3 from zope.interface import implements
4 from twisted.application import service
5 from twisted.internet import defer
6 from foolscap import Referenceable
7 from foolscap.eventual import eventually
8 from allmydata import upload, interfaces, storage, uri
9 from allmydata.util import idlib, log, observer, fileutil, hashutil
12 class NotEnoughWritersError(Exception):
16 class CHKCheckerAndUEBFetcher:
17 """I check to see if a file is already present in the grid. I also fetch
18 the URI Extension Block, which is useful for an uploading client who
19 wants to avoid the work of encryption and encoding.
21 I return False if the file is not completely healthy: i.e. if there are
22 less than 'N' shares present.
24 If the file is completely healthy, I return a tuple of (sharemap,
28 def __init__(self, peer_getter, storage_index, logparent=None):
29 self._peer_getter = peer_getter
30 self._found_shares = set()
31 self._storage_index = storage_index
36 self._logparent = logparent
38 def log(self, *args, **kwargs):
39 if 'facility' not in kwargs:
40 kwargs['facility'] = "tahoe.helper.chk.checkandUEBfetch"
41 if 'parent' not in kwargs:
42 kwargs['parent'] = self._logparent
43 return log.msg(*args, **kwargs)
46 d = self._get_all_shareholders(self._storage_index)
47 d.addCallback(self._get_uri_extension)
48 d.addCallback(self._done)
51 def _get_all_shareholders(self, storage_index):
53 for (peerid, ss) in self._peer_getter("storage", storage_index):
54 d = ss.callRemote("get_buckets", storage_index)
55 d.addCallbacks(self._got_response, self._got_error,
56 callbackArgs=(peerid,))
58 return defer.DeferredList(dl)
60 def _got_response(self, buckets, peerid):
61 # buckets is a dict: maps shum to an rref of the server who holds it
62 shnums_s = ",".join([str(shnum) for shnum in buckets])
63 self.log("got_response: [%s] has %d shares (%s)" %
64 (idlib.shortnodeid_b2a(peerid), len(buckets), shnums_s),
66 self._found_shares.update(buckets.keys())
68 if k not in self._sharemap:
69 self._sharemap[k] = []
70 self._sharemap[k].append(peerid)
71 self._readers.update( [ (bucket, peerid)
72 for bucket in buckets.values() ] )
74 def _got_error(self, f):
77 log.err(f, parent=self._logparent)
80 def _get_uri_extension(self, res):
81 # assume that we can pull the UEB from any share. If we get an error,
82 # declare the whole file unavailable.
84 self.log("no readers, so no UEB", level=log.NOISY)
86 b,peerid = self._readers.pop()
87 rbp = storage.ReadBucketProxy(b, peerid,
88 storage.si_b2a(self._storage_index))
89 d = rbp.startIfNecessary()
90 d.addCallback(lambda res: rbp.get_uri_extension())
91 d.addCallback(self._got_uri_extension)
92 d.addErrback(self._ueb_error)
95 def _got_uri_extension(self, ueb):
96 self.log("_got_uri_extension", level=log.NOISY)
97 self._ueb_hash = hashutil.uri_extension_hash(ueb)
98 self._ueb_data = uri.unpack_extension(ueb)
100 def _ueb_error(self, f):
101 # an error means the file is unavailable, but the overall check
103 self.log("UEB fetch failed", failure=f, level=log.WEIRD)
106 def _done(self, res):
108 found = len(self._found_shares)
109 total = self._ueb_data['total_shares']
110 self.log(format="got %(found)d shares of %(total)d",
111 found=found, total=total, level=log.NOISY)
113 # not all shares are present in the grid
114 self.log("not enough to qualify, file not found in grid",
117 # all shares are present
118 self.log("all shares present, file is found in grid",
120 return (self._sharemap, self._ueb_data, self._ueb_hash)
121 # no shares are present
122 self.log("unable to find UEB data, file not found in grid",
127 class CHKUploadHelper(Referenceable, upload.CHKUploader):
128 """I am the helper-server -side counterpart to AssistedUploader. I handle
129 peer selection, encoding, and share pushing. I read ciphertext from the
130 remote AssistedUploader.
