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 is what we hereby agree to accept. New
115 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.
129 def renew_lease(storage_index=StorageIndex, renew_secret=LeaseRenewSecret):
131 Renew the lease on a given bucket. Some networks will use this, some
136 def cancel_lease(storage_index=StorageIndex,
137 cancel_secret=LeaseCancelSecret):
139 Cancel the lease on a given bucket. If this was the last lease on the
140 bucket, the bucket will be deleted.
144 def get_buckets(storage_index=StorageIndex):
145 return DictOf(int, RIBucketReader, maxKeys=MAX_BUCKETS)
149 def slot_readv(storage_index=StorageIndex,
150 shares=ListOf(int), readv=ReadVector):
151 """Read a vector from the numbered shares associated with the given
152 storage index. An empty shares list means to return data from all
153 known shares. Returns a dictionary with one key per share."""
154 return DictOf(int, ReadData) # shnum -> results
156 def slot_testv_and_readv_and_writev(storage_index=StorageIndex,
157 secrets=TupleOf(WriteEnablerSecret,
160 tw_vectors=TestAndWriteVectorsForShares,
163 """General-purpose test-and-set operation for mutable slots. Perform
164 a bunch of comparisons against the existing shares. If they all pass,
165 then apply a bunch of write vectors to those shares. Then use the
166 read vectors to extract data from all the shares and return the data.
168 This method is, um, large. The goal is to allow clients to update all
169 the shares associated with a mutable file in a single round trip.
171 @param storage_index: the index of the bucket to be created or
173 @param write_enabler: a secret that is stored along with the slot.
174 Writes are accepted from any caller who can
175 present the matching secret. A different secret
176 should be used for each slot*server pair.
177 @param renew_secret: This is the secret used to protect bucket refresh
178 This secret is generated by the client and
179 stored for later comparison by the server. Each
180 server is given a different secret.
181 @param cancel_secret: Like renew_secret, but protects bucket decref.
183 The 'secrets' argument is a tuple of (write_enabler, renew_secret,
184 cancel_secret). The first is required to perform any write. The
185 latter two are used when allocating new shares. To simply acquire a
186 new lease on existing shares, use an empty testv and an empty writev.
188 Each share can have a separate test vector (i.e. a list of
189 comparisons to perform). If all vectors for all shares pass, then all
190 writes for all shares are recorded. Each comparison is a 4-tuple of
191 (offset, length, operator, specimen), which effectively does a bool(
192 (read(offset, length)) OPERATOR specimen ) and only performs the
193 write if all these evaluate to True. Basic test-and-set uses 'eq'.
194 Write-if-newer uses a seqnum and (offset, length, 'lt', specimen).
195 Write-if-same-or-newer uses 'le'.
197 Reads from the end of the container are truncated, and missing shares
198 behave like empty ones, so to assert that a share doesn't exist (for
199 use when creating a new share), use (0, 1, 'eq', '').
201 The write vector will be applied to the given share, expanding it if
202 necessary. A write vector applied to a share number that did not
203 exist previously will cause that share to be created.
205 Each write vector is accompanied by a 'new_length' argument. If
206 new_length is not None, use it to set the size of the container. This
207 can be used to pre-allocate space for a series of upcoming writes, or
208 truncate existing data. If the container is growing, new_length will
209 be applied before datav. If the container is shrinking, it will be
210 applied afterwards. If new_length==0, the share will be deleted.
212 The read vector is used to extract data from all known shares,
213 *before* any writes have been applied. The same vector is used for
214 all shares. This captures the state that was tested by the test
217 This method returns two values: a boolean and a dict. The boolean is
218 True if the write vectors were applied, False if not. The dict is
219 keyed by share number, and each value contains a list of strings, one
220 for each element of the read vector.
222 If the write_enabler is wrong, this will raise BadWriteEnablerError.
223 To enable share migration (using update_write_enabler), the exception
224 will have the nodeid used for the old write enabler embedded in it,
225 in the following string::
227 The write enabler was recorded by nodeid '%s'.
229 Note that the nodeid here is encoded using the same base32 encoding
230 used by Foolscap and allmydata.util.idlib.nodeid_b2a().
233 return TupleOf(bool, DictOf(int, ReadData))
235 def advise_corrupt_share(share_type=str, storage_index=StorageIndex,
236 shnum=int, reason=str):
237 """Clients who discover hash failures in shares that they have
238 downloaded from me will use this method to inform me about the
239 failures. I will record their concern so that my operator can
240 manually inspect the shares in question. I return None.
242 'share_type' is either 'mutable' or 'immutable'. 'storage_index' is a
243 (binary) storage index string, and 'shnum' is the integer share
244 number. 'reason' is a human-readable explanation of the problem,
245 probably including some expected hash values and the computed ones
246 which did not match. Corruption advisories for mutable shares should
247 include a hash of the public key (the same value that appears in the
248 mutable-file verify-cap), since the current share format does not
252 class IStorageBucketWriter(Interface):
254 Objects of this kind live on the client side.
256 def put_block(segmentnum=int, data=ShareData):
257 """@param data: For most segments, this data will be 'blocksize'
258 bytes in length. The last segment might be shorter.
259 @return: a Deferred that fires (with None) when the operation completes
262 def put_plaintext_hashes(hashes=ListOf(Hash)):
264 @return: a Deferred that fires (with None) when the operation completes
267 def put_crypttext_hashes(hashes=ListOf(Hash)):
269 @return: a Deferred that fires (with None) when the operation completes
272 def put_block_hashes(blockhashes=ListOf(Hash)):
274 @return: a Deferred that fires (with None) when the operation completes
277 def put_share_hashes(sharehashes=ListOf(TupleOf(int, Hash))):
279 @return: a Deferred that fires (with None) when the operation completes
282 def put_uri_extension(data=URIExtensionData):
283 """This block of data contains integrity-checking information (hashes
284 of plaintext, crypttext, and shares), as well as encoding parameters
285 that are necessary to recover the data. This is a serialized dict
286 mapping strings to other strings. The hash of this data is kept in
287 the URI and verified before any of the data is used. All buckets for
288 a given file contain identical copies of this data.
290 The serialization format is specified with the following pseudocode:
291 for k in sorted(dict.keys()):
292 assert re.match(r'^[a-zA-Z_\-]+$', k)
293 write(k + ':' + netstring(dict[k]))
295 @return: a Deferred that fires (with None) when the operation completes
299 """Finish writing and close the bucket. The share is not finalized
300 until this method is called: if the uploading client disconnects
301 before calling close(), the partially-written share will be
304 @return: a Deferred that fires (with None) when the operation completes
307 class IStorageBucketReader(Interface):
309 def get_block_data(blocknum=int, blocksize=int, size=int):
310 """Most blocks will be the same size. The last block might be shorter
316 def get_crypttext_hashes():
318 @return: ListOf(Hash)
321 def get_block_hashes(at_least_these=SetOf(int)):
323 @return: ListOf(Hash)
326 def get_share_hashes(at_least_these=SetOf(int)):
328 @return: ListOf(TupleOf(int, Hash))
331 def get_uri_extension():
333 @return: URIExtensionData
338 # hm, we need a solution for forward references in schemas
339 FileNode_ = Any() # TODO: foolscap needs constraints on copyables
340 DirectoryNode_ = Any() # TODO: same
341 AnyNode_ = ChoiceOf(FileNode_, DirectoryNode_)
344 class IURI(Interface):
345 def init_from_string(uri):
346 """Accept a string (as created by my to_string() method) and populate
347 this instance with its data. I am not normally called directly,
348 please use the module-level uri.from_string() function to convert
349 arbitrary URI strings into IURI-providing instances."""
352 """Return False if this URI be used to modify the data. Return True
353 if this URI cannot be used to modify the data."""
356 """Return True if the data can be modified by *somebody* (perhaps
357 someone who has a more powerful URI than this one)."""
360 """Return another IURI instance, which represents a read-only form of
361 this one. If is_readonly() is True, this returns self."""
363 def get_verify_cap():
364 """Return an instance that provides IVerifierURI, which can be used
365 to check on the availability of the file or directory, without
366 providing enough capabilities to actually read or modify the
367 contents. This may return None if the file does not need checking or
368 verification (e.g. LIT URIs).
372 """Return a string of printable ASCII characters, suitable for
373 passing into init_from_string."""
375 class IVerifierURI(Interface, IURI):
376 def init_from_string(uri):
377 """Accept a string (as created by my to_string() method) and populate
378 this instance with its data. I am not normally called directly,
379 please use the module-level uri.from_string() function to convert
380 arbitrary URI strings into IURI-providing instances."""
383 """Return a string of printable ASCII characters, suitable for
384 passing into init_from_string."""
386 class IDirnodeURI(Interface):
387 """I am a URI which represents a dirnode."""
390 class IFileURI(Interface):
391 """I am a URI which represents a filenode."""
393 """Return the length (in bytes) of the file that I represent."""
395 class IImmutableFileURI(IFileURI):
398 class IMutableFileURI(Interface):
399 """I am a URI which represents a mutable filenode."""
