1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
|
# -*- coding: utf-8 -*-
# Copyright (C) 2013 Yahoo! Inc. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License"); you may
# not use this file except in compliance with the License. You may obtain
# a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
# License for the specific language governing permissions and limitations
# under the License.
import collections
import contextlib
import functools
import sys
import threading
from concurrent import futures
from kazoo import exceptions as k_exceptions
from kazoo.protocol import paths as k_paths
from kazoo.recipe import watchers
from oslo_serialization import jsonutils
from oslo_utils import excutils
from oslo_utils import uuidutils
import six
from taskflow import exceptions as excp
from taskflow.jobs import base
from taskflow import logging
from taskflow import states
from taskflow.types import timing as tt
from taskflow.utils import kazoo_utils
from taskflow.utils import lock_utils
from taskflow.utils import misc
LOG = logging.getLogger(__name__)
def check_who(meth):
"""Decorator that checks the expected owner type & value restrictions."""
@six.wraps(meth)
def wrapper(self, job, who, *args, **kwargs):
if not isinstance(who, six.string_types):
raise TypeError("Job applicant must be a string type")
if len(who) == 0:
raise ValueError("Job applicant must be non-empty")
return meth(self, job, who, *args, **kwargs)
return wrapper
class ZookeeperJob(base.Job):
"""A zookeeper job."""
def __init__(self, name, board, client, backend, path,
uuid=None, details=None, book=None, book_data=None,
created_on=None):
super(ZookeeperJob, self).__init__(name, uuid=uuid, details=details)
self._board = board
self._book = book
if not book_data:
book_data = {}
self._book_data = book_data
self._client = client
self._backend = backend
if all((self._book, self._book_data)):
raise ValueError("Only one of 'book_data' or 'book'"
" can be provided")
self._path = k_paths.normpath(path)
self._lock_path = self._path + board.LOCK_POSTFIX
self._created_on = created_on
self._node_not_found = False
basename = k_paths.basename(self._path)
self._root = self._path[0:-len(basename)]
self._sequence = int(basename[len(board.JOB_PREFIX):])
@property
def lock_path(self):
return self._lock_path
@property
def path(self):
return self._path
@property
def sequence(self):
"""Sequence number of the current job."""
return self._sequence
@property
def root(self):
"""The parent path of the job in zookeeper."""
return self._root
def _get_node_attr(self, path, attr_name, trans_func=None):
try:
_data, node_stat = self._client.get(path)
attr = getattr(node_stat, attr_name)
if trans_func is not None:
return trans_func(attr)
else:
return attr
except k_exceptions.NoNodeError:
excp.raise_with_cause(
excp.NotFound,
"Can not fetch the %r attribute of job %s (%s),"
" path %s not found" % (attr_name, self.uuid,
self.path, path))
except self._client.handler.timeout_exception:
excp.raise_with_cause(
excp.JobFailure,
"Can not fetch the %r attribute of job %s (%s),"
" operation timed out" % (attr_name, self.uuid, self.path))
except k_exceptions.SessionExpiredError:
excp.raise_with_cause(
excp.JobFailure,
"Can not fetch the %r attribute of job %s (%s),"
" session expired" % (attr_name, self.uuid, self.path))
except (AttributeError, k_exceptions.KazooException):
excp.raise_with_cause(
excp.JobFailure,
"Can not fetch the %r attribute of job %s (%s),"
" internal error" % (attr_name, self.uuid, self.path))
@property
def last_modified(self):
modified_on = None
try:
if not self._node_not_found:
modified_on = self._get_node_attr(
self.path, 'mtime',
trans_func=misc.millis_to_datetime)
except excp.NotFound:
self._node_not_found = True
return modified_on
@property
def created_on(self):
# This one we can cache (since it won't change after creation).
if self._node_not_found:
return None
if self._created_on is None:
try:
self._created_on = self._get_node_attr(
self.path, 'ctime',
trans_func=misc.millis_to_datetime)
except excp.NotFound:
self._node_not_found = True
return self._created_on
@property
def board(self):
return self._board
def _load_book(self):
book_uuid = self.book_uuid
if self._backend is not None and book_uuid is not None:
