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
|
/**
* Copyright (C) 2018-present MongoDB, Inc.
*
* This program is free software: you can redistribute it and/or modify
* it under the terms of the Server Side Public License, version 1,
* as published by MongoDB, Inc.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* Server Side Public License for more details.
*
* You should have received a copy of the Server Side Public License
* along with this program. If not, see
* <http://www.mongodb.com/licensing/server-side-public-license>.
*
* As a special exception, the copyright holders give permission to link the
* code of portions of this program with the OpenSSL library under certain
* conditions as described in each individual source file and distribute
* linked combinations including the program with the OpenSSL library. You
* must comply with the Server Side Public License in all respects for
* all of the code used other than as permitted herein. If you modify file(s)
* with this exception, you may extend this exception to your version of the
* file(s), but you are not obligated to do so. If you do not wish to do so,
* delete this exception statement from your version. If you delete this
* exception statement from all source files in the program, then also delete
* it in the license file.
*/
#define MONGO_LOGV2_DEFAULT_COMPONENT ::mongo::logv2::LogComponent::kTransaction
#include "mongo/platform/basic.h"
#include "mongo/db/s/transaction_coordinator_util.h"
#include "mongo/client/remote_command_retry_scheduler.h"
#include "mongo/db/commands/txn_cmds_gen.h"
#include "mongo/db/commands/txn_two_phase_commit_cmds_gen.h"
#include "mongo/db/curop.h"
#include "mongo/db/dbdirectclient.h"
#include "mongo/db/ops/write_ops.h"
#include "mongo/db/repl/repl_client_info.h"
#include "mongo/db/s/transaction_coordinator_futures_util.h"
#include "mongo/db/s/transaction_coordinator_worker_curop_repository.h"
#include "mongo/db/storage/flow_control.h"
#include "mongo/db/write_concern.h"
#include "mongo/logv2/log.h"
#include "mongo/rpc/get_status_from_command_result.h"
#include "mongo/util/fail_point.h"
namespace mongo {
namespace txn {
namespace {
MONGO_FAIL_POINT_DEFINE(hangBeforeWritingParticipantList);
MONGO_FAIL_POINT_DEFINE(hangBeforeSendingPrepare);
MONGO_FAIL_POINT_DEFINE(hangBeforeWritingDecision);
MONGO_FAIL_POINT_DEFINE(hangBeforeSendingCommit);
MONGO_FAIL_POINT_DEFINE(hangBeforeSendingAbort);
MONGO_FAIL_POINT_DEFINE(hangBeforeDeletingCoordinatorDoc);
MONGO_FAIL_POINT_DEFINE(hangAfterDeletingCoordinatorDoc);
using ResponseStatus = executor::TaskExecutor::ResponseStatus;
using CoordinatorAction = TransactionCoordinatorWorkerCurOpRepository::CoordinatorAction;
const Backoff kExponentialBackoff(Seconds(1), Milliseconds::max());
const ReadPreferenceSetting kPrimaryReadPreference{ReadPreference::PrimaryOnly};
BSONArray buildParticipantListMatchesConditions(const std::vector<ShardId>& participantList) {
BSONArrayBuilder barr;
for (const auto& participant : participantList) {
barr.append(participant.toString());
}
const long long participantListLength = participantList.size();
BSONObj participantListHasSize = BSON(TransactionCoordinatorDocument::kParticipantsFieldName
<< BSON("$size" << participantListLength));
BSONObj participantListContains =
BSON(TransactionCoordinatorDocument::kParticipantsFieldName << BSON("$all" << barr.arr()));
return BSON_ARRAY(participantListContains << participantListHasSize);
}
std::string buildParticipantListString(const std::vector<ShardId>& participantList) {
StringBuilder ss;
ss << "[";
for (const auto& participant : participantList) {
ss << participant << " ";
}
ss << "]";
return ss.str();
}
template <typename T>
bool shouldRetryPersistingCoordinatorState(const StatusWith<T>& responseStatus) {
return !responseStatus.isOK() &&
responseStatus != ErrorCodes::TransactionCoordinatorSteppingDown;
}
} // namespace
namespace {
repl::OpTime persistParticipantListBlocking(OperationContext* opCtx,
const LogicalSessionId& lsid,
TxnNumber txnNumber,
const std::vector<ShardId>& participantList) {
LOGV2_DEBUG(22463,
3,
"{sessionId}:{txnNumber} Going to write participant list",
"Going to write participant list",
"sessionId"_attr = lsid.getId(),
"txnNumber"_attr = txnNumber);
if (MONGO_unlikely(hangBeforeWritingParticipantList.shouldFail())) {
LOGV2(22464, "Hit hangBeforeWritingParticipantList failpoint");
hangBeforeWritingParticipantList.pauseWhileSet(opCtx);
}
OperationSessionInfo sessionInfo;
sessionInfo.setSessionId(lsid);
sessionInfo.setTxnNumber(txnNumber);
DBDirectClient client(opCtx);
// Throws if serializing the request or deserializing the response fails.