132 implements(interfaces.RICHKUploadHelper)
134 def __init__(self, storage_index, helper,
135 incoming_file, encoding_file,
136 results, log_number):
137 self._storage_index = storage_index
138 self._helper = helper
139 self._incoming_file = incoming_file
140 self._encoding_file = encoding_file
141 upload_id = storage.si_b2a(storage_index)[:5]
142 self._log_number = log_number
143 self._results = results
144 self._upload_status = upload.UploadStatus()
145 self._upload_status.set_helper(False)
146 self._upload_status.set_storage_index(storage_index)
147 self._upload_status.set_status("fetching ciphertext")
148 self._upload_status.set_progress(0, 1.0)
149 self._helper.log("CHKUploadHelper starting for SI %s" % upload_id,
152 self._client = helper.parent
153 self._fetcher = CHKCiphertextFetcher(self, incoming_file, encoding_file,
155 self._reader = LocalCiphertextReader(self, storage_index, encoding_file)
156 self._finished_observers = observer.OneShotObserverList()
158 d = self._fetcher.when_done()
159 d.addCallback(lambda res: self._reader.start())
160 d.addCallback(lambda res: self.start_encrypted(self._reader))
161 d.addCallback(self._finished)
162 d.addErrback(self._failed)
164 def log(self, *args, **kwargs):
165 if 'facility' not in kwargs:
166 kwargs['facility'] = "tahoe.helper.chk"
167 return upload.CHKUploader.log(self, *args, **kwargs)
170 self._started = time.time()
171 # determine if we need to upload the file. If so, return ({},self) .
172 # If not, return (UploadResults,None) .
173 self.log("deciding whether to upload the file or not", level=log.NOISY)
174 if os.path.exists(self._encoding_file):
175 # we have the whole file, and we might be encoding it (or the
176 # encode/upload might have failed, and we need to restart it).
177 self.log("ciphertext already in place", level=log.UNUSUAL)
178 return (self._results, self)
179 if os.path.exists(self._incoming_file):
180 # we have some of the file, but not all of it (otherwise we'd be
181 # encoding). The caller might be useful.
182 self.log("partial ciphertext already present", level=log.UNUSUAL)
183 return (self._results, self)
184 # we don't remember uploading this file
185 self.log("no ciphertext yet", level=log.NOISY)
186 return (self._results, self)
188 def remote_upload(self, reader):
189 # reader is an RIEncryptedUploadable. I am specified to return an
190 # UploadResults dictionary.
192 # let our fetcher pull ciphertext from the reader.
193 self._fetcher.add_reader(reader)
195 self._reader.add_reader(reader)
197 # and inform the client when the upload has finished
198 return self._finished_observers.when_fired()
200 def _finished(self, res):
201 (uri_extension_hash, needed_shares, total_shares, size) = res
203 r.uri_extension_hash = uri_extension_hash
204 f_times = self._fetcher.get_times()
205 r.timings["cumulative_fetch"] = f_times["cumulative_fetch"]
206 r.ciphertext_fetched = self._fetcher.get_ciphertext_fetched()
207 r.timings["total_fetch"] = f_times["total"]
209 os.unlink(self._encoding_file)
210 self._finished_observers.fire(r)
211 self._helper.upload_finished(self._storage_index, size)
214 def _failed(self, f):
215 self.log(format="CHKUploadHelper(%(si)s) failed",
216 si=storage.si_b2a(self._storage_index)[:5],
219 self._finished_observers.fire(f)
220 self._helper.upload_finished(self._storage_index, 0)
223 class AskUntilSuccessMixin:
224 # create me with a _reader array
227 def add_reader(self, reader):
228 self._readers.append(reader)
230 def call(self, *args, **kwargs):
231 if not self._readers:
232 raise NotEnoughWritersError("ran out of assisted uploaders, last failure was %s" % self._last_failure)
233 rr = self._readers[0]
234 d = rr.callRemote(*args, **kwargs)
236 self._last_failure = f
237 if rr in self._readers:
238 self._readers.remove(rr)
239 self._upload_helper.log("call to assisted uploader %s failed" % rr,
240 failure=f, level=log.UNUSUAL)
241 # we can try again with someone else who's left
242 return self.call(*args, **kwargs)
246 class CHKCiphertextFetcher(AskUntilSuccessMixin):
247 """I use one or more remote RIEncryptedUploadable instances to gather
248 ciphertext on disk. When I'm done, the file I create can be used by a
249 LocalCiphertextReader to satisfy the ciphertext needs of a CHK upload
252 I begin pulling ciphertext as soon as a reader is added. I remove readers
253 when they have any sort of error. If the last reader is removed, I fire
254 my when_done() Deferred with a failure.