# TODO(harlowja): we are currently limited by assuming that the
# job posted has the same backend as this loader (to start this
# seems to be a ok assumption, and can be adjusted in the future
# if we determine there is a use-case for multi-backend loaders,
# aka a registry of loaders).
with contextlib.closing(self._backend.get_connection()) as conn:
return conn.get_logbook(book_uuid)
# No backend to fetch from or no uuid specified
return None
@property
def state(self):
return self._fetch_state()
def _fetch_state(self):
owner = self.board.find_owner(self)
job_data = {}
try:
raw_data, _data_stat = self._client.get(self.path)
job_data = misc.decode_json(raw_data)
except k_exceptions.NoNodeError:
pass
except k_exceptions.SessionExpiredError:
excp.raise_with_cause(
excp.JobFailure,
"Can not fetch the state of %s,"
" session expired" % (self.uuid))
except self._client.handler.timeout_exception:
excp.raise_with_cause(
excp.JobFailure,
"Can not fetch the state of %s,"
" operation timed out" % (self.uuid))
except k_exceptions.KazooException:
excp.raise_with_cause(
excp.JobFailure,
"Can not fetch the state of %s,"
" internal error" % (self.uuid))
if not job_data:
# No data this job has been completed (the owner that we might have
# fetched will not be able to be fetched again, since the job node
# is a parent node of the owner/lock node).
return states.COMPLETE
if not owner:
# No owner, but data, still work to be done.
return states.UNCLAIMED
return states.CLAIMED
def __lt__(self, other):
if self.root == other.root:
return self.sequence < other.sequence
else:
return self.root < other.root
def __hash__(self):
return hash(self.path)
@property
def book(self):
if self._book is None:
self._book = self._load_book()
return self._book
@property
def book_uuid(self):
if self._book:
return self._book.uuid
if self._book_data:
return self._book_data.get('uuid')
else:
return None
@property
def book_name(self):
if self._book:
return self._book.name
if self._book_data:
return self._book_data.get('name')
else:
return None
class ZookeeperJobBoardIterator(six.Iterator):
"""Iterator over a zookeeper jobboard that iterates over potential jobs.
It supports the following attributes/constructor arguments:
* ``ensure_fresh``: boolean that requests that during every fetch of a new
set of jobs this will cause the iterator to force the backend to
refresh (ensuring that the jobboard has the most recent job listings).
* ``only_unclaimed``: boolean that indicates whether to only iterate
over unclaimed jobs.
"""
_UNCLAIMED_JOB_STATES = (
states.UNCLAIMED,
)
_JOB_STATES = (
states.UNCLAIMED,
states.COMPLETE,
states.CLAIMED,
)
def __init__(self, board, only_unclaimed=False, ensure_fresh=False):
self._board = board
self._jobs = collections.deque()
self._fetched = False
self.ensure_fresh = ensure_fresh
self.only_unclaimed = only_unclaimed
@property
def board(self):
return self._board
def __iter__(self):
return self
def _next_job(self):
if self.only_unclaimed:
allowed_states = self._UNCLAIMED_JOB_STATES
else:
allowed_states = self._JOB_STATES
job = None
while self._jobs and job is None:
maybe_job = self._jobs.popleft()
try:
if maybe_job.state in allowed_states:
job = maybe_job
except excp.JobFailure:
LOG.warn("Failed determining the state of job: %s (%s)",
maybe_job.uuid, maybe_job.path, exc_info=True)
except excp.NotFound:
self._board._remove_job(maybe_job.path)
return job
def __next__(self):
if not self._jobs:
if not self._fetched:
jobs = self._board._fetch_jobs(ensure_fresh=self.ensure_fresh)
self._jobs.extend(jobs)
self._fetched = True
job = self._next_job()
if job is None:
raise StopIteration
else:
return job
class ZookeeperJobBoard(base.NotifyingJobBoard):
"""A jobboard backend by zookeeper.
Powered by the `kazoo <http://kazoo.readthedocs.org/>`_ library.