400 class INewDirectoryURI(Interface):
402 class IReadonlyNewDirectoryURI(Interface):
406 class IFilesystemNode(Interface):
409 Return the URI that can be used by others to get access to this
410 node. If this node is read-only, the URI will only offer read-only
411 access. If this node is read-write, the URI will offer read-write
414 If you have read-write access to a node and wish to share merely
415 read-only access with others, use get_readonly_uri().
418 def get_readonly_uri():
419 """Return the directory URI that can be used by others to get
420 read-only access to this directory node. The result is a read-only
421 URI, regardless of whether this dirnode is read-only or read-write.
423 If you have merely read-only access to this dirnode,
424 get_readonly_uri() will return the same thing as get_uri().
427 def get_repair_cap():
428 """Return an IURI instance that can be used to repair the file, or
429 None if this node cannot be repaired (either because it is not
430 distributed, like a LIT file, or because the node does not represent
431 sufficient authority to create a repair-cap, like a read-only RSA
432 mutable file node [which cannot create the correct write-enablers]).
435 def get_verify_cap():
436 """Return an IVerifierURI instance that represents the
437 'verifiy/refresh capability' for this node. The holder of this
438 capability will be able to renew the lease for this node, protecting
439 it from garbage-collection. They will also be able to ask a server if
440 it holds a share for the file or directory.
443 def get_storage_index():
444 """Return a string with the (binary) storage index in use on this
445 download. This may be None if there is no storage index (i.e. LIT
449 """Return True if this reference provides mutable access to the given
450 file or directory (i.e. if you can modify it), or False if not. Note
451 that even if this reference is read-only, someone else may hold a
452 read-write reference to it."""
455 """Return True if this file or directory is mutable (by *somebody*,
456 not necessarily you), False if it is is immutable. Note that a file
457 might be mutable overall, but your reference to it might be
458 read-only. On the other hand, all references to an immutable file
459 will be read-only; there are no read-write references to an immutable
463 class IMutableFilesystemNode(IFilesystemNode):
466 class IFileNode(IFilesystemNode):
467 def download(target):
468 """Download the file's contents to a given IDownloadTarget"""
470 def download_to_data():
471 """Download the file's contents. Return a Deferred that fires
472 with those contents."""
475 """Return the length (in bytes) of the data this node represents."""
477 def read(consumer, offset=0, size=None):
478 """Download a portion (possibly all) of the file's contents, making
479 them available to the given IConsumer. Return a Deferred that fires
480 (with the consumer) when the consumer is unregistered (either because
481 the last byte has been given to it, or because the consumer threw an
482 exception during write(), possibly because it no longer wants to
483 receive data). The portion downloaded will start at 'offset' and
484 contain 'size' bytes (or the remainder of the file if size==None).
486 The consumer will be used in non-streaming mode: an IPullProducer
487 will be attached to it.
489 The consumer will not receive data right away: several network trips
490 must occur first. The order of events will be::
492 consumer.registerProducer(p, streaming)
493 (if streaming == False)::
494 consumer does p.resumeProducing()
496 consumer does p.resumeProducing()
497 consumer.write(data).. (repeat until all data is written)
498 consumer.unregisterProducer()
499 deferred.callback(consumer)
501 If a download error occurs, or an exception is raised by
502 consumer.registerProducer() or consumer.write(), I will call
503 consumer.unregisterProducer() and then deliver the exception via
504 deferred.errback(). To cancel the download, the consumer should call
505 p.stopProducing(), which will result in an exception being delivered
506 via deferred.errback().
508 A simple download-to-memory consumer example would look like this::
510 class MemoryConsumer:
511 implements(IConsumer)
515 def registerProducer(self, p, streaming):
516 assert streaming == False
519 def write(self, data):
520 self.chunks.append(data)
521 def unregisterProducer(self):
523 d = filenode.read(MemoryConsumer())
524 d.addCallback(lambda mc: "".join(mc.chunks))
529 class IMutableFileNode(IFileNode, IMutableFilesystemNode):
530 """I provide access to a 'mutable file', which retains its identity
531 regardless of what contents are put in it.
533 The consistency-vs-availability problem means that there might be
534 multiple versions of a file present in the grid, some of which might be
535 unrecoverable (i.e. have fewer than 'k' shares). These versions are
536 loosely ordered: each has a sequence number and a hash, and any version
537 with seqnum=N was uploaded by a node which has seen at least one version
540 The 'servermap' (an instance of IMutableFileServerMap) is used to
541 describe the versions that are known to be present in the grid, and which
542 servers are hosting their shares. It is used to represent the 'state of
543 the world', and is used for this purpose by my test-and-set operations.
544 Downloading the contents of the mutable file will also return a
545 servermap. Uploading a new version into the mutable file requires a
546 servermap as input, and the semantics of the replace operation is
547 'replace the file with my new version if it looks like nobody else has
548 changed the file since my previous download'. Because the file is
549 distributed, this is not a perfect test-and-set operation, but it will do
550 its best. If the replace process sees evidence of a simultaneous write,
551 it will signal an UncoordinatedWriteError, so that the caller can take
555 Most readers will want to use the 'best' current version of the file, and
556 should use my 'download_best_version()' method.
558 To unconditionally replace the file, callers should use overwrite(). This
559 is the mode that user-visible mutable files will probably use.
561 To apply some delta to the file, call modify() with a callable modifier
562 function that can apply the modification that you want to make. This is
563 the mode that dirnodes will use, since most directory modification
564 operations can be expressed in terms of deltas to the directory state.
567 Three methods are available for users who need to perform more complex
568 operations. The first is get_servermap(), which returns an up-to-date
569 servermap using a specified mode. The second is download_version(), which
570 downloads a specific version (not necessarily the 'best' one). The third
571 is 'upload', which accepts new contents and a servermap (which must have
572 been updated with MODE_WRITE). The upload method will attempt to apply
573 the new contents as long as no other node has modified the file since the
574 servermap was updated. This might be useful to a caller who wants to
575 merge multiple versions into a single new one.
577 Note that each time the servermap is updated, a specific 'mode' is used,
578 which determines how many peers are queried. To use a servermap for my
579 replace() method, that servermap must have been updated in MODE_WRITE.
580 These modes are defined in allmydata.mutable.common, and consist of
581 MODE_READ, MODE_WRITE, MODE_ANYTHING, and MODE_CHECK. Please look in
582 allmydata/mutable/servermap.py for details about the differences.
584 Mutable files are currently limited in size (about 3.5MB max) and can
585 only be retrieved and updated all-at-once, as a single big string. Future
586 versions of our mutable files will remove this restriction.
589 def download_best_version():
590 """Download the 'best' available version of the file, meaning one of
591 the recoverable versions with the highest sequence number. If no
592 uncoordinated writes have occurred, and if enough shares are
593 available, then this will be the most recent version that has been
596 I update an internal servermap with MODE_READ, determine which
597 version of the file is indicated by
598 servermap.best_recoverable_version(), and return a Deferred that
599 fires with its contents. If no version is recoverable, the Deferred
600 will errback with UnrecoverableFileError.
603 def get_size_of_best_version():
604 """Find the size of the version that would be downloaded with
605 download_best_version(), without actually downloading the whole file.
607 I return a Deferred that fires with an integer.
610 def overwrite(new_contents):
611 """Unconditionally replace the contents of the mutable file with new
612 ones. This simply chains get_servermap(MODE_WRITE) and upload(). This
613 is only appropriate to use when the new contents of the file are
614 completely unrelated to the old ones, and you do not care about other
617 I return a Deferred that fires (with a PublishStatus object) when the
618 update has completed.
621 def modify(modifier_cb):
622 """Modify the contents of the file, by downloading the current
623 version, applying the modifier function (or bound method), then
624 uploading the new version. I return a Deferred that fires (with a
625 PublishStatus object) when the update is complete.
627 The modifier callable will be given three arguments: a string (with
628 the old contents), a 'first_time' boolean, and a servermap. As with
629 download_best_version(), the old contents will be from the best
630 recoverable version, but the modifier can use the servermap to make
631 other decisions (such as refusing to apply the delta if there are
632 multiple parallel versions, or if there is evidence of a newer
633 unrecoverable version). 'first_time' will be True the first time the
634 modifier is called, and False on any subsequent calls.
636 The callable should return a string with the new contents. The
637 callable must be prepared to be called multiple times, and must
638 examine the input string to see if the change that it wants to make
639 is already present in the old version. If it does not need to make
640 any changes, it can either return None, or return its input string.
642 If the modifier raises an exception, it will be returned in the
647 def get_servermap(mode):
648 """Return a Deferred that fires with an IMutableFileServerMap
649 instance, updated using the given mode.
652 def download_version(servermap, version):
653 """Download a specific version of the file, using the servermap
654 as a guide to where the shares are located.
656 I return a Deferred that fires with the requested contents, or
657 errbacks with UnrecoverableFileError. Note that a servermap which was
658 updated with MODE_ANYTHING or MODE_READ may not know about shares for
659 all versions (those modes stop querying servers as soon as they can
660 fulfil their goals), so you may want to use MODE_CHECK (which checks
661 everything) to get increased visibility.
664 def upload(new_contents, servermap):
665 """Replace the contents of the file with new ones. This requires a
666 servermap that was previously updated with MODE_WRITE.