const auto commandResponse = client.runCommand([&] {
write_ops::Update updateOp(NamespaceString::kTransactionCoordinatorsNamespace);
updateOp.setUpdates({[&] {
write_ops::UpdateOpEntry entry;
// Ensure that the document for the (lsid, txnNumber) either has no participant list or
// has the same participant list. The document may have the same participant list if an
// earlier attempt to write the participant list failed waiting for writeConcern.
BSONObj noParticipantList = BSON(TransactionCoordinatorDocument::kParticipantsFieldName
<< BSON("$exists" << false));
BSONObj sameParticipantList =
BSON("$and" << buildParticipantListMatchesConditions(participantList));
entry.setQ(BSON(TransactionCoordinatorDocument::kIdFieldName
<< sessionInfo.toBSON() << "$or"
<< BSON_ARRAY(noParticipantList << sameParticipantList)));
// Update with participant list.
TransactionCoordinatorDocument doc;
doc.setId(std::move(sessionInfo));
doc.setParticipants(std::move(participantList));
entry.setU(write_ops::UpdateModification::parseFromClassicUpdate(doc.toBSON()));
entry.setUpsert(true);
return entry;
}()});
return updateOp.serialize({});
}());
const auto upsertStatus = getStatusFromWriteCommandReply(commandResponse->getCommandReply());
// Convert a DuplicateKey error to an anonymous error.
if (upsertStatus.code() == ErrorCodes::DuplicateKey) {
// Attempt to include the document for this (lsid, txnNumber) in the error message, if one
// exists. Note that this is best-effort: the document may have been deleted or manually
// changed since the update above ran.
const auto doc = client.findOne(
NamespaceString::kTransactionCoordinatorsNamespace.toString(),
QUERY(TransactionCoordinatorDocument::kIdFieldName << sessionInfo.toBSON()));
uasserted(51025,
str::stream() << "While attempting to write participant list "
<< buildParticipantListString(participantList) << " for "
<< lsid.getId() << ':' << txnNumber
<< ", found document with a different participant list: " << doc);
}
// Throw any other error.