This jobboard creates *sequenced* persistent znodes in a directory in
zookeeper (that directory defaults to ``/taskflow/jobs``) and uses
zookeeper watches to notify other jobboards that the job which was posted
using the :meth:`.post` method (this creates a znode with contents/details
in json). The users of those jobboard(s) (potentially on disjoint sets of
machines) can then iterate over the available jobs and decide if they want
to attempt to claim one of the jobs they have iterated over. If so they
will then attempt to contact zookeeper and they will attempt to create a
ephemeral znode using the name of the persistent znode + ".lock" as a
postfix. If the entity trying to use the jobboard to :meth:`.claim` the
job is able to create a ephemeral znode with that name then it will be
allowed (and expected) to perform whatever *work* the contents of that
job described. Once finished the ephemeral znode and persistent znode may
be deleted (if successfully completed) in a single transaction or if not
successful (or the entity that claimed the znode dies) the ephemeral
znode will be released (either manually by using :meth:`.abandon` or
automatically by zookeeper when the ephemeral node and associated session
is deemed to have been lost).
"""
#: Transaction support was added in 3.4.0 so we need at least that version.
MIN_ZK_VERSION = (3, 4, 0)
#: Znode **postfix** that lock entries have.
LOCK_POSTFIX = ".lock"
#: Znode child path created under root path that contains trashed jobs.
TRASH_FOLDER = ".trash"
#: Znode **prefix** that job entries have.
JOB_PREFIX = 'job'
#: Default znode path used for jobs (data, locks...).
DEFAULT_PATH = "/taskflow/jobs"
def __init__(self, name, conf,
client=None, persistence=None, emit_notifications=True):
super(ZookeeperJobBoard, self).__init__(name, conf)
if client is not None:
self._client = client
self._owned = False
else:
self._client = kazoo_utils.make_client(self._conf)
self._owned = True
path = str(conf.get("path", self.DEFAULT_PATH))
if not path:
raise ValueError("Empty zookeeper path is disallowed")
if not k_paths.isabs(path):
raise ValueError("Zookeeper path must be absolute")
self._path = path
self._trash_path = self._path.replace(k_paths.basename(self._path),
self.TRASH_FOLDER)
# The backend to load the full logbooks from, since whats sent over
# the zookeeper data connection is only the logbook uuid and name, and
# not currently the full logbook (later when a zookeeper backend
# appears we can likely optimize for that backend usage by directly
# reading from the path where the data is stored, if we want).
self._persistence = persistence
# Misc. internal details
self._known_jobs = {}
self._job_cond = threading.Condition()
self._open_close_lock = threading.RLock()
self._client.add_listener(self._state_change_listener)
self._bad_paths = frozenset([path])
self._job_watcher = None
# Since we use sequenced ids this will be the path that the sequences
# are prefixed with, for example, job0000000001, job0000000002, ...
self._job_base = k_paths.join(path, self.JOB_PREFIX)
self._worker = None
self._emit_notifications = bool(emit_notifications)
self._connected = False
def _emit(self, state, details):
# Submit the work to the executor to avoid blocking the kazoo threads
# and queue(s)...
worker = self._worker
if worker is None:
return
try:
worker.submit(self.notifier.notify, state, details)
except RuntimeError:
# Notification thread is shutdown just skip submitting a
# notification...
pass
@property
def path(self):
return self._path
@property
def trash_path(self):
return self._trash_path
@property
def job_count(self):
return len(self._known_jobs)
def _fetch_jobs(self, ensure_fresh=False):
if ensure_fresh:
self._force_refresh()
with self._job_cond:
return sorted(six.itervalues(self._known_jobs))
def _force_refresh(self):
try:
children = self._client.get_children(self.path)
except self._client.handler.timeout_exception:
excp.raise_with_cause(excp.JobFailure,
"Refreshing failure, operation timed out")
except k_exceptions.SessionExpiredError:
excp.raise_with_cause(excp.JobFailure,
"Refreshing failure, session expired")
except k_exceptions.NoNodeError:
pass
except k_exceptions.KazooException:
excp.raise_with_cause(excp.JobFailure,
"Refreshing failure, internal error")
else:
self._on_job_posting(children, delayed=False)
def iterjobs(self, only_unclaimed=False, ensure_fresh=False):
return ZookeeperJobBoardIterator(self,
only_unclaimed=only_unclaimed,
ensure_fresh=ensure_fresh)
def _remove_job(self, path):
if path not in self._known_jobs:
return
with self._job_cond:
job = self._known_jobs.pop(path, None)
if job is not None:
LOG.debug("Removed job that was at path '%s'", path)
self._emit(base.REMOVAL, details={'job': job})
def _process_child(self, path, request):
"""Receives the result of a child data fetch request."""