668 I attempt to provide test-and-set semantics, in that I will avoid
669 modifying any share that is different than the version I saw in the
670 servermap. However, if another node is writing to the file at the
671 same time as me, I may manage to update some shares while they update
672 others. If I see any evidence of this, I will signal
673 UncoordinatedWriteError, and the file will be left in an inconsistent
674 state (possibly the version you provided, possibly the old version,
675 possibly somebody else's version, and possibly a mix of shares from
678 The recommended response to UncoordinatedWriteError is to either
679 return it to the caller (since they failed to coordinate their
680 writes), or to attempt some sort of recovery. It may be sufficient to
681 wait a random interval (with exponential backoff) and repeat your
682 operation. If I do not signal UncoordinatedWriteError, then I was
683 able to write the new version without incident.
685 I return a Deferred that fires (with a PublishStatus object) when the
686 publish has completed. I will update the servermap in-place with the
687 location of all new shares.
691 """Return this filenode's writekey, or None if the node does not have
692 write-capability. This may be used to assist with data structures
693 that need to make certain data available only to writers, such as the
694 read-write child caps in dirnodes. The recommended process is to have
695 reader-visible data be submitted to the filenode in the clear (where
696 it will be encrypted by the filenode using the readkey), but encrypt
697 writer-visible data using this writekey.
700 class NotEnoughSharesError(Exception):
703 class ExistingChildError(Exception):
704 """A directory node was asked to add or replace a child that already
705 exists, and overwrite= was set to False."""
707 class NoSuchChildError(Exception):
708 """A directory node was asked to fetch a child which does not exist."""
710 class IDirectoryNode(IMutableFilesystemNode):
711 """I represent a name-to-child mapping, holding the tahoe equivalent of a
712 directory. All child names are unicode strings, and all children are some
713 sort of IFilesystemNode (either files or subdirectories).
718 The dirnode ('1') URI returned by this method can be used in
719 set_uri() on a different directory ('2') to 'mount' a reference to
720 this directory ('1') under the other ('2'). This URI is just a
721 string, so it can be passed around through email or other out-of-band
725 def get_readonly_uri():
727 The dirnode ('1') URI returned by this method can be used in
728 set_uri() on a different directory ('2') to 'mount' a reference to
729 this directory ('1') under the other ('2'). This URI is just a
730 string, so it can be passed around through email or other out-of-band
735 """I return a Deferred that fires with a dictionary mapping child
736 name (a unicode string) to (node, metadata_dict) tuples, in which
737 'node' is either an IFileNode or IDirectoryNode, and 'metadata_dict'
738 is a dictionary of metadata."""
741 """I return a Deferred that fires with a boolean, True if there
742 exists a child of the given name, False if not. The child name must
743 be a unicode string."""
746 """I return a Deferred that fires with a specific named child node,
747 either an IFileNode or an IDirectoryNode. The child name must be a
748 unicode string. I raise NoSuchChildError if I do not have a child by
751 def get_metadata_for(name):
752 """I return a Deferred that fires with the metadata dictionary for a
753 specific named child node. This metadata is stored in the *edge*, not
754 in the child, so it is attached to the parent dirnode rather than the
755 child dir-or-file-node. The child name must be a unicode string. I
756 raise NoSuchChildError if I do not have a child by that name."""
758 def set_metadata_for(name, metadata):
759 """I replace any existing metadata for the named child with the new
760 metadata. The child name must be a unicode string. This metadata is
761 stored in the *edge*, not in the child, so it is attached to the
762 parent dirnode rather than the child dir-or-file-node. I return a
763 Deferred (that fires with this dirnode) when the operation is
764 complete. I raise NoSuchChildError if I do not have a child by that
767 def get_child_at_path(path):
768 """Transform a child path into an IDirectoryNode or IFileNode.
770 I perform a recursive series of 'get' operations to find the named
771 descendant node. I return a Deferred that fires with the node, or
772 errbacks with NoSuchChildError if the node could not be found.
774 The path can be either a single string (slash-separated) or a list of
775 path-name elements. All elements must be unicode strings.
778 def get_child_and_metadata_at_path(path):
779 """Transform a child path into an IDirectoryNode/IFileNode and
782 I am like get_child_at_path(), but my Deferred fires with a tuple of
783 (node, metadata). The metadata comes from the last edge. If the path
784 is empty, the metadata will be an empty dictionary.
787 def set_uri(name, child_uri, metadata=None, overwrite=True):
788 """I add a child (by URI) at the specific name. I return a Deferred
789 that fires when the operation finishes. If overwrite= is True, I will
790 replace any existing child of the same name, otherwise an existing
791 child will cause me to return ExistingChildError. The child name must
794 The child_uri could be for a file, or for a directory (either
795 read-write or read-only, using a URI that came from get_uri() ).
797 If metadata= is provided, I will use it as the metadata for the named
798 edge. This will replace any existing metadata. If metadata= is left
799 as the default value of None, I will set ['mtime'] to the current
800 time, and I will set ['ctime'] to the current time if there was not
801 already a child by this name present. This roughly matches the
802 ctime/mtime semantics of traditional filesystems.
804 If this directory node is read-only, the Deferred will errback with a
807 def set_children(entries, overwrite=True):
808 """Add multiple (name, child_uri) pairs (or (name, child_uri,
809 metadata) triples) to a directory node. Returns a Deferred that fires
810 (with None) when the operation finishes. This is equivalent to
811 calling set_uri() multiple times, but is much more efficient. All
812 child names must be unicode strings.
815 def set_node(name, child, metadata=None, overwrite=True):
816 """I add a child at the specific name. I return a Deferred that fires
817 when the operation finishes. This Deferred will fire with the child
818 node that was just added. I will replace any existing child of the
819 same name. The child name must be a unicode string. The 'child'
820 instance must be an instance providing IDirectoryNode or IFileNode.
822 If metadata= is provided, I will use it as the metadata for the named
823 edge. This will replace any existing metadata. If metadata= is left
824 as the default value of None, I will set ['mtime'] to the current
825 time, and I will set ['ctime'] to the current time if there was not
826 already a child by this name present. This roughly matches the
827 ctime/mtime semantics of traditional filesystems.
829 If this directory node is read-only, the Deferred will errback with a
832 def set_nodes(entries, overwrite=True):
833 """Add multiple (name, child_node) pairs (or (name, child_node,
834 metadata) triples) to a directory node. Returns a Deferred that fires
835 (with None) when the operation finishes. This is equivalent to
836 calling set_node() multiple times, but is much more efficient. All
837 child names must be unicode strings."""
840 def add_file(name, uploadable, metadata=None, overwrite=True):
841 """I upload a file (using the given IUploadable), then attach the
842 resulting FileNode to the directory at the given name. I set metadata
843 the same way as set_uri and set_node. The child name must be a
846 I return a Deferred that fires (with the IFileNode of the uploaded
847 file) when the operation completes."""
850 """I remove the child at the specific name. I return a Deferred that
851 fires when the operation finishes. The child name must be a unicode
852 string. I raise NoSuchChildError if I do not have a child by that
855 def create_empty_directory(name, overwrite=True):
856 """I create and attach an empty directory at the given name. The
857 child name must be a unicode string. I return a Deferred that fires
858 when the operation finishes."""
860 def move_child_to(current_child_name, new_parent, new_child_name=None,
862 """I take one of my children and move them to a new parent. The child
863 is referenced by name. On the new parent, the child will live under
864 'new_child_name', which defaults to 'current_child_name'. TODO: what
865 should we do about metadata? I return a Deferred that fires when the
866 operation finishes. The child name must be a unicode string. I raise
867 NoSuchChildError if I do not have a child by that name."""
869 def build_manifest():
870 """I generate a table of everything reachable from this directory.
871 I also compute deep-stats as described below.
873 I return a Monitor. The Monitor's results will be a dictionary with
876 res['manifest']: a list of (path, cap) tuples for all nodes
877 (directories and files) reachable from this one.
878 'path' will be a tuple of unicode strings. The
879 origin dirnode will be represented by an empty path
881 res['verifycaps']: a list of (printable) verifycap strings, one for
882 each reachable non-LIT node. This is a set:
883 it will contain no duplicates.
884 res['storage-index']: a list of (base32) storage index strings,
885 one for each reachable non-LIT node. This is
886 a set: it will contain no duplicates.
887 res['stats']: a dictionary, the same that is generated by
888 start_deep_stats() below.
890 The Monitor will also have an .origin_si attribute with the (binary)
891 storage index of the starting point.
894 def start_deep_stats():
895 """Return a Monitor, examining all nodes (directories and files)
896 reachable from this one. The Monitor's results will be a dictionary
897 with the following keys::
899 count-immutable-files: count of how many CHK files are in the set
900 count-mutable-files: same, for mutable files (does not include
902 count-literal-files: same, for LIT files
903 count-files: sum of the above three
905 count-directories: count of directories
907 size-immutable-files: total bytes for all CHK files in the set
908 size-mutable-files (TODO): same, for current version of all mutable
909 files, does not include directories
910 size-literal-files: same, for LIT files
911 size-directories: size of mutable files used by directories
913 largest-directory: number of bytes in the largest directory
914 largest-directory-children: number of children in the largest
916 largest-immutable-file: number of bytes in the largest CHK file
918 size-mutable-files is not yet implemented, because it would involve
919 even more queries than deep_stats does.