uassertStatusOK(upsertStatus);
LOGV2_DEBUG(22465,
3,
"{sessionId}:{txnNumber} Wrote participant list",
"sessionId"_attr = lsid.getId(),
"txnNumber"_attr = txnNumber);
return repl::ReplClientInfo::forClient(opCtx->getClient()).getLastOp();
}
} // namespace
Future<repl::OpTime> persistParticipantsList(txn::AsyncWorkScheduler& scheduler,
const LogicalSessionId& lsid,
TxnNumber txnNumber,
const txn::ParticipantsList& participants) {
return txn::doWhile(
scheduler,
boost::none /* no need for a backoff */,
[](const StatusWith<repl::OpTime>& s) { return shouldRetryPersistingCoordinatorState(s); },
[&scheduler, lsid, txnNumber, participants] {
return scheduler.scheduleWork([lsid, txnNumber, participants](OperationContext* opCtx) {
FlowControl::Bypass flowControlBypass(opCtx);
getTransactionCoordinatorWorkerCurOpRepository()->set(
opCtx, lsid, txnNumber, CoordinatorAction::kWritingParticipantList);
return persistParticipantListBlocking(opCtx, lsid, txnNumber, participants);
});
});
}
void PrepareVoteConsensus::registerVote(const PrepareResponse& vote) {
if (vote.vote == PrepareVote::kCommit) {
++_numCommitVotes;
_maxPrepareTimestamp = std::max(_maxPrepareTimestamp, *vote.prepareTimestamp);
} else {
vote.vote == PrepareVote::kAbort ? ++_numAbortVotes : ++_numNoVotes;
if (!_abortStatus)
_abortStatus.emplace(*vote.abortReason);
}
}
CoordinatorCommitDecision PrepareVoteConsensus::decision() const {
invariant(_numShards == _numCommitVotes + _numAbortVotes + _numNoVotes);
CoordinatorCommitDecision decision;
if (_numCommitVotes == _numShards) {
decision.setDecision(CommitDecision::kCommit);
decision.setCommitTimestamp(_maxPrepareTimestamp);
} else {
invariant(_abortStatus);
decision.setDecision(CommitDecision::kAbort);
decision.setAbortStatus(*_abortStatus);
}
return decision;
}
Future<PrepareVoteConsensus> sendPrepare(ServiceContext* service,
txn::AsyncWorkScheduler& scheduler,
const LogicalSessionId& lsid,
TxnNumber txnNumber,
const txn::ParticipantsList& participants) {
PrepareTransaction prepareTransaction;
prepareTransaction.setDbName(NamespaceString::kAdminDb);
auto prepareObj = prepareTransaction.toBSON(
BSON("lsid" << lsid.toBSON() << "txnNumber" << txnNumber << "autocommit" << false
<< WriteConcernOptions::kWriteConcernField << WriteConcernOptions::Majority));
std::vector<Future<PrepareResponse>> responses;
// Send prepare to all participants asynchronously and collect their future responses in a
// vector of responses.
auto prepareScheduler = scheduler.makeChildScheduler();
OperationContextFn operationContextFn = [lsid, txnNumber](OperationContext* opCtx) {
invariant(opCtx);
getTransactionCoordinatorWorkerCurOpRepository()->set(
opCtx, lsid, txnNumber, CoordinatorAction::kSendingPrepare);
if (MONGO_unlikely(hangBeforeSendingPrepare.shouldFail())) {
LOGV2(22466, "Hit hangBeforeSendingPrepare failpoint");
hangBeforeSendingPrepare.pauseWhileSet(opCtx);
}
};
for (const auto& participant : participants) {
responses.emplace_back(sendPrepareToShard(service,
*prepareScheduler,
lsid,
txnNumber,
participant,
prepareObj,
operationContextFn));
}
// Asynchronously aggregate all prepare responses to find the decision and max prepare timestamp
// (used for commit), stopping the aggregation and preventing any further retries as soon as an
// abort decision is received. Return a future containing the result.
return txn::collect(
std::move(responses),
// Initial value
PrepareVoteConsensus{int(participants.size())},
// Aggregates an incoming response (next) with the existing aggregate value (result)
[&prepareScheduler = *prepareScheduler, txnNumber](PrepareVoteConsensus& result,
const PrepareResponse& next) {
result.registerVote(next);
if (next.vote == PrepareVote::kAbort) {
LOGV2_DEBUG(5141701,
1,
"Received abort prepare vote from node",
"shardId"_attr = next.shardId,
"txnNumber"_attr = txnNumber,
"error"_attr = (next.abortReason.has_value()
? next.abortReason.value().reason()
: ""));
prepareScheduler.shutdown(
{ErrorCodes::TransactionCoordinatorReachedAbortDecision,
str::stream() << "Received abort vote from " << next.shardId});
}
return txn::ShouldStopIteration::kNo;
})
.tapAll([prepareScheduler = std::move(prepareScheduler)](auto&& unused) mutable {
// Destroy the prepare scheduler before the rest of the future chain has executed so
// that any parent schedulers can be destroyed without dangling children
prepareScheduler.reset();
});
}
namespace {
repl::OpTime persistDecisionBlocking(OperationContext* opCtx,
const LogicalSessionId& lsid,
TxnNumber txnNumber,
const std::vector<ShardId>& participantList,
const txn::CoordinatorCommitDecision& decision) {
const bool isCommit = decision.getDecision() == txn::CommitDecision::kCommit;
LOGV2_DEBUG(22467,
3,
"{sessionId}:{txnNumber} Going to write decision {decision}",
"sessionId"_attr = lsid.getId(),
"txnNumber"_attr = txnNumber,
"decision"_attr = (isCommit ? "commit" : "abort"));
if (MONGO_unlikely(hangBeforeWritingDecision.shouldFail())) {
LOGV2(22468, "Hit hangBeforeWritingDecision failpoint");
hangBeforeWritingDecision.pauseWhileSet(opCtx);
}
OperationSessionInfo sessionInfo;
sessionInfo.setSessionId(lsid);
sessionInfo.setTxnNumber(txnNumber);
DBDirectClient client(opCtx);
// Throws if serializing the request or deserializing the response fails.
