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
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
|
/**
* 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::kReplication
#include "mongo/platform/basic.h"
#include "mongo/db/op_observer_impl.h"
#include <limits>
#include "mongo/bson/bsonobjbuilder.h"
#include "mongo/db/bson/dotted_path_support.h"
#include "mongo/db/catalog/collection_options.h"
#include "mongo/db/catalog/database.h"
#include "mongo/db/catalog/database_holder.h"
#include "mongo/db/commands/feature_compatibility_version.h"
#include "mongo/db/commands/feature_compatibility_version_parser.h"
#include "mongo/db/commands/txn_cmds_gen.h"
#include "mongo/db/concurrency/d_concurrency.h"
#include "mongo/db/concurrency/write_conflict_exception.h"
#include "mongo/db/index/index_descriptor.h"
#include "mongo/db/logical_time_validator.h"
#include "mongo/db/namespace_string.h"
#include "mongo/db/op_observer_util.h"
#include "mongo/db/operation_context.h"
#include "mongo/db/read_write_concern_defaults.h"
#include "mongo/db/repl/oplog.h"
#include "mongo/db/repl/oplog_entry_gen.h"
#include "mongo/db/repl/replication_coordinator.h"
#include "mongo/db/s/collection_sharding_state.h"
#include "mongo/db/server_options.h"
#include "mongo/db/session_catalog_mongod.h"
#include "mongo/db/storage/durable_catalog.h"
#include "mongo/db/transaction_participant.h"
#include "mongo/db/transaction_participant_gen.h"
#include "mongo/db/views/durable_view_catalog.h"
#include "mongo/logv2/log.h"
#include "mongo/s/client/shard_registry.h"
#include "mongo/s/grid.h"
#include "mongo/scripting/engine.h"
#include "mongo/util/assert_util.h"
#include "mongo/util/fail_point.h"
namespace mongo {
using repl::MutableOplogEntry;
using repl::OplogEntry;
namespace {
MONGO_FAIL_POINT_DEFINE(failCollectionUpdates);
MONGO_FAIL_POINT_DEFINE(hangAndFailUnpreparedCommitAfterReservingOplogSlot);
const auto documentKeyDecoration =
OperationContext::declareDecoration<boost::optional<OpObserverImpl::DocumentKey>>();
constexpr auto kNumRecordsFieldName = "numRecords"_sd;
constexpr auto kMsgFieldName = "msg"_sd;
constexpr long long kInvalidNumRecords = -1LL;
Date_t getWallClockTimeForOpLog(OperationContext* opCtx) {
auto const clockSource = opCtx->getServiceContext()->getFastClockSource();
return clockSource->now();
}
repl::OpTime logOperation(OperationContext* opCtx, MutableOplogEntry* oplogEntry) {
oplogEntry->setWallClockTime(getWallClockTimeForOpLog(opCtx));
auto& times = OpObserver::Times::get(opCtx).reservedOpTimes;
auto opTime = repl::logOp(opCtx, oplogEntry);
times.push_back(opTime);
return opTime;
}
/**
* Updates the session state with the last write timestamp and transaction for that session.
*
* In the case of writes with transaction/statement id, this method will be recursively entered a
* second time for the actual write to the transactions table. Since this write does not generate an
* oplog entry, the recursion will stop at this point.
*/
void onWriteOpCompleted(OperationContext* opCtx,
std::vector<StmtId> stmtIdsWritten,
SessionTxnRecord sessionTxnRecord) {
if (sessionTxnRecord.getLastWriteOpTime().isNull())
return;
auto txnParticipant = TransactionParticipant::get(opCtx);
if (!txnParticipant)
return;
// We add these here since they may not exist if we return early.
sessionTxnRecord.setSessionId(*opCtx->getLogicalSessionId());
sessionTxnRecord.setTxnNum(*opCtx->getTxnNumber());
txnParticipant.onWriteOpCompletedOnPrimary(opCtx, std::move(stmtIdsWritten), sessionTxnRecord);
}
/**
* Given the collection count from Collection::numRecords(), create and return the object for the
* 'o2' field of a drop or rename oplog entry. If the collection count exceeds the upper limit of a
* BSON NumberLong (long long), we will add a count of -1 and append a message with the original
* collection count.