921 The Monitor will also have an .origin_si attribute with the (binary)
922 storage index of the starting point.
924 This operation will visit every directory node underneath this one,
925 and can take a long time to run. On a typical workstation with good
926 bandwidth, this can examine roughly 15 directories per second (and
927 takes several minutes of 100% CPU for ~1700 directories).
930 class ICodecEncoder(Interface):
931 def set_params(data_size, required_shares, max_shares):
932 """Set up the parameters of this encoder.
934 This prepares the encoder to perform an operation that converts a
935 single block of data into a number of shares, such that a future
936 ICodecDecoder can use a subset of these shares to recover the
937 original data. This operation is invoked by calling encode(). Once
938 the encoding parameters are set up, the encode operation can be
939 invoked multiple times.
941 set_params() prepares the encoder to accept blocks of input data that
942 are exactly 'data_size' bytes in length. The encoder will be prepared
943 to produce 'max_shares' shares for each encode() operation (although
944 see the 'desired_share_ids' to use less CPU). The encoding math will
945 be chosen such that the decoder can get by with as few as
946 'required_shares' of these shares and still reproduce the original
947 data. For example, set_params(1000, 5, 5) offers no redundancy at
948 all, whereas set_params(1000, 1, 10) provides 10x redundancy.
950 Numerical Restrictions: 'data_size' is required to be an integral
951 multiple of 'required_shares'. In general, the caller should choose
952 required_shares and max_shares based upon their reliability
953 requirements and the number of peers available (the total storage
954 space used is roughly equal to max_shares*data_size/required_shares),
955 then choose data_size to achieve the memory footprint desired (larger
956 data_size means more efficient operation, smaller data_size means
957 smaller memory footprint).
959 In addition, 'max_shares' must be equal to or greater than
960 'required_shares'. Of course, setting them to be equal causes
961 encode() to degenerate into a particularly slow form of the 'split'
964 See encode() for more details about how these parameters are used.
966 set_params() must be called before any other ICodecEncoder methods
971 """Return the 3-tuple of data_size, required_shares, max_shares"""
973 def get_encoder_type():
974 """Return a short string that describes the type of this encoder.
976 There is required to be a global table of encoder classes. This method
977 returns an index into this table; the value at this index is an
978 encoder class, and this encoder is an instance of that class.
981 def get_block_size():
982 """Return the length of the shares that encode() will produce.
985 def encode_proposal(data, desired_share_ids=None):
988 'data' must be a string (or other buffer object), and len(data) must
989 be equal to the 'data_size' value passed earlier to set_params().
991 This will return a Deferred that will fire with two lists. The first
992 is a list of shares, each of which is a string (or other buffer
993 object) such that len(share) is the same as what get_share_size()
994 returned earlier. The second is a list of shareids, in which each is
995 an integer. The lengths of the two lists will always be equal to each
996 other. The user should take care to keep each share closely
997 associated with its shareid, as one is useless without the other.
999 The length of this output list will normally be the same as the value
1000 provided to the 'max_shares' parameter of set_params(). This may be
1001 different if 'desired_share_ids' is provided.
1003 'desired_share_ids', if provided, is required to be a sequence of
1004 ints, each of which is required to be >= 0 and < max_shares. If not
1005 provided, encode() will produce 'max_shares' shares, as if
1006 'desired_share_ids' were set to range(max_shares). You might use this
1007 if you initially thought you were going to use 10 peers, started
1008 encoding, and then two of the peers dropped out: you could use
1009 desired_share_ids= to skip the work (both memory and CPU) of
1010 producing shares for the peers which are no longer available.
1014 def encode(inshares, desired_share_ids=None):
1015 """Encode some data. This may be called multiple times. Each call is
1018 inshares is a sequence of length required_shares, containing buffers
1019 (i.e. strings), where each buffer contains the next contiguous
1020 non-overlapping segment of the input data. Each buffer is required to
1021 be the same length, and the sum of the lengths of the buffers is
1022 required to be exactly the data_size promised by set_params(). (This
1023 implies that the data has to be padded before being passed to
1024 encode(), unless of course it already happens to be an even multiple
1025 of required_shares in length.)
1027 ALSO: the requirement to break up your data into 'required_shares'
1028 chunks before calling encode() feels a bit surprising, at least from
1029 the point of view of a user who doesn't know how FEC works. It feels
1030 like an implementation detail that has leaked outside the
1031 abstraction barrier. Can you imagine a use case in which the data to
1032 be encoded might already be available in pre-segmented chunks, such
1033 that it is faster or less work to make encode() take a list rather
1034 than splitting a single string?
1036 ALSO ALSO: I think 'inshares' is a misleading term, since encode()
1037 is supposed to *produce* shares, so what it *accepts* should be
1038 something other than shares. Other places in this interface use the
1039 word 'data' for that-which-is-not-shares.. maybe we should use that
1042 'desired_share_ids', if provided, is required to be a sequence of
1043 ints, each of which is required to be >= 0 and < max_shares. If not
1044 provided, encode() will produce 'max_shares' shares, as if
1045 'desired_share_ids' were set to range(max_shares). You might use this
1046 if you initially thought you were going to use 10 peers, started
1047 encoding, and then two of the peers dropped out: you could use
1048 desired_share_ids= to skip the work (both memory and CPU) of
1049 producing shares for the peers which are no longer available.
1051 For each call, encode() will return a Deferred that fires with two
1052 lists, one containing shares and the other containing the shareids.
1053 The get_share_size() method can be used to determine the length of
1054 the share strings returned by encode(). Each shareid is a small
1055 integer, exactly as passed into 'desired_share_ids' (or
1056 range(max_shares), if desired_share_ids was not provided).
1058 The shares and their corresponding shareids are required to be kept
1059 together during storage and retrieval. Specifically, the share data is
1060 useless by itself: the decoder needs to be told which share is which
1061 by providing it with both the shareid and the actual share data.
1063 This function will allocate an amount of memory roughly equal to::
1065 (max_shares - required_shares) * get_share_size()
1067 When combined with the memory that the caller must allocate to
1068 provide the input data, this leads to a memory footprint roughly
1069 equal to the size of the resulting encoded shares (i.e. the expansion
1070 factor times the size of the input segment).
1075 # returning a list of (shareidN,shareN) tuples instead of a pair of
1076 # lists (shareids..,shares..). Brian thought the tuples would
1077 # encourage users to keep the share and shareid together throughout
1078 # later processing, Zooko pointed out that the code to iterate
1079 # through two lists is not really more complicated than using a list
1080 # of tuples and there's also a performance improvement
1082 # having 'data_size' not required to be an integral multiple of
1083 # 'required_shares'. Doing this would require encode() to perform
1084 # padding internally, and we'd prefer to have any padding be done
1085 # explicitly by the caller. Yes, it is an abstraction leak, but
1086 # hopefully not an onerous one.
1089 class ICodecDecoder(Interface):
1090 def set_params(data_size, required_shares, max_shares):
1091 """Set the params. They have to be exactly the same ones that were used for encoding. """
1093 def get_needed_shares():
1094 """Return the number of shares needed to reconstruct the data.
1095 set_params() is required to be called before this."""
1097 def decode(some_shares, their_shareids):
1098 """Decode a partial list of shares into data.
1100 'some_shares' is required to be a sequence of buffers of sharedata, a
1101 subset of the shares returned by ICodecEncode.encode(). Each share is
1102 required to be of the same length. The i'th element of their_shareids
1103 is required to be the shareid of the i'th buffer in some_shares.
1105 This returns a Deferred which fires with a sequence of buffers. This
1106 sequence will contain all of the segments of the original data, in
1107 order. The sum of the lengths of all of the buffers will be the
1108 'data_size' value passed into the original ICodecEncode.set_params()
1109 call. To get back the single original input block of data, use
1110 ''.join(output_buffers), or you may wish to simply write them in
1111 order to an output file.
1113 Note that some of the elements in the result sequence may be
1114 references to the elements of the some_shares input sequence. In
1115 particular, this means that if those share objects are mutable (e.g.
1116 arrays) and if they are changed, then both the input (the
1117 'some_shares' parameter) and the output (the value given when the
1118 deferred is triggered) will change.
1120 The length of 'some_shares' is required to be exactly the value of
1121 'required_shares' passed into the original ICodecEncode.set_params()
1125 class IEncoder(Interface):
1126 """I take an object that provides IEncryptedUploadable, which provides
1127 encrypted data, and a list of shareholders. I then encode, hash, and
1128 deliver shares to those shareholders. I will compute all the necessary
1129 Merkle hash trees that are necessary to validate the crypttext that
1130 eventually comes back from the shareholders. I provide the URI Extension
1131 Block Hash, and the encoding parameters, both of which must be included
1134 I do not choose shareholders, that is left to the IUploader. I must be
1135 given a dict of RemoteReferences to storage buckets that are ready and
1136 willing to receive data.
1140 """Specify the number of bytes that will be encoded. This must be
1141 peformed before get_serialized_params() can be called.
