needed = needed
2206 return "InsufficientVersionError(need '%s', got %s)" % (self.needed,