const auto commandResponse = client.runCommand([&] {
write_ops::Update updateOp(NamespaceString::kTransactionCoordinatorsNamespace);
updateOp.setUpdates({[&] {
write_ops::UpdateOpEntry entry;
// Ensure that the document for the (lsid, txnNumber) has the same participant list and
// either has no decision or the same decision. The document may have the same decision
// if an earlier attempt to write the decision failed waiting for writeConcern.
BSONObj noDecision = BSON(TransactionCoordinatorDocument::kDecisionFieldName
<< BSON("$exists" << false));
BSONObj sameDecision =
BSON(TransactionCoordinatorDocument::kDecisionFieldName << decision.toBSON());
entry.setQ(BSON(TransactionCoordinatorDocument::kIdFieldName
<< sessionInfo.toBSON() << "$and"
<< buildParticipantListMatchesConditions(participantList) << "$or"
<< BSON_ARRAY(noDecision << sameDecision)));
entry.setUpsert(true);
entry.setU(write_ops::UpdateModification::parseFromClassicUpdate([&] {
TransactionCoordinatorDocument doc;
doc.setId(sessionInfo);
doc.setParticipants(std::move(participantList));
doc.setDecision(decision);
return doc.toBSON();
}()));
return entry;
}()});
return updateOp.serialize({});
}());
const auto commandReply = commandResponse->getCommandReply();
uassertStatusOK(getStatusFromWriteCommandReply(commandReply));
// If no document matched, throw an anonymous error. (The update itself will not have thrown an
// error, because it's legal for an update to match no documents.)
if (commandReply.getIntField("n") != 1) {
// Attempt to include the document for this (lsid, txnNumber) in the error message, if one
// exists. Note that this is best-effort: the document may have been deleted or manually
// changed since the update above ran.
const auto doc = client.findOne(
NamespaceString::kTransactionCoordinatorsNamespace.ns(),
QUERY(TransactionCoordinatorDocument::kIdFieldName << sessionInfo.toBSON()));
uasserted(51026,
str::stream() << "While attempting to write decision "
<< (isCommit ? "'commit'" : "'abort'") << " for" << lsid.getId()
<< ':' << txnNumber
<< ", either failed to find document for this lsid:txnNumber or "
"document existed with a different participant list, decision "
"or commitTimestamp: "
<< doc);
}
LOGV2_DEBUG(22469,
3,
"{sessionId}:{txnNumber} Wrote decision {decision}",
"Wrote decision",
"sessionId"_attr = lsid.getId(),
"txnNumber"_attr = txnNumber,
"decision"_attr = (isCommit ? "commit" : "abort"));
return repl::ReplClientInfo::forClient(opCtx->getClient()).getLastOp();
}
} // namespace
Future<repl::OpTime> persistDecision(txn::AsyncWorkScheduler& scheduler,
const LogicalSessionId& lsid,
TxnNumber txnNumber,
const txn::ParticipantsList& participants,
const txn::CoordinatorCommitDecision& decision) {
return txn::doWhile(
scheduler,
boost::none /* no need for a backoff */,
[](const StatusWith<repl::OpTime>& s) { return shouldRetryPersistingCoordinatorState(s); },
[&scheduler, lsid, txnNumber, participants, decision] {
return scheduler.scheduleWork(
[lsid, txnNumber, participants, decision](OperationContext* opCtx) {
FlowControl::Bypass flowControlBypass(opCtx);
getTransactionCoordinatorWorkerCurOpRepository()->set(
opCtx, lsid, txnNumber, CoordinatorAction::kWritingDecision);
return persistDecisionBlocking(opCtx, lsid, txnNumber, participants, decision);