*
* Replication rollback uses this field to correct correction counts on drop-pending collections.
*/
BSONObj makeObject2ForDropOrRename(uint64_t numRecords) {
BSONObjBuilder obj2Builder;
if (numRecords > static_cast<uint64_t>(std::numeric_limits<long long>::max())) {
obj2Builder.appendNumber(kNumRecordsFieldName, kInvalidNumRecords);
std::string msg = str::stream() << "Collection count " << numRecords
<< " is larger than the "
"maximum int64_t value. Setting numRecords to -1.";
obj2Builder.append(kMsgFieldName, msg);
} else {
obj2Builder.appendNumber(kNumRecordsFieldName, static_cast<long long>(numRecords));
}
auto obj = obj2Builder.obj();
return obj;
}
struct OpTimeBundle {
repl::OpTime writeOpTime;
repl::OpTime prePostImageOpTime;
Date_t wallClockTime;
};
/**
* Write oplog entry(ies) for the update operation.
*/
OpTimeBundle replLogUpdate(OperationContext* opCtx, const OplogUpdateEntryArgs& args) {
MutableOplogEntry oplogEntry;
oplogEntry.setNss(args.nss);
oplogEntry.setUuid(args.uuid);
repl::OplogLink oplogLink;
repl::appendOplogEntryChainInfo(opCtx, &oplogEntry, &oplogLink, args.updateArgs.stmtId);
OpTimeBundle opTimes;
const auto storePreImageForRetryableWrite =
(args.updateArgs.storeDocOption == CollectionUpdateArgs::StoreDocOption::PreImage &&
opCtx->getTxnNumber());
if (storePreImageForRetryableWrite || args.updateArgs.preImageRecordingEnabledForCollection) {
MutableOplogEntry noopEntry = oplogEntry;
invariant(args.updateArgs.preImageDoc);
noopEntry.setOpType(repl::OpTypeEnum::kNoop);
noopEntry.setObject(*args.updateArgs.preImageDoc);
oplogLink.preImageOpTime = logOperation(opCtx, &noopEntry);
if (storePreImageForRetryableWrite) {
opTimes.prePostImageOpTime = oplogLink.preImageOpTime;
}
}
// This case handles storing the post image for retryable findAndModify's.
if (args.updateArgs.storeDocOption == CollectionUpdateArgs::StoreDocOption::PostImage &&
opCtx->getTxnNumber()) {
MutableOplogEntry noopEntry = oplogEntry;
noopEntry.setOpType(repl::OpTypeEnum::kNoop);
noopEntry.setObject(args.updateArgs.updatedDoc);
oplogLink.postImageOpTime = logOperation(opCtx, &noopEntry);
invariant(opTimes.prePostImageOpTime.isNull());
opTimes.prePostImageOpTime = oplogLink.postImageOpTime;
}
oplogEntry.setOpType(repl::OpTypeEnum::kUpdate);
oplogEntry.setObject(args.updateArgs.update);
oplogEntry.setObject2(args.updateArgs.criteria);
oplogEntry.setFromMigrateIfTrue(args.updateArgs.fromMigrate);
// oplogLink could have been changed to include pre/postImageOpTime by the previous no-op write.
repl::appendOplogEntryChainInfo(opCtx, &oplogEntry, &oplogLink, args.updateArgs.stmtId);
opTimes.writeOpTime = logOperation(opCtx, &oplogEntry);
opTimes.wallClockTime = oplogEntry.getWallClockTime();
return opTimes;
}
/**
* Write oplog entry(ies) for the delete operation.