job = None
try:
raw_data, node_stat = request.get()
job_data = misc.decode_json(raw_data)
created_on = misc.millis_to_datetime(node_stat.ctime)
except (ValueError, TypeError, KeyError):
LOG.warn("Incorrectly formatted job data found at path: %s",
path, exc_info=True)
except self._client.handler.timeout_exception:
LOG.warn("Operation timed out fetching job data from path: %s",
path, exc_info=True)
except k_exceptions.SessionExpiredError:
LOG.warn("Session expired fetching job data from path: %s", path,
exc_info=True)
except k_exceptions.NoNodeError:
LOG.debug("No job node found at path: %s, it must have"
" disappeared or was removed", path)
except k_exceptions.KazooException:
LOG.warn("Internal error fetching job data from path: %s",
path, exc_info=True)
else:
with self._job_cond:
# Now we can offically check if someone already placed this
# jobs information into the known job set (if it's already
# existing then just leave it alone).
if path not in self._known_jobs:
job = ZookeeperJob(job_data['name'], self,
self._client, self._persistence, path,
uuid=job_data['uuid'],
book_data=job_data.get("book"),
details=job_data.get("details", {}),
created_on=created_on)
self._known_jobs[path] = job
self._job_cond.notify_all()
if job is not None:
self._emit(base.POSTED, details={'job': job})
def _on_job_posting(self, children, delayed=True):
LOG.debug("Got children %s under path %s", children, self.path)
child_paths = []
for c in children:
if (c.endswith(self.LOCK_POSTFIX) or
not c.startswith(self.JOB_PREFIX)):
# Skip lock paths or non-job-paths (these are not valid jobs)
continue
child_paths.append(k_paths.join(self.path, c))
# Figure out what we really should be investigating and what we
# shouldn't (remove jobs that exist in our local version, but don't
# exist in the children anymore) and accumulate all paths that we
# need to trigger population of (without holding the job lock).
investigate_paths = []
pending_removals = []
with self._job_cond:
for path in six.iterkeys(self._known_jobs):
if path not in child_paths:
pending_removals.append(path)
for path in child_paths:
if path in self._bad_paths:
continue
# This pre-check will *not* guarantee that we will not already
# have the job (if it's being populated elsewhere) but it will
# reduce the amount of duplicated requests in general; later when
# the job information has been populated we will ensure that we
# are not adding duplicates into the currently known jobs...
if path in self._known_jobs:
continue
if path not in investigate_paths:
investigate_paths.append(path)
if pending_removals:
with self._job_cond:
for path in pending_removals:
self._remove_job(path)
for path in investigate_paths:
# Fire off the request to populate this job.
#
# This method is *usually* called from a asynchronous handler so
# it's better to exit from this quickly to allow other asynchronous
# handlers to be executed.
request = self._client.get_async(path)
if delayed:
request.rawlink(functools.partial(self._process_child, path))
else:
self._process_child(path, request)
def post(self, name, book=None, details=None):
def format_posting(job_uuid):
posting = {
'uuid': job_uuid,
'name': name,
}
if details:
posting['details'] = details
else:
posting['details'] = {}
if book is not None:
posting['book'] = {
'name': book.name,
'uuid': book.uuid,
}
return posting
# NOTE(harlowja): Jobs are not ephemeral, they will persist until they
# are consumed (this may change later, but seems safer to do this until
# further notice).
job_uuid = uuidutils.generate_uuid()
with self._wrap(job_uuid, None,
"Posting failure: %s", ensure_known=False):
job_posting = format_posting(job_uuid)
job_posting = misc.binary_encode(jsonutils.dumps(job_posting))
job_path = self._client.create(self._job_base,
value=job_posting,
sequence=True,
ephemeral=False)
job = ZookeeperJob(name, self, self._client,
self._persistence, job_path,
book=book, details=details,
uuid=job_uuid)
with self._job_cond:
self._known_jobs[job_path] = job
self._job_cond.notify_all()
self._emit(base.POSTED, details={'job': job})
return job
@check_who
def claim(self, job, who):
def _unclaimable_try_find_owner(cause):
try:
owner = self.find_owner(job)
except Exception:
owner = None
if owner:
message = "Job %s already claimed by '%s'" % (job.uuid, owner)
else:
message = "Job %s already claimed" % (job.uuid)
excp.raise_with_cause(excp.UnclaimableJob,
message, cause=cause)
with self._wrap(job.uuid, job.path, "Claiming failure: %s"):