});
});
}
Future<void> sendCommit(ServiceContext* service,
txn::AsyncWorkScheduler& scheduler,
const LogicalSessionId& lsid,
TxnNumber txnNumber,
const txn::ParticipantsList& participants,
Timestamp commitTimestamp) {
CommitTransaction commitTransaction;
commitTransaction.setDbName(NamespaceString::kAdminDb);
commitTransaction.setCommitTimestamp(commitTimestamp);
auto commitObj = commitTransaction.toBSON(
BSON("lsid" << lsid.toBSON() << "txnNumber" << txnNumber << "autocommit" << false
<< WriteConcernOptions::kWriteConcernField << WriteConcernOptions::Majority));
OperationContextFn operationContextFn = [lsid, txnNumber](OperationContext* opCtx) {
invariant(opCtx);
getTransactionCoordinatorWorkerCurOpRepository()->set(
opCtx, lsid, txnNumber, CoordinatorAction::kSendingCommit);
if (MONGO_unlikely(hangBeforeSendingCommit.shouldFail())) {
LOGV2(22470, "Hit hangBeforeSendingCommit failpoint");
hangBeforeSendingCommit.pauseWhileSet(opCtx);
}
};
std::vector<Future<void>> responses;
for (const auto& participant : participants) {
responses.push_back(sendDecisionToShard(
service, scheduler, lsid, txnNumber, participant, commitObj, operationContextFn));
}
return txn::whenAll(responses);
}
Future<void> sendAbort(ServiceContext* service,
txn::AsyncWorkScheduler& scheduler,
const LogicalSessionId& lsid,
TxnNumber txnNumber,
const txn::ParticipantsList& participants) {
AbortTransaction abortTransaction;
abortTransaction.setDbName(NamespaceString::kAdminDb);
auto abortObj = abortTransaction.toBSON(
BSON("lsid" << lsid.toBSON() << "txnNumber" << txnNumber << "autocommit" << false
<< WriteConcernOptions::kWriteConcernField << WriteConcernOptions::Majority));
OperationContextFn operationContextFn = [lsid, txnNumber](OperationContext* opCtx) {
invariant(opCtx);
getTransactionCoordinatorWorkerCurOpRepository()->set(
opCtx, lsid, txnNumber, CoordinatorAction::kSendingAbort);
if (MONGO_unlikely(hangBeforeSendingAbort.shouldFail())) {
LOGV2(22471, "Hit hangBeforeSendingAbort failpoint");
hangBeforeSendingAbort.pauseWhileSet(opCtx);
}
};
std::vector<Future<void>> responses;
for (const auto& participant : participants) {
responses.push_back(sendDecisionToShard(
service, scheduler, lsid, txnNumber, participant, abortObj, operationContextFn));
}
return txn::whenAll(responses);
}
namespace {
void deleteCoordinatorDocBlocking(OperationContext* opCtx,
const LogicalSessionId& lsid,
TxnNumber txnNumber) {
LOGV2_DEBUG(22472,
3,
"{sessionId}:{txnNumber} Going to delete coordinator doc",
"Going to delete coordinator doc",
"sessionId"_attr = lsid.getId(),
"txnNumber"_attr = txnNumber);
if (MONGO_unlikely(hangBeforeDeletingCoordinatorDoc.shouldFail())) {
LOGV2(22473, "Hit hangBeforeDeletingCoordinatorDoc failpoint");
hangBeforeDeletingCoordinatorDoc.pauseWhileSet(opCtx);
}
OperationSessionInfo sessionInfo;
sessionInfo.setSessionId(lsid);
sessionInfo.setTxnNumber(txnNumber);
DBDirectClient client(opCtx);
// Throws if serializing the request or deserializing the response fails.
auto commandResponse = client.runCommand([&] {
write_ops::Delete deleteOp(NamespaceString::kTransactionCoordinatorsNamespace);
deleteOp.setDeletes({[&] {
write_ops::DeleteOpEntry entry;
// Ensure the document is only deleted after a decision has been made.