*/
OpTimeBundle replLogDelete(OperationContext* opCtx,
const NamespaceString& nss,
OptionalCollectionUUID uuid,
StmtId stmtId,
bool fromMigrate,
const boost::optional<BSONObj>& deletedDoc) {
MutableOplogEntry oplogEntry;
oplogEntry.setNss(nss);
oplogEntry.setUuid(uuid);
repl::OplogLink oplogLink;
repl::appendOplogEntryChainInfo(opCtx, &oplogEntry, &oplogLink, stmtId);
OpTimeBundle opTimes;
if (deletedDoc) {
MutableOplogEntry noopEntry = oplogEntry;
noopEntry.setOpType(repl::OpTypeEnum::kNoop);
noopEntry.setObject(*deletedDoc);
auto noteOplog = logOperation(opCtx, &noopEntry);
opTimes.prePostImageOpTime = noteOplog;
oplogLink.preImageOpTime = noteOplog;
}
oplogEntry.setOpType(repl::OpTypeEnum::kDelete);
oplogEntry.setObject(documentKeyDecoration(opCtx).get().getShardKeyAndId());
oplogEntry.setFromMigrateIfTrue(fromMigrate);
// oplogLink could have been changed to include preImageOpTime by the previous no-op write.
repl::appendOplogEntryChainInfo(opCtx, &oplogEntry, &oplogLink, stmtId);
opTimes.writeOpTime = logOperation(opCtx, &oplogEntry);
opTimes.wallClockTime = oplogEntry.getWallClockTime();
return opTimes;
}
} // namespace
BSONObj OpObserverImpl::DocumentKey::getId() const {
return _id;
}
BSONObj OpObserverImpl::DocumentKey::getShardKeyAndId() const {
if (_shardKey) {
BSONObjBuilder builder(_shardKey.get());
builder.appendElementsUnique(_id);
return builder.obj();
}
// _shardKey is not set so just return the _id.
return getId();
}
OpObserverImpl::DocumentKey OpObserverImpl::getDocumentKey(OperationContext* opCtx,
NamespaceString const& nss,
BSONObj const& doc) {
BSONObj id = doc["_id"] ? doc["_id"].wrap() : doc;
boost::optional<BSONObj> shardKey;
// Extract the shard key from the collection description in the CollectionShardingState
// if running on standalone or primary. Skip this completely on secondaries since they are
// not expected to have the collection metadata cached.
if (opCtx->writesAreReplicated()) {
auto collDesc = CollectionShardingState::get(opCtx, nss)->getCollectionDescription();
if (collDesc.isSharded()) {
shardKey =
dotted_path_support::extractElementsBasedOnTemplate(doc, collDesc.getKeyPattern())
.getOwned();
}
}
return {std::move(id), std::move(shardKey)};
}
void OpObserverImpl::onCreateIndex(OperationContext* opCtx,
const NamespaceString& nss,
CollectionUUID uuid,
BSONObj indexDoc,
bool fromMigrate) {
auto txnParticipant = TransactionParticipant::get(opCtx);
const bool inMultiDocumentTransaction =
txnParticipant && opCtx->writesAreReplicated() && txnParticipant.transactionIsOpen();
if (inMultiDocumentTransaction) {
auto operation = MutableOplogEntry::makeCreateIndexesCommand(nss, uuid, indexDoc);
txnParticipant.addTransactionOperation(opCtx, operation);
} else {
BSONObjBuilder builder;
builder.append("createIndexes", nss.coll());
builder.appendElements(indexDoc);
MutableOplogEntry oplogEntry;
oplogEntry.setOpType(repl::OpTypeEnum::kCommand);
oplogEntry.setNss(nss.getCommandNS());
oplogEntry.setUuid(uuid);
oplogEntry.setObject(builder.done());
oplogEntry.setFromMigrateIfTrue(fromMigrate);
logOperation(opCtx, &oplogEntry);
}
}
void OpObserverImpl::onStartIndexBuild(OperationContext* opCtx,
const NamespaceString& nss,
CollectionUUID collUUID,
const UUID& indexBuildUUID,
const std::vector<BSONObj>& indexes,
bool fromMigrate) {