1143 def set_params(params):
1144 """Override the default encoding parameters. 'params' is a tuple of
1145 (k,d,n), where 'k' is the number of required shares, 'd' is the
1146 shares_of_happiness, and 'n' is the total number of shares that will
1149 Encoding parameters can be set in three ways. 1: The Encoder class
1150 provides defaults (3/7/10). 2: the Encoder can be constructed with
1151 an 'options' dictionary, in which the
1152 needed_and_happy_and_total_shares' key can be a (k,d,n) tuple. 3:
1153 set_params((k,d,n)) can be called.
1155 If you intend to use set_params(), you must call it before
1156 get_share_size or get_param are called.
1159 def set_encrypted_uploadable(u):
1160 """Provide a source of encrypted upload data. 'u' must implement
1161 IEncryptedUploadable.
1163 When this is called, the IEncryptedUploadable will be queried for its
1164 length and the storage_index that should be used.
1166 This returns a Deferred that fires with this Encoder instance.
1168 This must be performed before start() can be called.
1171 def get_param(name):
1172 """Return an encoding parameter, by name.
1174 'storage_index': return a string with the (16-byte truncated SHA-256
1175 hash) storage index to which these shares should be
1178 'share_counts': return a tuple describing how many shares are used:
1179 (needed_shares, shares_of_happiness, total_shares)
1181 'num_segments': return an int with the number of segments that
1184 'segment_size': return an int with the size of each segment.
1186 'block_size': return the size of the individual blocks that will
1187 be delivered to a shareholder's put_block() method. By
1188 knowing this, the shareholder will be able to keep all
1189 blocks in a single file and still provide random access
1190 when reading them. # TODO: can we avoid exposing this?
1192 'share_size': an int with the size of the data that will be stored
1193 on each shareholder. This is aggregate amount of data
1194 that will be sent to the shareholder, summed over all
1195 the put_block() calls I will ever make. It is useful to
1196 determine this size before asking potential
1197 shareholders whether they will grant a lease or not,
1198 since their answers will depend upon how much space we
1199 need. TODO: this might also include some amount of
1200 overhead, like the size of all the hashes. We need to
1201 decide whether this is useful or not.
1203 'serialized_params': a string with a concise description of the
1204 codec name and its parameters. This may be passed
1205 into the IUploadable to let it make sure that
1206 the same file encoded with different parameters
1207 will result in different storage indexes.
1209 Once this is called, set_size() and set_params() may not be called.
1212 def set_shareholders(shareholders):
1213 """Tell the encoder where to put the encoded shares. 'shareholders'
1214 must be a dictionary that maps share number (an integer ranging from
1215 0 to n-1) to an instance that provides IStorageBucketWriter. This
1216 must be performed before start() can be called."""
1219 """Begin the encode/upload process. This involves reading encrypted
1220 data from the IEncryptedUploadable, encoding it, uploading the shares
1221 to the shareholders, then sending the hash trees.
1223 set_encrypted_uploadable() and set_shareholders() must be called
1224 before this can be invoked.
1226 This returns a Deferred that fires with a verify cap when the upload process is
1227 complete. The verifycap, plus the encryption key, is sufficient to construct the read
1231 class IDecoder(Interface):
1232 """I take a list of shareholders and some setup information, then
1233 download, validate, decode, and decrypt data from them, writing the
1234 results to an output file.
1236 I do not locate the shareholders, that is left to the IDownloader. I must
1237 be given a dict of RemoteReferences to storage buckets that are ready to
1242 """I take a file-like object (providing write and close) to which all
1243 the plaintext data will be written.
1245 TODO: producer/consumer . Maybe write() should return a Deferred that
1246 indicates when it will accept more data? But probably having the
1247 IDecoder be a producer is easier to glue to IConsumer pieces.
1250 def set_shareholders(shareholders):
1251 """I take a dictionary that maps share identifiers (small integers)
1252 to RemoteReferences that provide RIBucketReader. This must be called
1256 """I start the download. This process involves retrieving data and
1257 hash chains from the shareholders, using the hashes to validate the
1258 data, decoding the shares into segments, decrypting the segments,
1259 then writing the resulting plaintext to the output file.
1261 I return a Deferred that will fire (with self) when the download is
1265 class IDownloadTarget(Interface):
1266 # Note that if the IDownloadTarget is also an IConsumer, the downloader
1267 # will register itself as a producer. This allows the target to invoke
1268 # downloader.pauseProducing, resumeProducing, and stopProducing.
1270 """Called before any calls to write() or close(). If an error
1271 occurs before any data is available, fail() may be called without
1272 a previous call to open().
1274 'size' is the length of the file being downloaded, in bytes."""
1277 """Output some data to the target."""
1279 """Inform the target that there is no more data to be written."""
1281 """fail() is called to indicate that the download has failed. 'why'
1282 is a Failure object indicating what went wrong. No further methods
1283 will be invoked on the IDownloadTarget after fail()."""
1284 def register_canceller(cb):
1285 """The CiphertextDownloader uses this to register a no-argument function
1286 that the target can call to cancel the download. Once this canceller
1287 is invoked, no further calls to write() or close() will be made."""
1289 """When the CiphertextDownloader is done, this finish() function will be
1290 called. Whatever it returns will be returned to the invoker of
1291 Downloader.download.
1293 # The following methods are just because that target might be a repairer.DownUpConnector,
1294 # and just because the current CHKUpload object expects to find the storage index and
1295 # encoding parameters in its Uploadable.
1296 def set_storageindex(storageindex):
1297 """ Set the storage index. """
1298 def set_encodingparams(encodingparams):
1299 """ Set the encoding parameters. """
1301 class IDownloader(Interface):
1302 def download(uri, target):
1303 """Perform a CHK download, sending the data to the given target.
1304 'target' must provide IDownloadTarget.
1306 Returns a Deferred that fires (with the results of target.finish)
1307 when the download is finished, or errbacks if something went wrong."""
1309 class IEncryptedUploadable(Interface):
1310 def set_upload_status(upload_status):
1311 """Provide an IUploadStatus object that should be filled with status
1312 information. The IEncryptedUploadable is responsible for setting
1313 key-determination progress ('chk'), size, storage_index, and
1314 ciphertext-fetch progress. It may delegate some of this
1315 responsibility to others, in particular to the IUploadable."""
1318 """This behaves just like IUploadable.get_size()."""
1320 def get_all_encoding_parameters():
1321 """Return a Deferred that fires with a tuple of
1322 (k,happy,n,segment_size). The segment_size will be used as-is, and
1323 must match the following constraints: it must be a multiple of k, and
1324 it shouldn't be unreasonably larger than the file size (if
1325 segment_size is larger than filesize, the difference must be stored
1328 This usually passes through to the IUploadable method of the same
1331 The encoder strictly obeys the values returned by this method. To
1332 make an upload use non-default encoding parameters, you must arrange
1333 to control the values that this method returns.
1336 def get_storage_index():
1337 """Return a Deferred that fires with a 16-byte storage index.
1340 def read_encrypted(length, hash_only):
1341 """This behaves just like IUploadable.read(), but returns crypttext
1342 instead of plaintext. If hash_only is True, then this discards the
1343 data (and returns an empty list); this improves efficiency when
1344 resuming an interrupted upload (where we need to compute the
1345 plaintext hashes, but don't need the redundant encrypted data)."""
1347 def get_plaintext_hashtree_leaves(first, last, num_segments):
1348 """OBSOLETE; Get the leaf nodes of a merkle hash tree over the plaintext
1349 segments, i.e. get the tagged hashes of the given segments. The
1350 segment size is expected to be generated by the IEncryptedUploadable
1351 before any plaintext is read or ciphertext produced, so that the
1352 segment hashes can be generated with only a single pass.
1354 This returns a Deferred which fires with a sequence of hashes, using:
1356 tuple(segment_hashes[first:last])
1358 'num_segments' is used to assert that the number of segments that the
1359 IEncryptedUploadable handled matches the number of segments that the
1360 encoder was expecting.
1362 This method must not be called until the final byte has been read
1363 from read_encrypted(). Once this method is called, read_encrypted()
1364 can never be called again.
1367 def get_plaintext_hash():
1368 """OBSOLETE; Get the hash of the whole plaintext.
1370 This returns a Deferred which fires with a tagged SHA-256 hash of the
1371 whole plaintext, obtained from hashutil.plaintext_hash(data).
1375 """Just like IUploadable.close()."""
1377 class IUploadable(Interface):
1378 def set_upload_status(upload_status):
1379 """Provide an IUploadStatus object that should be filled with status
1380 information. The IUploadable is responsible for setting
1381 key-determination progress ('chk')."""
1383 def set_default_encoding_parameters(params):
1384 """Set the default encoding parameters, which must be a dict mapping
1385 strings to ints. The meaningful keys are 'k', 'happy', 'n', and
1386 'max_segment_size'. These might have an influence on the final
1387 encoding parameters returned by get_all_encoding_parameters(), if the
1388 Uploadable doesn't have more specific preferences.
1390 This call is optional: if it is not used, the Uploadable will use
1391 some built-in defaults. If used, this method must be called before
1392 any other IUploadable methods to have any effect.
1396 """Return a Deferred that will fire with the length of the data to be
1397 uploaded, in bytes. This will be called before the data is actually
1398 used, to compute encoding parameters.