BSONObj abortDecision =
BSON(TransactionCoordinatorDocument::kDecisionFieldName << "abort");
BSONObj commitDecision =
BSON(TransactionCoordinatorDocument::kDecisionFieldName << "commit");
entry.setQ(BSON(TransactionCoordinatorDocument::kIdFieldName
<< sessionInfo.toBSON()
<< TransactionCoordinatorDocument::kDecisionFieldName
<< BSON("$exists" << true)));
entry.setMulti(false);
return entry;
}()});
return deleteOp.serialize({});
}());
const auto commandReply = commandResponse->getCommandReply();
uassertStatusOK(getStatusFromWriteCommandReply(commandReply));
// If no document matched, throw an anonymous error. (The delete itself will not have thrown an
// error, because it's legal for a delete to match no documents.)
if (commandReply.getIntField("n") != 1) {
// Attempt to include the document for this (lsid, txnNumber) in the error message, if one
// exists. Note that this is best-effort: the document may have been deleted or manually
// changed since the update above ran.
const auto doc = client.findOne(
NamespaceString::kTransactionCoordinatorsNamespace.toString(),
QUERY(TransactionCoordinatorDocument::kIdFieldName << sessionInfo.toBSON()));
uasserted(51027,
str::stream() << "While attempting to delete document for " << lsid.getId() << ':'
<< txnNumber
<< ", either failed to find document for this lsid:txnNumber or "
"document existed without a decision: "
<< doc);
}
LOGV2_DEBUG(22474,
3,
"{sessionId}:{txnNumber} Deleted coordinator doc",
"Deleted coordinator doc",
"sessionId"_attr = lsid.getId(),
"txnNumber"_attr = txnNumber);
hangAfterDeletingCoordinatorDoc.execute([&](const BSONObj& data) {
LOGV2(22475, "Hit hangAfterDeletingCoordinatorDoc failpoint");
if (!data["useUninterruptibleSleep"].eoo()) {
hangAfterDeletingCoordinatorDoc.pauseWhileSet();
} else {
hangAfterDeletingCoordinatorDoc.pauseWhileSet(opCtx);
}
});
}
} // namespace
Future<void> deleteCoordinatorDoc(txn::AsyncWorkScheduler& scheduler,
const LogicalSessionId& lsid,
TxnNumber txnNumber) {
return txn::doWhile(
scheduler,
boost::none /* no need for a backoff */,
[](const Status& s) { return s == ErrorCodes::Interrupted; },
[&scheduler, lsid, txnNumber] {
return scheduler.scheduleWork([lsid, txnNumber](OperationContext* opCtx) {
FlowControl::Bypass flowControlBypass(opCtx);
getTransactionCoordinatorWorkerCurOpRepository()->set(
opCtx, lsid, txnNumber, CoordinatorAction::kDeletingCoordinatorDoc);
deleteCoordinatorDocBlocking(opCtx, lsid, txnNumber);
});
});
}
std::vector<TransactionCoordinatorDocument> readAllCoordinatorDocs(OperationContext* opCtx) {
std::vector<TransactionCoordinatorDocument> allCoordinatorDocs;
DBDirectClient client(opCtx);
auto coordinatorDocsCursor =
client.query(NamespaceString::kTransactionCoordinatorsNamespace, Query{});
while (coordinatorDocsCursor->more()) {
// TODO (SERVER-38307): Try/catch around parsing the document and skip the document if it
// fails to parse.