BSONObjBuilder oplogEntryBuilder;
oplogEntryBuilder.append("startIndexBuild", nss.coll());
indexBuildUUID.appendToBuilder(&oplogEntryBuilder, "indexBuildUUID");
BSONArrayBuilder indexesArr(oplogEntryBuilder.subarrayStart("indexes"));
for (auto indexDoc : indexes) {
indexesArr.append(indexDoc);
}
indexesArr.done();
MutableOplogEntry oplogEntry;
oplogEntry.setOpType(repl::OpTypeEnum::kCommand);
oplogEntry.setNss(nss.getCommandNS());
oplogEntry.setUuid(collUUID);
oplogEntry.setObject(oplogEntryBuilder.done());
oplogEntry.setFromMigrateIfTrue(fromMigrate);
logOperation(opCtx, &oplogEntry);
}
void OpObserverImpl::onStartIndexBuildSinglePhase(OperationContext* opCtx,
const NamespaceString& nss) {
// This function sets a timestamp for the initial catalog write when beginning an index
// build, if necessary. There are four scenarios:
// 1. A timestamp is already set -- replication application sets a timestamp ahead of time.
// This could include the phase of initial sync where it applies oplog entries. Also,
// primaries performing an index build via `applyOps` may have a wrapping commit timestamp.
if (!opCtx->recoveryUnit()->getCommitTimestamp().isNull())
return;
// 2. If the node is initial syncing, we do not set a timestamp.
auto replCoord = repl::ReplicationCoordinator::get(opCtx);
if (replCoord->isReplEnabled() && replCoord->getMemberState().startup2())
return;
// 3. If the index build is on the local database, do not timestamp.
if (nss.isLocal())
return;
// 4. All other cases, we generate a timestamp by writing a no-op oplog entry. This is
// better than using a ghost timestamp. Writing an oplog entry ensures this node is
// primary.
onInternalOpMessage(
opCtx,
{},
boost::none,
BSON("msg" << std::string(str::stream() << "Creating indexes. Coll: " << nss)),
boost::none);
}
void OpObserverImpl::onCommitIndexBuild(OperationContext* opCtx,
const NamespaceString& nss,
CollectionUUID collUUID,
const UUID& indexBuildUUID,
const std::vector<BSONObj>& indexes,
bool fromMigrate) {
BSONObjBuilder oplogEntryBuilder;
oplogEntryBuilder.append("commitIndexBuild", nss.coll());
indexBuildUUID.appendToBuilder(&oplogEntryBuilder, "indexBuildUUID");
BSONArrayBuilder indexesArr(oplogEntryBuilder.subarrayStart("indexes"));
for (auto indexDoc : indexes) {
indexesArr.append(indexDoc);
}
indexesArr.done();
MutableOplogEntry oplogEntry;
oplogEntry.setOpType(repl::OpTypeEnum::kCommand);
oplogEntry.setNss(nss.getCommandNS());
oplogEntry.setUuid(collUUID);
oplogEntry.setObject(oplogEntryBuilder.done());
oplogEntry.setFromMigrateIfTrue(fromMigrate);
logOperation(opCtx, &oplogEntry);
}
void OpObserverImpl::onAbortIndexBuild(OperationContext* opCtx,
const NamespaceString& nss,
CollectionUUID collUUID,
const UUID& indexBuildUUID,
const std::vector<BSONObj>& indexes,
const Status& cause,
bool fromMigrate) {
BSONObjBuilder oplogEntryBuilder;
oplogEntryBuilder.append("abortIndexBuild", nss.coll());
indexBuildUUID.appendToBuilder(&oplogEntryBuilder, "indexBuildUUID");
BSONArrayBuilder indexesArr(oplogEntryBuilder.subarrayStart("indexes"));
for (auto indexDoc : indexes) {
indexesArr.append(indexDoc);
}
indexesArr.done();
BSONObjBuilder causeBuilder(oplogEntryBuilder.subobjStart("cause"));
// Some functions that extract a Status from a BSONObj, such as getStatusFromCommandResult(),
// expect the 'ok' field.