1401 def get_all_encoding_parameters():
1402 """Return a Deferred that fires with a tuple of
1403 (k,happy,n,segment_size). The segment_size will be used as-is, and
1404 must match the following constraints: it must be a multiple of k, and
1405 it shouldn't be unreasonably larger than the file size (if
1406 segment_size is larger than filesize, the difference must be stored
1409 The relative values of k and n allow some IUploadables to request
1410 better redundancy than others (in exchange for consuming more space
1413 Larger values of segment_size reduce hash overhead, while smaller
1414 values reduce memory footprint and cause data to be delivered in
1415 smaller pieces (which may provide a smoother and more predictable
1416 download experience).
1418 The encoder strictly obeys the values returned by this method. To
1419 make an upload use non-default encoding parameters, you must arrange
1420 to control the values that this method returns. One way to influence
1421 them may be to call set_encoding_parameters() before calling
1422 get_all_encoding_parameters().
1425 def get_encryption_key():
1426 """Return a Deferred that fires with a 16-byte AES key. This key will
1427 be used to encrypt the data. The key will also be hashed to derive
1430 Uploadables which want to achieve convergence should hash their file
1431 contents and the serialized_encoding_parameters to form the key
1432 (which of course requires a full pass over the data). Uploadables can
1433 use the upload.ConvergentUploadMixin class to achieve this
1436 Uploadables which do not care about convergence (or do not wish to
1437 make multiple passes over the data) can simply return a
1438 strongly-random 16 byte string.
1440 get_encryption_key() may be called multiple times: the IUploadable is
1441 required to return the same value each time.
1445 """Return a Deferred that fires with a list of strings (perhaps with
1446 only a single element) which, when concatenated together, contain the
1447 next 'length' bytes of data. If EOF is near, this may provide fewer
1448 than 'length' bytes. The total number of bytes provided by read()
1449 before it signals EOF must equal the size provided by get_size().
1451 If the data must be acquired through multiple internal read
1452 operations, returning a list instead of a single string may help to
1453 reduce string copies.
1455 'length' will typically be equal to (min(get_size(),1MB)/req_shares),
1456 so a 10kB file means length=3kB, 100kB file means length=30kB,
1457 and >=1MB file means length=300kB.
1459 This method provides for a single full pass through the data. Later
1460 use cases may desire multiple passes or access to only parts of the
1461 data (such as a mutable file making small edits-in-place). This API
1462 will be expanded once those use cases are better understood.
1466 """The upload is finished, and whatever filehandle was in use may be
1469 class IUploadResults(Interface):
1470 """I am returned by upload() methods. I contain a number of public
1471 attributes which can be read to determine the results of the upload. Some
1472 of these are functional, some are timing information. All of these may be
1475 .file_size : the size of the file, in bytes
1476 .uri : the CHK read-cap for the file
1477 .ciphertext_fetched : how many bytes were fetched by the helper
1478 .sharemap: dict mapping share identifier to set of serverids
1479 (binary strings). This indicates which servers were given
1480 which shares. For immutable files, the shareid is an
1481 integer (the share number, from 0 to N-1). For mutable
1482 files, it is a string of the form 'seq%d-%s-sh%d',
1483 containing the sequence number, the roothash, and the
1485 .servermap : dict mapping server peerid to a set of share numbers
1486 .timings : dict of timing information, mapping name to seconds (float)
1487 total : total upload time, start to finish
1488 storage_index : time to compute the storage index
1489 peer_selection : time to decide which peers will be used
1490 contacting_helper : initial helper query to upload/no-upload decision
1491 existence_check : helper pre-upload existence check
1492 helper_total : initial helper query to helper finished pushing
1493 cumulative_fetch : helper waiting for ciphertext requests
1494 total_fetch : helper start to last ciphertext response
1495 cumulative_encoding : just time spent in zfec
1496 cumulative_sending : just time spent waiting for storage servers
1497 hashes_and_close : last segment push to shareholder close
1498 total_encode_and_push : first encode to shareholder close
1502 class IDownloadResults(Interface):
1503 """I am created internally by download() methods. I contain a number of
1504 public attributes which contain details about the download process.::
1506 .file_size : the size of the file, in bytes
1507 .servers_used : set of server peerids that were used during download
1508 .server_problems : dict mapping server peerid to a problem string. Only
1509 servers that had problems (bad hashes, disconnects) are
1511 .servermap : dict mapping server peerid to a set of share numbers. Only
1512 servers that had any shares are listed here.
1513 .timings : dict of timing information, mapping name to seconds (float)
1514 peer_selection : time to ask servers about shares
1515 servers_peer_selection : dict of peerid to DYHB-query time
1516 uri_extension : time to fetch a copy of the URI extension block
1517 hashtrees : time to fetch the hash trees
1518 segments : time to fetch, decode, and deliver segments
1519 cumulative_fetch : time spent waiting for storage servers
1520 cumulative_decode : just time spent in zfec
1521 cumulative_decrypt : just time spent in decryption
1522 total : total download time, start to finish
1523 fetch_per_server : dict of peerid to list of per-segment fetch times
1527 class IUploader(Interface):
1528 def upload(uploadable):
1529 """Upload the file. 'uploadable' must impement IUploadable. This
1530 returns a Deferred which fires with an UploadResults instance, from
1531 which the URI of the file can be obtained as results.uri ."""
1533 def upload_ssk(write_capability, new_version, uploadable):
1534 """TODO: how should this work?"""
1536 class ICheckable(Interface):
1537 def check(monitor, verify=False):
1538 """Check upon my health, optionally repairing any problems.
1540 This returns a Deferred that fires with an instance that provides
1541 ICheckResults, or None if the object is non-distributed (i.e. LIT
1544 The monitor will be checked periodically to see if the operation has
1545 been cancelled. If so, no new queries will be sent, and the Deferred
1546 will fire (with a OperationCancelledError) immediately.
1548 Filenodes and dirnodes (which provide IFilesystemNode) are also
1549 checkable. Instances that represent verifier-caps will be checkable
1550 but not downloadable. Some objects (like LIT files) do not actually
1551 live in the grid, and their checkers return None (non-distributed
1552 files are always healthy).
1554 If verify=False, a relatively lightweight check will be performed: I
1555 will ask all servers if they have a share for me, and I will believe
1556 whatever they say. If there are at least N distinct shares on the
1557 grid, my results will indicate r.is_healthy()==True. This requires a
1558 roundtrip to each server, but does not transfer very much data, so
1559 the network bandwidth is fairly low.
1561 If verify=True, a more resource-intensive check will be performed:
1562 every share will be downloaded, and the hashes will be validated on
1563 every bit. I will ignore any shares that failed their hash checks. If
1564 there are at least N distinct valid shares on the grid, my results
1565 will indicate r.is_healthy()==True. This requires N/k times as much
1566 download bandwidth (and server disk IO) as a regular download. If a
1567 storage server is holding a corrupt share, or is experiencing memory
1568 failures during retrieval, or is malicious or buggy, then
1569 verification will detect the problem, but checking will not.
1571 TODO: any problems seen during checking will be reported to the
1572 health-manager.furl, a centralized object which is responsible for
1573 figuring out why files are unhealthy so corrective action can be
1577 def check_and_repair(monitor, verify=False):
1578 """Like check(), but if the file/directory is not healthy, attempt to
1581 Any non-healthy result will cause an immediate repair operation, to
1582 generate and upload new shares. After repair, the file will be as
1583 healthy as we can make it. Details about what sort of repair is done
1584 will be put in the check-and-repair results. The Deferred will not
1585 fire until the repair is complete.
1587 This returns a Deferred which fires with an instance of
1588 ICheckAndRepairResults."""
1590 class IDeepCheckable(Interface):
1591 def start_deep_check(verify=False):
1592 """Check upon the health of me and everything I can reach.
1594 This is a recursive form of check(), useable only on dirnodes.
1596 I return a Monitor, with results that are an IDeepCheckResults
1600 def start_deep_check_and_repair(verify=False):
1601 """Check upon the health of me and everything I can reach. Repair
1602 anything that isn't healthy.
1604 This is a recursive form of check_and_repair(), useable only on
1607 I return a Monitor, with results that are an
1608 IDeepCheckAndRepairResults object.
1611 class ICheckResults(Interface):
1612 """I contain the detailed results of a check/verify operation.
1615 def get_storage_index():
1616 """Return a string with the (binary) storage index."""
1617 def get_storage_index_string():
1618 """Return a string with the (printable) abbreviated storage index."""
1620 """Return the (string) URI of the object that was checked."""
1623 """Return a boolean, True if the file/dir is fully healthy, False if
1624 it is damaged in any way. Non-distributed LIT files always return
1627 def is_recoverable():
1628 """Return a boolean, True if the file/dir can be recovered, False if
1629 not. Unrecoverable files are obviously unhealthy. Non-distributed LIT
1630 files always return True."""
1632 def needs_rebalancing():
1633 """Return a boolean, True if the file/dir's reliability could be
1634 improved by moving shares to new servers. Non-distributed LIT files
1635 always return False."""