256 I fire my when_done() Deferred (with None) immediately after I have moved
257 the ciphertext to 'encoded_file'.
260 def __init__(self, helper, incoming_file, encoded_file, logparent):
261 self._upload_helper = helper
262 self._incoming_file = incoming_file
263 self._encoding_file = encoded_file
264 self._log_parent = logparent
265 self._done_observers = observer.OneShotObserverList()
267 self._started = False
270 "cumulative_fetch": 0.0,
273 self._ciphertext_fetched = 0
275 def log(self, *args, **kwargs):
276 if "facility" not in kwargs:
277 kwargs["facility"] = "tahoe.helper.chkupload.fetch"
278 if "parent" not in kwargs:
279 kwargs["parent"] = self._log_parent
280 return log.msg(*args, **kwargs)
282 def add_reader(self, reader):
283 AskUntilSuccessMixin.add_reader(self, reader)
284 eventually(self._start)
290 started = time.time()
292 if os.path.exists(self._encoding_file):
293 self.log("ciphertext already present, bypassing fetch",
295 # we'll still need the plaintext hashes (when
296 # LocalCiphertextReader.get_plaintext_hashtree_leaves() is
297 # called), and currently the easiest way to get them is to ask
298 # the sender for the last byte of ciphertext. That will provoke
299 # them into reading and hashing (but not sending) everything
301 have = os.stat(self._encoding_file)[stat.ST_SIZE]
302 d = self.call("read_encrypted", have-1, 1)
303 d.addCallback(self._done2, started)
306 # first, find out how large the file is going to be
307 d = self.call("get_size")
308 d.addCallback(self._got_size)
309 d.addCallback(self._start_reading)
310 d.addCallback(self._done)
311 d.addCallback(self._done2, started)
312 d.addErrback(self._failed)
314 def _got_size(self, size):
315 self.log("total size is %d bytes" % size, level=log.NOISY)
316 self._upload_helper._upload_status.set_size(size)
317 self._expected_size = size
319 def _start_reading(self, res):
320 # then find out how much crypttext we have on disk
321 if os.path.exists(self._incoming_file):
322 self._have = os.stat(self._incoming_file)[stat.ST_SIZE]
323 self.log("we already have %d bytes" % self._have, level=log.NOISY)
326 self.log("we do not have any ciphertext yet", level=log.NOISY)
327 self.log("starting ciphertext fetch", level=log.NOISY)
328 self._f = open(self._incoming_file, "ab")
330 # now loop to pull the data from the readers
333 # this Deferred will be fired once the last byte has been written to
337 # read data in 50kB chunks. We should choose a more considered number
338 # here, possibly letting the client specify it. The goal should be to
339 # keep the RTT*bandwidth to be less than 10% of the chunk size, to reduce
340 # the upload bandwidth lost because this protocol is non-windowing. Too
341 # large, however, means more memory consumption for both ends. Something
342 # that can be transferred in, say, 10 seconds sounds about right. On my
343 # home DSL line (50kBps upstream), that suggests 500kB. Most lines are
344 # slower, maybe 10kBps, which suggests 100kB, and that's a bit more
345 # memory than I want to hang on to, so I'm going to go with 50kB and see
349 def _loop(self, fire_when_done):
350 # this slightly weird structure is needed because Deferreds don't do
351 # tail-recursion, so it is important to let each one retire promptly.
352 # Simply chaining them will cause a stack overflow at the end of a
353 # transfer that involves more than a few hundred chunks.
354 # 'fire_when_done' lives a long time, but the Deferreds returned by
355 # the inner _fetch() call do not.