causeBuilder.appendBool("ok", 0);
cause.serializeErrorToBSON(&causeBuilder);
causeBuilder.done();
MutableOplogEntry oplogEntry;
oplogEntry.setOpType(repl::OpTypeEnum::kCommand);
oplogEntry.setNss(nss.getCommandNS());
oplogEntry.setUuid(collUUID);
oplogEntry.setObject(oplogEntryBuilder.done());
oplogEntry.setFromMigrateIfTrue(fromMigrate);
logOperation(opCtx, &oplogEntry);
}
void OpObserverImpl::onInserts(OperationContext* opCtx,
const NamespaceString& nss,
OptionalCollectionUUID uuid,
std::vector<InsertStatement>::const_iterator first,
std::vector<InsertStatement>::const_iterator last,
bool fromMigrate) {
auto txnParticipant = TransactionParticipant::get(opCtx);
const bool inMultiDocumentTransaction =
txnParticipant && opCtx->writesAreReplicated() && txnParticipant.transactionIsOpen();
Date_t lastWriteDate;
std::vector<repl::OpTime> opTimeList;
repl::OpTime lastOpTime;
if (inMultiDocumentTransaction) {
// Do not add writes to the profile collection to the list of transaction operations, since
// these are done outside the transaction. There is no top-level WriteUnitOfWork when we are
// in a SideTransactionBlock.
if (!opCtx->getWriteUnitOfWork()) {
invariant(nss.isSystemDotProfile());
return;
}
for (auto iter = first; iter != last; iter++) {
auto operation = MutableOplogEntry::makeInsertOperation(nss, uuid.get(), iter->doc);
txnParticipant.addTransactionOperation(opCtx, operation);
}
} else {
MutableOplogEntry oplogEntryTemplate;
oplogEntryTemplate.setNss(nss);
oplogEntryTemplate.setUuid(uuid);
oplogEntryTemplate.setFromMigrateIfTrue(fromMigrate);
lastWriteDate = getWallClockTimeForOpLog(opCtx);
oplogEntryTemplate.setWallClockTime(lastWriteDate);
opTimeList = repl::logInsertOps(opCtx, &oplogEntryTemplate, first, last);
if (!opTimeList.empty())
lastOpTime = opTimeList.back();
auto& times = OpObserver::Times::get(opCtx).reservedOpTimes;
using std::begin;
using std::end;
times.insert(end(times), begin(opTimeList), end(opTimeList));
std::vector<StmtId> stmtIdsWritten;
std::transform(first,
last,
std::back_inserter(stmtIdsWritten),
[](const InsertStatement& stmt) { return stmt.stmtId; });
SessionTxnRecord sessionTxnRecord;
sessionTxnRecord.setLastWriteOpTime(lastOpTime);
sessionTxnRecord.setLastWriteDate(lastWriteDate);
onWriteOpCompleted(opCtx, stmtIdsWritten, sessionTxnRecord);
}
size_t index = 0;
for (auto it = first; it != last; it++, index++) {
auto opTime = opTimeList.empty() ? repl::OpTime() : opTimeList[index];
shardObserveInsertOp(opCtx, nss, it->doc, opTime, fromMigrate, inMultiDocumentTransaction);
}
if (nss.coll() == "system.js") {
Scope::storedFuncMod(opCtx);
} else if (nss.coll() == DurableViewCatalog::viewsCollectionName()) {
DurableViewCatalog::onExternalChange(opCtx, nss);
} else if (nss == NamespaceString::kServerConfigurationNamespace) {
// We must check server configuration collection writes for featureCompatibilityVersion
// document changes.