1639 """Return a dictionary that describes the state of the file/dir. LIT
1640 files always return an empty dictionary. Normal files and directories return a
1641 dictionary with the following keys (note that these use binary strings rather than
1642 base32-encoded ones) (also note that for mutable files, these counts are for the 'best'
1645 count-shares-good: the number of distinct good shares that were found
1646 count-shares-needed: 'k', the number of shares required for recovery
1647 count-shares-expected: 'N', the number of total shares generated
1648 count-good-share-hosts: the number of distinct storage servers with
1649 good shares. If this number is less than
1650 count-shares-good, then some shares are
1651 doubled up, increasing the correlation of
1652 failures. This indicates that one or more
1653 shares should be moved to an otherwise unused
1654 server, if one is available.
1655 count-corrupt-shares: the number of shares with integrity failures
1656 list-corrupt-shares: a list of 'share locators', one for each share
1657 that was found to be corrupt. Each share
1658 locator is a list of (serverid, storage_index,
1660 count-incompatible-shares: the number of shares which are of a share format unknown to
1662 list-incompatible-shares: a list of 'share locators', one for each share that was found
1663 to be of an unknown format. Each share locator is a list of
1664 (serverid, storage_index, sharenum).
1665 servers-responding: list of (binary) storage server identifiers,
1666 one for each server which responded to the share
1667 query (even if they said they didn't have shares,
1668 and even if they said they did have shares but then
1669 didn't send them when asked, or dropped the
1670 connection, or returned a Failure, and even if they
1671 said they did have shares and sent incorrect ones
1673 sharemap: dict mapping share identifier to list of serverids
1674 (binary strings). This indicates which servers are holding
1675 which shares. For immutable files, the shareid is an
1676 integer (the share number, from 0 to N-1). For mutable
1677 files, it is a string of the form 'seq%d-%s-sh%d',
1678 containing the sequence number, the roothash, and the
1681 The following keys are most relevant for mutable files, but immutable
1682 files will provide sensible values too::
1684 count-wrong-shares: the number of shares for versions other than the
1685 'best' one (which is defined as being the
1686 recoverable version with the highest sequence
1687 number, then the highest roothash). These are
1688 either leftover shares from an older version
1689 (perhaps on a server that was offline when an
1690 update occurred), shares from an unrecoverable
1691 newer version, or shares from an alternate
1692 current version that results from an
1693 uncoordinated write collision. For a healthy
1694 file, this will equal 0.
1696 count-recoverable-versions: the number of recoverable versions of
1697 the file. For a healthy file, this will
1700 count-unrecoverable-versions: the number of unrecoverable versions
1701 of the file. For a healthy file, this
1707 """Return a string with a brief (one-line) summary of the results."""
1710 """Return a list of strings with more detailed results."""
1712 class ICheckAndRepairResults(Interface):
1713 """I contain the detailed results of a check/verify/repair operation.
1715 The IFilesystemNode.check()/verify()/repair() methods all return
1716 instances that provide ICheckAndRepairResults.
1719 def get_storage_index():
1720 """Return a string with the (binary) storage index."""
1721 def get_storage_index_string():
1722 """Return a string with the (printable) abbreviated storage index."""
1723 def get_repair_attempted():
1724 """Return a boolean, True if a repair was attempted."""
1725 def get_repair_successful():
1726 """Return a boolean, True if repair was attempted and the file/dir
1727 was fully healthy afterwards. False if no repair was attempted or if
1728 a repair attempt failed."""
1729 def get_pre_repair_results():
1730 """Return an ICheckResults instance that describes the state of the
1731 file/dir before any repair was attempted."""
1732 def get_post_repair_results():
1733 """Return an ICheckResults instance that describes the state of the
1734 file/dir after any repair was attempted. If no repair was attempted,
1735 the pre-repair and post-repair results will be identical."""
1738 class IDeepCheckResults(Interface):
1739 """I contain the results of a deep-check operation.
1741 This is returned by a call to ICheckable.deep_check().
1744 def get_root_storage_index_string():
1745 """Return the storage index (abbreviated human-readable string) of
1746 the first object checked."""
1748 """Return a dictionary with the following keys::
1750 count-objects-checked: count of how many objects were checked
1751 count-objects-healthy: how many of those objects were completely
1753 count-objects-unhealthy: how many were damaged in some way
1754 count-objects-unrecoverable: how many were unrecoverable
1755 count-corrupt-shares: how many shares were found to have
1756 corruption, summed over all objects
1760 def get_corrupt_shares():
1761 """Return a set of (serverid, storage_index, sharenum) for all shares
1762 that were found to be corrupt. Both serverid and storage_index are
1765 def get_all_results():
1766 """Return a dictionary mapping pathname (a tuple of strings, ready to
1767 be slash-joined) to an ICheckResults instance, one for each object
1768 that was checked."""
1770 def get_results_for_storage_index(storage_index):
1771 """Retrive the ICheckResults instance for the given (binary)
1772 storage index. Raises KeyError if there are no results for that
1776 """Return a dictionary with the same keys as
1777 IDirectoryNode.deep_stats()."""
1779 class IDeepCheckAndRepairResults(Interface):
1780 """I contain the results of a deep-check-and-repair operation.
1782 This is returned by a call to ICheckable.deep_check_and_repair().
1785 def get_root_storage_index_string():
1786 """Return the storage index (abbreviated human-readable string) of
1787 the first object checked."""
1789 """Return a dictionary with the following keys::
1791 count-objects-checked: count of how many objects were checked
1792 count-objects-healthy-pre-repair: how many of those objects were
1793 completely healthy (before any
1795 count-objects-unhealthy-pre-repair: how many were damaged in
1797 count-objects-unrecoverable-pre-repair: how many were unrecoverable
1798 count-objects-healthy-post-repair: how many of those objects were
1799 completely healthy (after any
1801 count-objects-unhealthy-post-repair: how many were damaged in
1803 count-objects-unrecoverable-post-repair: how many were
1805 count-repairs-attempted: repairs were attempted on this many
1806 objects. The count-repairs- keys will
1807 always be provided, however unless
1808 repair=true is present, they will all
1810 count-repairs-successful: how many repairs resulted in healthy
1812 count-repairs-unsuccessful: how many repairs resulted did not
1813 results in completely healthy objects
1814 count-corrupt-shares-pre-repair: how many shares were found to
1815 have corruption, summed over all
1816 objects examined (before any
1818 count-corrupt-shares-post-repair: how many shares were found to
1819 have corruption, summed over all
1820 objects examined (after any
1825 """Return a dictionary with the same keys as
1826 IDirectoryNode.deep_stats()."""
1828 def get_corrupt_shares():
1829 """Return a set of (serverid, storage_index, sharenum) for all shares
1830 that were found to be corrupt before any repair was attempted. Both
1831 serverid and storage_index are binary.
1833 def get_remaining_corrupt_shares():
1834 """Return a set of (serverid, storage_index, sharenum) for all shares
1835 that were found to be corrupt after any repair was completed. Both
1836 serverid and storage_index are binary. These are shares that need
1837 manual inspection and probably deletion.
1839 def get_all_results():
1840 """Return a dictionary mapping pathname (a tuple of strings, ready to
1841 be slash-joined) to an ICheckAndRepairResults instance, one for each
1842 object that was checked."""
1844 def get_results_for_storage_index(storage_index):
1845 """Retrive the ICheckAndRepairResults instance for the given (binary)
1846 storage index. Raises KeyError if there are no results for that
1850 class IRepairable(Interface):
1851 def repair(check_results):
1852 """Attempt to repair the given object. Returns a Deferred that fires
1853 with a IRepairResults object.
1855 I must be called with an object that implements ICheckResults, as
1856 proof that you have actually discovered a problem with this file. I
1857 will use the data in the checker results to guide the repair process,
1858 such as which servers provided bad data and should therefore be
1859 avoided. The ICheckResults object is inside the
1860 ICheckAndRepairResults object, which is returned by the
1861 ICheckable.check() method::
1863 d = filenode.check(repair=False)
1864 def _got_results(check_and_repair_results):
1865 check_results = check_and_repair_results.get_pre_repair_results()
1866 return filenode.repair(check_results)
1867 d.addCallback(_got_results)
1871 class IRepairResults(Interface):
1872 """I contain the results of a repair operation."""
1875 class IClient(Interface):
1876 def upload(uploadable):
1877 """Upload some data into a CHK, get back the UploadResults for it.
1878 @param uploadable: something that implements IUploadable
1879 @return: a Deferred that fires with the UploadResults instance.
1880 To get the URI for this file, use results.uri .
1883 def create_mutable_file(contents=""):
1884 """Create a new mutable file with contents, get back the URI string.
1885 @param contents: the initial contents to place in the file.
1886 @return: a Deferred that fires with tne (string) SSK URI for the new
1890 def create_empty_dirnode():
1891 """Create a new dirnode, empty and unattached.
1892 @return: a Deferred that fires with the new IDirectoryNode instance.
1895 def create_node_from_uri(uri):
1896 """Create a new IFilesystemNode instance from the uri, synchronously.