auto nextDecision = TransactionCoordinatorDocument::parse(IDLParserErrorContext(""),
coordinatorDocsCursor->next());
allCoordinatorDocs.push_back(nextDecision);
}
return allCoordinatorDocs;
}
Future<PrepareResponse> sendPrepareToShard(ServiceContext* service,
txn::AsyncWorkScheduler& scheduler,
const LogicalSessionId& lsid,
TxnNumber txnNumber,
const ShardId& shardId,
const BSONObj& commandObj,
OperationContextFn operationContextFn) {
const bool isLocalShard = (shardId == txn::getLocalShardId(service));
auto f = txn::doWhile(
scheduler,
kExponentialBackoff,
[](const StatusWith<PrepareResponse>& swPrepareResponse) {
// *Always* retry until hearing a conclusive response or being told to stop via a
// coordinator-specific code.
return !swPrepareResponse.isOK() &&
swPrepareResponse != ErrorCodes::TransactionCoordinatorSteppingDown &&
swPrepareResponse != ErrorCodes::TransactionCoordinatorReachedAbortDecision;
},
[&scheduler,
lsid,
txnNumber,
shardId,
isLocalShard,
commandObj = commandObj.getOwned(),
operationContextFn] {
LOGV2_DEBUG(22476,
3,
"{sessionId}:{txnNumber} Coordinator going to send command "
"{command} to {localOrRemote} shard {shardId}",
"Coordinator going to send command to shard",
"sessionId"_attr = lsid.getId(),
"txnNumber"_attr = txnNumber,
"command"_attr = commandObj,
"localOrRemote"_attr = (isLocalShard ? "local" : "remote"),
"shardId"_attr = shardId);
return scheduler
.scheduleRemoteCommand(
shardId, kPrimaryReadPreference, commandObj, operationContextFn)
.then([lsid, txnNumber, shardId, commandObj = commandObj.getOwned()](
ResponseStatus response) {
auto status = getStatusFromCommandResult(response.data);
auto wcStatus = getWriteConcernStatusFromCommandResult(response.data);
// There must be no writeConcern error in order for us to interpret the command
// response.
if (!wcStatus.isOK()) {
status = wcStatus;
}
if (status.isOK()) {
auto prepareTimestampField = response.data["prepareTimestamp"];
if (prepareTimestampField.eoo() ||
prepareTimestampField.timestamp().isNull()) {
Status abortStatus(ErrorCodes::Error(50993),
str::stream()
<< "Coordinator shard received an OK response "
"to prepareTransaction without a "
"prepareTimestamp from shard "
<< shardId
<< ", which is not an expected behavior. "
"Interpreting the response as vote to abort");
LOGV2(22477,
"{sessionId}:{txnNumber} {error}",
"Coordinator received error from transaction participant",
"sessionId"_attr = lsid.getId(),
"txnNumber"_attr = txnNumber,
"error"_attr = redact(abortStatus));
return PrepareResponse{
shardId, PrepareVote::kAbort, boost::none, abortStatus};
}
LOGV2_DEBUG(
22478,
3,
"{sessionId}:{txnNumber} Coordinator shard received a "
"vote to commit from shard {shardId} with prepareTimestamp: "
"{prepareTimestamp}",
"Coordinator shard received a vote to commit from participant shard",
"sessionId"_attr = lsid.getId(),
"txnNumber"_attr = txnNumber,
"shardId"_attr = shardId,
"prepareTimestampField"_attr = prepareTimestampField.timestamp());
return PrepareResponse{shardId,
PrepareVote::kCommit,
prepareTimestampField.timestamp(),
boost::none};
}
LOGV2_DEBUG(22479,
3,
"{sessionId}:{txnNumber} Coordinator shard received "
"{error} from shard {shardId} for {command}",
"Coordinator shard received response from shard",
"sessionId"_attr = lsid.getId(),
"txnNumber"_attr = txnNumber,
"error"_attr = status,
"shardId"_attr = shardId,
"command"_attr = commandObj);
if (ErrorCodes::isVoteAbortError(status.code())) {
return PrepareResponse{
shardId,
PrepareVote::kAbort,
boost::none,
status.withContext(str::stream() << "from shard " << shardId)};
}
uassertStatusOK(status);
MONGO_UNREACHABLE;
})
.onError<ErrorCodes::ShardNotFound>([shardId, isLocalShard](const Status& status) {
invariant(!isLocalShard);
// ShardNotFound may indicate that the participant shard has been removed (it
// could also mean the participant shard was recently added and this node
// refreshed its ShardRegistry from a stale config secondary).