for (auto it = first; it != last; it++) {
FeatureCompatibilityVersion::onInsertOrUpdate(opCtx, it->doc);
}
} else if (nss == NamespaceString::kSessionTransactionsTableNamespace && !lastOpTime.isNull()) {
for (auto it = first; it != last; it++) {
MongoDSessionCatalog::observeDirectWriteToConfigTransactions(opCtx, it->doc);
}
} else if (nss == NamespaceString::kConfigSettingsNamespace) {
for (auto it = first; it != last; it++) {
ReadWriteConcernDefaults::get(opCtx).observeDirectWriteToConfigSettings(
opCtx, it->doc["_id"], it->doc);
}
}
}
void OpObserverImpl::onUpdate(OperationContext* opCtx, const OplogUpdateEntryArgs& args) {
failCollectionUpdates.executeIf(
[&](const BSONObj&) {
uasserted(40654,
str::stream() << "failCollectionUpdates failpoint enabled, namespace: "
<< args.nss.ns() << ", update: " << args.updateArgs.update
<< " on document with " << args.updateArgs.criteria);
},
[&](const BSONObj& data) {
// If the failpoint specifies no collection or matches the existing one, fail.
auto collElem = data["collectionNS"];
return !collElem || args.nss.ns() == collElem.String();
});
// Do not log a no-op operation; see SERVER-21738
if (args.updateArgs.update.isEmpty()) {
return;
}
auto txnParticipant = TransactionParticipant::get(opCtx);
const bool inMultiDocumentTransaction =
txnParticipant && opCtx->writesAreReplicated() && txnParticipant.transactionIsOpen();
OpTimeBundle opTime;
if (inMultiDocumentTransaction) {
auto operation = MutableOplogEntry::makeUpdateOperation(
args.nss, args.uuid, args.updateArgs.update, args.updateArgs.criteria);
if (args.updateArgs.preImageRecordingEnabledForCollection) {
invariant(args.updateArgs.preImageDoc);
operation.setPreImage(args.updateArgs.preImageDoc->getOwned());
}
txnParticipant.addTransactionOperation(opCtx, operation);
} else {
opTime = replLogUpdate(opCtx, args);
SessionTxnRecord sessionTxnRecord;
sessionTxnRecord.setLastWriteOpTime(opTime.writeOpTime);
sessionTxnRecord.setLastWriteDate(opTime.wallClockTime);
onWriteOpCompleted(opCtx, std::vector<StmtId>{args.updateArgs.stmtId}, sessionTxnRecord);
}
if (args.nss != NamespaceString::kSessionTransactionsTableNamespace) {
if (!args.updateArgs.fromMigrate) {
shardObserveUpdateOp(opCtx,
args.nss,
args.updateArgs.preImageDoc,
args.updateArgs.updatedDoc,
opTime.writeOpTime,
opTime.prePostImageOpTime,
inMultiDocumentTransaction);
}
}
if (args.nss.coll() == "system.js") {
Scope::storedFuncMod(opCtx);
} else if (args.nss.coll() == DurableViewCatalog::viewsCollectionName()) {
DurableViewCatalog::onExternalChange(opCtx, args.nss);
} else if (args.nss == NamespaceString::kServerConfigurationNamespace) {
// We must check server configuration collection writes for featureCompatibilityVersion
// document changes.