357 d = defer.maybeDeferred(self._fetch)
359 elapsed = time.time() - start
360 self._times["cumulative_fetch"] += elapsed
362 self.log("finished reading ciphertext", level=log.NOISY)
363 fire_when_done.callback(None)
365 self._loop(fire_when_done)
367 self.log("ciphertext read failed", failure=f, level=log.UNUSUAL)
368 fire_when_done.errback(f)
369 d.addCallbacks(_done, _err)
373 needed = self._expected_size - self._have
374 fetch_size = min(needed, self.CHUNK_SIZE)
376 self._upload_helper._upload_status.set_progress(1, 1.0)
377 return True # all done
379 if self._expected_size:
380 percent = 1.0 * (self._have+fetch_size) / self._expected_size
381 self.log(format="fetching %(start)d-%(end)d of %(total)d (%(percent)d%%)",
383 end=self._have+fetch_size,
384 total=self._expected_size,
385 percent=int(100.0*percent),
387 d = self.call("read_encrypted", self._have, fetch_size)
388 def _got_data(ciphertext_v):
389 for data in ciphertext_v:
391 self._have += len(data)
392 self._ciphertext_fetched += len(data)
393 self._upload_helper._helper.count("chk_upload_helper.fetched_bytes", len(data))
394 self._upload_helper._upload_status.set_progress(1, percent)
395 return False # not done
396 d.addCallback(_got_data)
399 def _done(self, res):
402 self.log(format="done fetching ciphertext, size=%(size)d",
403 size=os.stat(self._incoming_file)[stat.ST_SIZE],
405 os.rename(self._incoming_file, self._encoding_file)
407 def _done2(self, _ignored, started):
408 self.log("done2", level=log.NOISY)
409 elapsed = time.time() - started
410 self._times["total"] = elapsed
412 self._done_observers.fire(None)
414 def _failed(self, f):
418 self._done_observers.fire(f)
421 return self._done_observers.when_fired()
426 def get_ciphertext_fetched(self):
427 return self._ciphertext_fetched
430 class LocalCiphertextReader(AskUntilSuccessMixin):
431 implements(interfaces.IEncryptedUploadable)
433 def __init__(self, upload_helper, storage_index, encoding_file):
435 self._upload_helper = upload_helper
436 self._storage_index = storage_index
437 self._encoding_file = encoding_file
440 def set_upload_status(self, upload_status):
441 self._status = interfaces.IUploadStatus(upload_status)
444 self._upload_helper._upload_status.set_status("pushing")
445 self._size = os.stat(self._encoding_file)[stat.ST_SIZE]
446 self.f = open(self._encoding_file, "rb")
449 return defer.succeed(self._size)
451 def get_all_encoding_parameters(self):
452 return self.call("get_all_encoding_parameters")
454 def get_storage_index(self):
455 return defer.succeed(self._storage_index)
457 def read_encrypted(self, length, hash_only):
458 assert hash_only is False
459 d = defer.maybeDeferred(self.f.read, length)
460 d.addCallback(lambda data: [data])
462 def get_plaintext_hashtree_leaves(self, first, last, num_segments):
463 return self.call("get_plaintext_hashtree_leaves", first, last,
465 def get_plaintext_hash(self):
466 return self.call("get_plaintext_hash")
469 # ??. I'm not sure if it makes sense to forward the close message.
470 return self.call("close")
474 class Helper(Referenceable, service.MultiService):
475 implements(interfaces.RIHelper, interfaces.IStatsProducer)
476 # this is the non-distributed version. When we need to have multiple
477 # helpers, this object will become the HelperCoordinator, and will query
478 # the farm of Helpers to see if anyone has the storage_index of interest,
479 # and send the request off to them. If nobody has it, we'll choose a
483 chk_upload_helper_class = CHKUploadHelper
484 MAX_UPLOAD_STATUSES = 10
486 def __init__(self, basedir, stats_provider=None):
487 self._basedir = basedir
488 self._chk_incoming = os.path.join(basedir, "CHK_incoming")
489 self._chk_encoding = os.path.join(basedir, "CHK_encoding")
490 fileutil.make_dirs(self._chk_incoming)
491 fileutil.make_dirs(self._chk_encoding)
492 self._active_uploads = {}
493 self._recent_upload_statuses = []
494 self.stats_provider = stats_provider
496 stats_provider.register_producer(self)
497 self._counters = {"chk_upload_helper.upload_requests": 0,
498 "chk_upload_helper.upload_already_present": 0,
499 "chk_upload_helper.upload_need_upload": 0,
500 "chk_upload_helper.fetched_bytes": 0,
501 "chk_upload_helper.encoded_bytes": 0,
503 service.MultiService.__init__(self)