# NOTE(harlowja): post as json which will allow for future changes
# more easily than a raw string/text.
value = jsonutils.dumps({
'owner': who,
})
# Ensure the target job is still existent (at the right version).
job_data, job_stat = self._client.get(job.path)
txn = self._client.transaction()
# This will abort (and not create the lock) if the job has been
# removed (somehow...) or updated by someone else to a different
# version...
txn.check(job.path, version=job_stat.version)
txn.create(job.lock_path, value=misc.binary_encode(value),
ephemeral=True)
try:
kazoo_utils.checked_commit(txn)
except k_exceptions.NodeExistsError as e:
_unclaimable_try_find_owner(e)
except kazoo_utils.KazooTransactionException as e:
if len(e.failures) < 2:
raise
else:
if isinstance(e.failures[0], k_exceptions.NoNodeError):
excp.raise_with_cause(
excp.NotFound,
"Job %s not found to be claimed" % job.uuid,
cause=e.failures[0])
if isinstance(e.failures[1], k_exceptions.NodeExistsError):
_unclaimable_try_find_owner(e.failures[1])
else:
excp.raise_with_cause(
excp.UnclaimableJob,
"Job %s claim failed due to transaction"
" not succeeding" % (job.uuid), cause=e)
@contextlib.contextmanager
def _wrap(self, job_uuid, job_path,
fail_msg_tpl="Failure: %s", ensure_known=True):
if job_path:
fail_msg_tpl += " (%s)" % (job_path)
if ensure_known:
if not job_path:
raise ValueError("Unable to check if %r is a known path"
% (job_path))
if job_path not in self._known_jobs:
fail_msg_tpl += ", unknown job"
raise excp.NotFound(fail_msg_tpl % (job_uuid))
try:
yield
except self._client.handler.timeout_exception:
fail_msg_tpl += ", operation timed out"
excp.raise_with_cause(excp.JobFailure, fail_msg_tpl % (job_uuid))
except k_exceptions.SessionExpiredError:
fail_msg_tpl += ", session expired"
excp.raise_with_cause(excp.JobFailure, fail_msg_tpl % (job_uuid))
except k_exceptions.NoNodeError:
fail_msg_tpl += ", unknown job"
excp.raise_with_cause(excp.NotFound, fail_msg_tpl % (job_uuid))
except k_exceptions.KazooException:
fail_msg_tpl += ", internal error"
excp.raise_with_cause(excp.JobFailure, fail_msg_tpl % (job_uuid))
def find_owner(self, job):
with self._wrap(job.uuid, job.path, "Owner query failure: %s"):
try:
self._client.sync(job.lock_path)
raw_data, _lock_stat = self._client.get(job.lock_path)
data = misc.decode_json(raw_data)
owner = data.get("owner")
except k_exceptions.NoNodeError:
owner = None
return owner
def _get_owner_and_data(self, job):
lock_data, lock_stat = self._client.get(job.lock_path)
job_data, job_stat = self._client.get(job.path)
return (misc.decode_json(lock_data), lock_stat,
misc.decode_json(job_data), job_stat)
@check_who
def consume(self, job, who):
with self._wrap(job.uuid, job.path, "Consumption failure: %s"):
try:
owner_data = self._get_owner_and_data(job)
lock_data, lock_stat, data, data_stat = owner_data
except k_exceptions.NoNodeError:
excp.raise_with_cause(excp.JobFailure,
"Can not consume a job %s"
" which we can not determine"
" the owner of" % (job.uuid))
if lock_data.get("owner") != who:
raise excp.JobFailure("Can not consume a job %s"
" which is not owned by %s"
% (job.uuid, who))
txn = self._client.transaction()
txn.delete(job.lock_path, version=lock_stat.version)
txn.delete(job.path, version=data_stat.version)
kazoo_utils.checked_commit(txn)
self._remove_job(job.path)
@check_who
def abandon(self, job, who):
with self._wrap(job.uuid, job.path, "Abandonment failure: %s"):
try:
owner_data = self._get_owner_and_data(job)
lock_data, lock_stat, data, data_stat = owner_data
except k_exceptions.NoNodeError:
excp.raise_with_cause(excp.JobFailure,
"Can not abandon a job %s"
" which we can not determine"
" the owner of" % (job.uuid))
if lock_data.get("owner") != who:
raise excp.JobFailure("Can not abandon a job %s"