1897 @param uri: a string or IURI-providing instance. This could be for a
1898 LiteralFileNode, a CHK file node, a mutable file node, or
1900 @return: an instance that provides IFilesystemNode (or more usefully one
1901 of its subclasses). File-specifying URIs will result in
1902 IFileNode or IMutableFileNode -providing instances, like
1903 FileNode, LiteralFileNode, or MutableFileNode.
1904 Directory-specifying URIs will result in
1905 IDirectoryNode-providing instances, like NewDirectoryNode.
1908 class IClientStatus(Interface):
1909 def list_all_uploads():
1910 """Return a list of uploader objects, one for each upload which
1911 currently has an object available (tracked with weakrefs). This is
1912 intended for debugging purposes."""
1913 def list_active_uploads():
1914 """Return a list of active IUploadStatus objects."""
1915 def list_recent_uploads():
1916 """Return a list of IUploadStatus objects for the most recently
1919 def list_all_downloads():
1920 """Return a list of downloader objects, one for each download which
1921 currently has an object available (tracked with weakrefs). This is
1922 intended for debugging purposes."""
1923 def list_active_downloads():
1924 """Return a list of active IDownloadStatus objects."""
1925 def list_recent_downloads():
1926 """Return a list of IDownloadStatus objects for the most recently
1927 started downloads."""
1929 class IUploadStatus(Interface):
1931 """Return a timestamp (float with seconds since epoch) indicating
1932 when the operation was started."""
1933 def get_storage_index():
1934 """Return a string with the (binary) storage index in use on this
1935 upload. Returns None if the storage index has not yet been
1938 """Return an integer with the number of bytes that will eventually
1939 be uploaded for this file. Returns None if the size is not yet known.
1942 """Return True if this upload is using a Helper, False if not."""
1944 """Return a string describing the current state of the upload
1947 """Returns a tuple of floats, (chk, ciphertext, encode_and_push),
1948 each from 0.0 to 1.0 . 'chk' describes how much progress has been
1949 made towards hashing the file to determine a CHK encryption key: if
1950 non-convergent encryption is in use, this will be trivial, otherwise
1951 the whole file must be hashed. 'ciphertext' describes how much of the
1952 ciphertext has been pushed to the helper, and is '1.0' for non-helper
1953 uploads. 'encode_and_push' describes how much of the encode-and-push
1954 process has finished: for helper uploads this is dependent upon the
1955 helper providing progress reports. It might be reasonable to add all
1956 three numbers and report the sum to the user."""
1958 """Return True if the upload is currently active, False if not."""
1960 """Return an instance of UploadResults (which contains timing and
1961 sharemap information). Might return None if the upload is not yet
1964 """Each upload status gets a unique number: this method returns that
1965 number. This provides a handle to this particular upload, so a web
1966 page can generate a suitable hyperlink."""
1968 class IDownloadStatus(Interface):
1970 """Return a timestamp (float with seconds since epoch) indicating
1971 when the operation was started."""
1972 def get_storage_index():
1973 """Return a string with the (binary) storage index in use on this
1974 download. This may be None if there is no storage index (i.e. LIT
1977 """Return an integer with the number of bytes that will eventually be
1978 retrieved for this file. Returns None if the size is not yet known.
1981 """Return True if this download is using a Helper, False if not."""
1983 """Return a string describing the current state of the download
1986 """Returns a float (from 0.0 to 1.0) describing the amount of the
1987 download that has completed. This value will remain at 0.0 until the
1988 first byte of plaintext is pushed to the download target."""
1990 """Return True if the download is currently active, False if not."""
1992 """Each download status gets a unique number: this method returns
1993 that number. This provides a handle to this particular download, so a
1994 web page can generate a suitable hyperlink."""
1996 class IServermapUpdaterStatus(Interface):
1998 class IPublishStatus(Interface):
2000 class IRetrieveStatus(Interface):
2003 class NotCapableError(Exception):
2004 """You have tried to write to a read-only node."""
2006 class BadWriteEnablerError(Exception):
2009 class RIControlClient(RemoteInterface):
2011 def wait_for_client_connections(num_clients=int):
2012 """Do not return until we have connections to at least NUM_CLIENTS
2016 def upload_from_file_to_uri(filename=str, convergence=ChoiceOf(None, StringConstraint(2**20))):
2017 """Upload a file to the grid. This accepts a filename (which must be
2018 absolute) that points to a file on the node's local disk. The node will
2019 read the contents of this file, upload it to the grid, then return the
2020 URI at which it was uploaded. If convergence is None then a random
2021 encryption key will be used, else the plaintext will be hashed, then
2022 that hash will be mixed together with the "convergence" string to form
2027 def download_from_uri_to_file(uri=URI, filename=str):
2028 """Download a file from the grid, placing it on the node's local disk
2029 at the given filename (which must be absolute[?]). Returns the
2030 absolute filename where the file was written."""
2035 def get_memory_usage():
2036 """Return a dict describes the amount of memory currently in use. The
2037 keys are 'VmPeak', 'VmSize', and 'VmData'. The values are integers,
2038 measuring memory consupmtion in bytes."""
2039 return DictOf(str, int)
2041 def speed_test(count=int, size=int, mutable=Any()):
2042 """Write 'count' tempfiles to disk, all of the given size. Measure
2043 how long (in seconds) it takes to upload them all to the servers.
2044 Then measure how long it takes to download all of them. If 'mutable'
2045 is 'create', time creation of mutable files. If 'mutable' is
2046 'upload', then time access to the same mutable file instead of
2049 Returns a tuple of (upload_time, download_time).
2051 return (float, float)
2053 def measure_peer_response_time():
2054 """Send a short message to each connected peer, and measure the time
2055 it takes for them to respond to it. This is a rough measure of the
2056 application-level round trip time.
2058 @return: a dictionary mapping peerid to a float (RTT time in seconds)
2061 return DictOf(Nodeid, float)
2063 UploadResults = Any() #DictOf(str, str)
2065 class RIEncryptedUploadable(RemoteInterface):
2066 __remote_name__ = "RIEncryptedUploadable.tahoe.allmydata.com"
2071 def get_all_encoding_parameters():
2072 return (int, int, int, long)
2074 def read_encrypted(offset=Offset, length=ReadSize):
2077 def get_plaintext_hashtree_leaves(first=int, last=int, num_segments=int):
2080 def get_plaintext_hash():
2087 class RICHKUploadHelper(RemoteInterface):
2088 __remote_name__ = "RIUploadHelper.tahoe.allmydata.com"
2092 Return a dictionary of version information.
2094 return DictOf(str, Any())
2096 def upload(reader=RIEncryptedUploadable):
2097 return UploadResults
2100 class RIHelper(RemoteInterface):
2101 __remote_name__ = "RIHelper.tahoe.allmydata.com"
2105 Return a dictionary of version information.
2107 return DictOf(str, Any())
2109 def upload_chk(si=StorageIndex):
2110 """See if a file with a given storage index needs uploading. The
2111 helper will ask the appropriate storage servers to see if the file
2112 has already been uploaded. If so, the helper will return a set of
2113 'upload results' that includes whatever hashes are needed to build
2114 the read-cap, and perhaps a truncated sharemap.
2116 If the file has not yet been uploaded (or if it was only partially
2117 uploaded), the helper will return an empty upload-results dictionary
2118 and also an RICHKUploadHelper object that will take care of the
2119 upload process. The client should call upload() on this object and
2120 pass it a reference to an RIEncryptedUploadable object that will
2121 provide ciphertext. When the upload is finished, the upload() method
2122 will finish and return the upload results.
2124 return (UploadResults, ChoiceOf(RICHKUploadHelper, None))
2127 class RIStatsProvider(RemoteInterface):
2128 __remote_name__ = "RIStatsProvider.tahoe.allmydata.com"
2130 Provides access to statistics and monitoring information.
2135 returns a dictionary containing 'counters' and 'stats', each a dictionary
2136 with string counter/stat name keys, and numeric values. counters are
2137 monotonically increasing measures of work done, and stats are instantaneous
2138 measures (potentially time averaged internally)
2140 return DictOf(str, DictOf(str, ChoiceOf(float, int, long)))
2142 class RIStatsGatherer(RemoteInterface):
2143 __remote_name__ = "RIStatsGatherer.tahoe.allmydata.com"
2145 Provides a monitoring service for centralised collection of stats
2148 def provide(provider=RIStatsProvider, nickname=str):
2150 @param provider: a stats collector instance which should be polled
2151 periodically by the gatherer to collect stats.
2152 @param nickname: a name useful to identify the provided client
2157 class IStatsProducer(Interface):
2160 returns a dictionary, with str keys representing the names of stats
2161 to be monitored, and numeric values.
2164 class RIKeyGenerator(RemoteInterface):
2165 __remote_name__ = "RIKeyGenerator.tahoe.allmydata.com"
2167 Provides a service offering to make RSA key pairs.
2170 def get_rsa_key_pair(key_size=int):
2172 @param key_size: the size of the signature key.
2173 @return: tuple(verifying_key, signing_key)
2175 return TupleOf(str, str)
2178 class FileTooLargeError(Exception):
2181 class IValidatedThingProxy(Interface):
2183 """ Acquire a thing and validate it. Return a deferred which is eventually fired with
2184 self if the thing is valid or errbacked if it can't be acquired or validated. """
2186 class InsufficientVersionError(Exception):
2187 def __init__(self, needed, got):
2188 self.needed = needed
2191 return "InsufficientVersionError(need '%s', got %s)" % (self.needed,