//
// Since this node can't know which is the case, it is safe to pessimistically
// treat ShardNotFound as a vote to abort, which is always safe since the node
// must then send abort.
return Future<PrepareResponse>::makeReady(
{shardId, CommitDecision::kAbort, boost::none, status});
});
});
return std::move(f).onError<ErrorCodes::TransactionCoordinatorReachedAbortDecision>(
[lsid, txnNumber, shardId](const Status& status) {
LOGV2_DEBUG(22480,
3,
"{sessionId}:{txnNumber} Prepare stopped retrying due to retrying "
"being cancelled",
"Prepare stopped retrying due to retrying being cancelled",
"sessionId"_attr = lsid.getId(),
"txnNumber"_attr = txnNumber);
return PrepareResponse{shardId,
boost::none,
boost::none,
Status(ErrorCodes::NoSuchTransaction, status.reason())};
});
}
Future<void> sendDecisionToShard(ServiceContext* service,
txn::AsyncWorkScheduler& scheduler,
const LogicalSessionId& lsid,
TxnNumber txnNumber,
const ShardId& shardId,
const BSONObj& commandObj,
OperationContextFn operationContextFn) {
const bool isLocalShard = (shardId == txn::getLocalShardId(service));
return txn::doWhile(
scheduler,
kExponentialBackoff,
[](const Status& s) {
// *Always* retry until hearing a conclusive response or being told to stop via a
// coordinator-specific code.
return !s.isOK() && s != ErrorCodes::TransactionCoordinatorSteppingDown;
},
[&scheduler,
lsid,
txnNumber,
shardId,
isLocalShard,
operationContextFn,
commandObj = commandObj.getOwned()] {
LOGV2_DEBUG(22481,
3,
"{sessionId}:{txnNumber} Coordinator going to send command "
"{command} to {localOrRemote} shard {shardId}",
"Coordinator going to send command to shard",
"sessionId"_attr = lsid.getId(),
"txnNumber"_attr = txnNumber,
"command"_attr = commandObj,
"localOrRemote"_attr = (isLocalShard ? "local" : "remote"),
"shardId"_attr = shardId);
return scheduler
.scheduleRemoteCommand(
shardId, kPrimaryReadPreference, commandObj, operationContextFn)
.then([lsid, txnNumber, shardId, commandObj = commandObj.getOwned()](
ResponseStatus response) {
auto status = getStatusFromCommandResult(response.data);
auto wcStatus = getWriteConcernStatusFromCommandResult(response.data);
// There must be no writeConcern error in order for us to interpret the command
// response.
if (!wcStatus.isOK()) {
status = wcStatus;
}
LOGV2_DEBUG(22482,
3,
"{sessionId}:{txnNumber} Coordinator shard received "
"{status} in response to {command} from shard {shardId}",
"Coordinator shard received response from shard",
"sessionId"_attr = lsid.getId(),
"txnNumber"_attr = txnNumber,
"status"_attr = status,
"command"_attr = commandObj,
"shardId"_attr = shardId);
if (ErrorCodes::isVoteAbortError(status.code())) {
// Interpret voteAbort errors as an ack.
status = Status::OK();
}
return status;
})
.onError<ErrorCodes::ShardNotFound>([isLocalShard](const Status& s) {
invariant(!isLocalShard);
// TODO (SERVER-38918): Unlike for prepare, there is no pessimistic way to
// handle ShardNotFound. It's not safe to treat ShardNotFound as an ack, because
// this node may have refreshed its ShardRegistry from a stale config secondary.
fassert(51068, false);
});
});
}
std::string txnIdToString(const LogicalSessionId& lsid, TxnNumber txnNumber) {
return str::stream() << lsid.getId() << ':' << txnNumber;
}
} // namespace txn
} // namespace mongo
|