" which is not owned by %s"
% (job.uuid, who))
txn = self._client.transaction()
txn.delete(job.lock_path, version=lock_stat.version)
kazoo_utils.checked_commit(txn)
@check_who
def trash(self, job, who):
with self._wrap(job.uuid, job.path, "Trash failure: %s"):
try:
owner_data = self._get_owner_and_data(job)
lock_data, lock_stat, data, data_stat = owner_data
except k_exceptions.NoNodeError:
excp.raise_with_cause(excp.JobFailure,
"Can not trash a job %s"
" which we can not determine"
" the owner of" % (job.uuid))
if lock_data.get("owner") != who:
raise excp.JobFailure("Can not trash a job %s"
" which is not owned by %s"
% (job.uuid, who))
trash_path = job.path.replace(self.path, self.trash_path)
value = misc.binary_encode(jsonutils.dumps(data))
txn = self._client.transaction()
txn.create(trash_path, value=value)
txn.delete(job.lock_path, version=lock_stat.version)
txn.delete(job.path, version=data_stat.version)
kazoo_utils.checked_commit(txn)
def _state_change_listener(self, state):
LOG.debug("Kazoo client has changed to state: %s", state)
def wait(self, timeout=None):
# Wait until timeout expires (or forever) for jobs to appear.
watch = tt.StopWatch(duration=timeout)
watch.start()
with self._job_cond:
while True:
if not self._known_jobs:
if watch.expired():
raise excp.NotFound("Expired waiting for jobs to"
" arrive; waited %s seconds"
% watch.elapsed())
# This is done since the given timeout can not be provided
# to the condition variable, since we can not ensure that
# when we acquire the condition that there will actually
# be jobs (especially if we are spuriously awaken), so we
# must recalculate the amount of time we really have left.
self._job_cond.wait(watch.leftover(return_none=True))
else:
it = ZookeeperJobBoardIterator(self)
it._jobs.extend(self._fetch_jobs())
it._fetched = True
return it
@property
def connected(self):
return self._connected and self._client.connected
@lock_utils.locked(lock='_open_close_lock')
def close(self):
if self._owned:
LOG.debug("Stopping client")
kazoo_utils.finalize_client(self._client)
if self._worker is not None:
LOG.debug("Shutting down the notifier")
self._worker.shutdown()
self._worker = None
with self._job_cond:
self._known_jobs.clear()
LOG.debug("Stopped & cleared local state")
self._connected = False
@lock_utils.locked(lock='_open_close_lock')
def connect(self, timeout=10.0):
def try_clean():
# Attempt to do the needed cleanup if post-connection setup does
# not succeed (maybe the connection is lost right after it is
# obtained).
try:
self.close()
except k_exceptions.KazooException:
LOG.exception("Failed cleaning-up after post-connection"
" initialization failed")
try:
if timeout is not None:
timeout = float(timeout)
self._client.start(timeout=timeout)
except (self._client.handler.timeout_exception,
k_exceptions.KazooException):
excp.raise_with_cause(excp.JobFailure,
"Failed to connect to zookeeper")
try:
if self._conf.get('check_compatible', True):
kazoo_utils.check_compatible(self._client, self.MIN_ZK_VERSION)
if self._worker is None and self._emit_notifications:
self._worker = futures.ThreadPoolExecutor(max_workers=1)
self._client.ensure_path(self.path)
self._client.ensure_path(self.trash_path)
if self._job_watcher is None:
self._job_watcher = watchers.ChildrenWatch(
self._client,
self.path,
func=self._on_job_posting,
allow_session_lost=True)
self._connected = True
except excp.IncompatibleVersion:
with excutils.save_and_reraise_exception():
try_clean()
except (self._client.handler.timeout_exception,
k_exceptions.KazooException):
exc_type, exc, exc_tb = sys.exc_info()
try:
try_clean()
excp.raise_with_cause(excp.JobFailure,
"Failed to do post-connection"
" initialization", cause=exc)
finally:
del(exc_type, exc, exc_tb)
|