FeatureCompatibilityVersion::onInsertOrUpdate(opCtx, args.updateArgs.updatedDoc);
} else if (args.nss == NamespaceString::kSessionTransactionsTableNamespace &&
!opTime.writeOpTime.isNull()) {
MongoDSessionCatalog::observeDirectWriteToConfigTransactions(opCtx,
args.updateArgs.updatedDoc);
} else if (args.nss == NamespaceString::kConfigSettingsNamespace) {
ReadWriteConcernDefaults::get(opCtx).observeDirectWriteToConfigSettings(
opCtx, args.updateArgs.updatedDoc["_id"], args.updateArgs.updatedDoc);
}
}
void OpObserverImpl::aboutToDelete(OperationContext* opCtx,
NamespaceString const& nss,
BSONObj const& doc) {
documentKeyDecoration(opCtx).emplace(getDocumentKey(opCtx, nss, doc));
shardObserveAboutToDelete(opCtx, nss, doc);
}
void OpObserverImpl::onDelete(OperationContext* opCtx,
const NamespaceString& nss,
OptionalCollectionUUID uuid,
StmtId stmtId,
bool fromMigrate,
const boost::optional<BSONObj>& deletedDoc) {
auto optDocKey = documentKeyDecoration(opCtx);
invariant(optDocKey, nss.ns());
auto& documentKey = optDocKey.get();
auto txnParticipant = TransactionParticipant::get(opCtx);
const bool inMultiDocumentTransaction =
txnParticipant && opCtx->writesAreReplicated() && txnParticipant.transactionIsOpen();
OpTimeBundle opTime;
if (inMultiDocumentTransaction) {
auto operation =
MutableOplogEntry::makeDeleteOperation(nss, uuid.get(), documentKey.getShardKeyAndId());
if (deletedDoc) {
operation.setPreImage(deletedDoc->getOwned());
}
txnParticipant.addTransactionOperation(opCtx, operation);
} else {
opTime = replLogDelete(opCtx, nss, uuid, stmtId, fromMigrate, deletedDoc);
SessionTxnRecord sessionTxnRecord;
sessionTxnRecord.setLastWriteOpTime(opTime.writeOpTime);
sessionTxnRecord.setLastWriteDate(opTime.wallClockTime);
onWriteOpCompleted(opCtx, std::vector<StmtId>{stmtId}, sessionTxnRecord);
}
if (nss != NamespaceString::kSessionTransactionsTableNamespace) {
if (!fromMigrate) {
shardObserveDeleteOp(opCtx,
nss,
documentKey.getId(),
opTime.writeOpTime,
opTime.prePostImageOpTime,
inMultiDocumentTransaction);
}
}
if (nss.coll() == "system.js") {
Scope::storedFuncMod(opCtx);
} else if (nss.coll() == DurableViewCatalog::viewsCollectionName()) {
DurableViewCatalog::onExternalChange(opCtx, nss);
} else if (nss.isServerConfigurationCollection()) {
auto _id = documentKey.getId().firstElement();
if (_id.type() == BSONType::String &&
_id.String() == FeatureCompatibilityVersionParser::kParameterName)
uasserted(40670, "removing FeatureCompatibilityVersion document is not allowed");
} else if (nss == NamespaceString::kSessionTransactionsTableNamespace &&
!opTime.writeOpTime.isNull()) {
MongoDSessionCatalog::observeDirectWriteToConfigTransactions(opCtx, documentKey.getId());
} else if (nss == NamespaceString::kConfigSettingsNamespace) {
ReadWriteConcernDefaults::get(opCtx).observeDirectWriteToConfigSettings(
opCtx, documentKey.getId().firstElement(), boost::none);
}
}
void OpObserverImpl::onInternalOpMessage(OperationContext* opCtx,
const NamespaceString& nss,
const boost::optional<UUID> uuid,
const BSONObj& msgObj,
const boost::optional<BSONObj> o2MsgObj) {
MutableOplogEntry oplogEntry;
oplogEntry.setOpType(repl::OpTypeEnum::kNoop);
oplogEntry.setNss(nss);
oplogEntry.setUuid(uuid);
oplogEntry.setObject(msgObj);
oplogEntry.setObject2(o2MsgObj);
logOperation(opCtx, &oplogEntry);
}
void OpObserverImpl::onCreateCollection(OperationContext* opCtx,
Collection* coll,