505 def setServiceParent(self, parent):
506 service.MultiService.setServiceParent(self, parent)
508 def log(self, *args, **kwargs):
509 if 'facility' not in kwargs:
510 kwargs['facility'] = "tahoe.helper"
511 return self.parent.log(*args, **kwargs)
513 def count(self, key, value=1):
514 if self.stats_provider:
515 self.stats_provider.count(key, value)
516 self._counters[key] += value
519 OLD = 86400*2 # 48hours
521 inc_count = inc_size = inc_size_old = 0
522 enc_count = enc_size = enc_size_old = 0
523 inc = os.listdir(self._chk_incoming)
524 enc = os.listdir(self._chk_encoding)
526 s = os.stat(os.path.join(self._chk_incoming, f))
527 size = s[stat.ST_SIZE]
528 mtime = s[stat.ST_MTIME]
531 if now - mtime > OLD:
534 s = os.stat(os.path.join(self._chk_encoding, f))
535 size = s[stat.ST_SIZE]
536 mtime = s[stat.ST_MTIME]
539 if now - mtime > OLD:
541 stats = { 'chk_upload_helper.active_uploads': len(self._active_uploads),
542 'chk_upload_helper.incoming_count': inc_count,
543 'chk_upload_helper.incoming_size': inc_size,
544 'chk_upload_helper.incoming_size_old': inc_size_old,
545 'chk_upload_helper.encoding_count': enc_count,
546 'chk_upload_helper.encoding_size': enc_size,
547 'chk_upload_helper.encoding_size_old': enc_size_old,
549 stats.update(self._counters)
552 def remote_upload_chk(self, storage_index):
553 self.count("chk_upload_helper.upload_requests")
554 r = upload.UploadResults()
555 started = time.time()
556 si_s = storage.si_b2a(storage_index)
557 lp = self.log(format="helper: upload_chk query for SI %(si)s", si=si_s)
558 incoming_file = os.path.join(self._chk_incoming, si_s)
559 encoding_file = os.path.join(self._chk_encoding, si_s)
560 if storage_index in self._active_uploads:
561 self.log("upload is currently active", parent=lp)
562 uh = self._active_uploads[storage_index]
565 d = self._check_for_chk_already_in_grid(storage_index, r, lp)
566 def _checked(already_present):
567 elapsed = time.time() - started
568 r.timings['existence_check'] = elapsed
570 # the necessary results are placed in the UploadResults
571 self.count("chk_upload_helper.upload_already_present")
572 self.log("file already found in grid", parent=lp)
575 self.count("chk_upload_helper.upload_need_upload")
576 # the file is not present in the grid, by which we mean there are
577 # less than 'N' shares available.
578 self.log("unable to find file in the grid", parent=lp,
580 # We need an upload helper. Check our active uploads again in
581 # case there was a race.
582 if storage_index in self._active_uploads:
583 self.log("upload is currently active", parent=lp)
584 uh = self._active_uploads[storage_index]
586 self.log("creating new upload helper", parent=lp)
587 uh = self.chk_upload_helper_class(storage_index, self,
588 incoming_file, encoding_file,
590 self._active_uploads[storage_index] = uh
592 d.addCallback(_checked)
594 self.log("error while checking for chk-already-in-grid",
595 failure=f, level=log.WEIRD, parent=lp)
600 def _check_for_chk_already_in_grid(self, storage_index, results, lp):
601 # see if this file is already in the grid
602 lp2 = self.log("doing a quick check+UEBfetch",
603 parent=lp, level=log.NOISY)
604 c = CHKCheckerAndUEBFetcher(self.parent.get_permuted_peers,
609 (sharemap, ueb_data, ueb_hash) = res
610 self.log("found file in grid", level=log.NOISY, parent=lp)
611 results.uri_extension_hash = ueb_hash
612 results.sharemap = {}
613 for shnum, peerids in sharemap.items():
614 peers_s = ",".join(["[%s]" % idlib.shortnodeid_b2a(peerid)
615 for peerid in peerids])
616 results.sharemap[shnum] = "Found on " + peers_s
617 results.uri_extension_data = ueb_data
618 results.preexisting_shares = len(sharemap)
619 results.pushed_shares = 0
622 d.addCallback(_checked)
625 def upload_finished(self, storage_index, size):
626 self.count("chk_upload_helper.encoded_bytes", size)
627 uh = self._active_uploads[storage_index]
628 del self._active_uploads[storage_index]
629 s = uh.get_upload_status()
631 self._recent_upload_statuses.append(s)
632 while len(self._recent_upload_statuses) > self.MAX_UPLOAD_STATUSES:
633 self._recent_upload_statuses.pop(0)
635 def get_active_upload_statuses(self):
636 return [u.get_upload_status() for u in self._active_uploads.values()]
638 def get_recent_upload_statuses(self):
639 return self._recent_upload_statuses