const NamespaceString& collectionName,
const CollectionOptions& options,
const BSONObj& idIndex,
const OplogSlot& createOpTime) {
// do not replicate system.profile modifications
if (collectionName.isSystemDotProfile()) {
return;
}
auto txnParticipant = TransactionParticipant::get(opCtx);
const bool inMultiDocumentTransaction =
txnParticipant && opCtx->writesAreReplicated() && txnParticipant.transactionIsOpen();
if (inMultiDocumentTransaction) {
auto operation = MutableOplogEntry::makeCreateCommand(collectionName, options, idIndex);
txnParticipant.addTransactionOperation(opCtx, operation);
} else {
MutableOplogEntry oplogEntry;
oplogEntry.setOpType(repl::OpTypeEnum::kCommand);
oplogEntry.setNss(collectionName.getCommandNS());
oplogEntry.setUuid(options.uuid);
oplogEntry.setObject(
MutableOplogEntry::makeCreateCollCmdObj(collectionName, options, idIndex));
oplogEntry.setOpTime(createOpTime);
logOperation(opCtx, &oplogEntry);
}
}
void OpObserverImpl::onCollMod(OperationContext* opCtx,
const NamespaceString& nss,
OptionalCollectionUUID uuid,
const BSONObj& collModCmd,
const CollectionOptions& oldCollOptions,
boost::optional<IndexCollModInfo> indexInfo) {
if (!nss.isSystemDotProfile()) {
// do not replicate system.profile modifications
// Create the 'o2' field object. We save the old collection metadata and TTL expiration.
BSONObjBuilder o2Builder;
o2Builder.append("collectionOptions_old", oldCollOptions.toBSON());
if (indexInfo) {
if (indexInfo->oldExpireAfterSeconds) {
auto oldExpireAfterSeconds =
durationCount<Seconds>(indexInfo->oldExpireAfterSeconds.get());
o2Builder.append("expireAfterSeconds_old", oldExpireAfterSeconds);
}
if (indexInfo->oldHidden) {
auto oldHidden = indexInfo->oldHidden.get();
o2Builder.append("hidden_old", oldHidden);
}
}
MutableOplogEntry oplogEntry;
oplogEntry.setOpType(repl::OpTypeEnum::kCommand);
oplogEntry.setNss(nss.getCommandNS());
oplogEntry.setUuid(uuid);
oplogEntry.setObject(makeCollModCmdObj(collModCmd, oldCollOptions, indexInfo));
oplogEntry.setObject2(o2Builder.done());
logOperation(opCtx, &oplogEntry);
}
// Make sure the UUID values in the Collection metadata, the Collection object, and the UUID
// catalog are all present and equal.
invariant(opCtx->lockState()->isCollectionLockedForMode(nss, MODE_X));
auto databaseHolder = DatabaseHolder::get(opCtx);
auto db = databaseHolder->getDb(opCtx, nss.db());
// Some unit tests call the op observer on an unregistered Database.
if (!db) {
return;
}
Collection* coll = CollectionCatalog::get(opCtx).lookupCollectionByNamespace(opCtx, nss);
invariant(coll->uuid() == uuid);
invariant(DurableCatalog::get(opCtx)->isEqualToMetadataUUID(opCtx, coll->getCatalogId(), uuid));
}
void OpObserverImpl::onDropDatabase(OperationContext* opCtx, const std::string& dbName) {
MutableOplogEntry oplogEntry;
oplogEntry.setOpType(repl::OpTypeEnum::kCommand);
oplogEntry.setNss({dbName, "$cmd"});
oplogEntry.setObject(BSON("dropDatabase" << 1));
logOperation(opCtx, &oplogEntry);
uassert(
50714, "dropping the admin database is not allowed.", dbName != NamespaceString::kAdminDb);
if (dbName == NamespaceString::kSessionTransactionsTableNamespace.db()) {
MongoDSessionCatalog::invalidateAllSessions(opCtx);
}
}
repl::OpTime OpObserverImpl::onDropCollection(OperationContext* opCtx,
const NamespaceString& collectionName,
OptionalCollectionUUID uuid,
std::uint64_t numRecords,
const CollectionDropType dropType) {
if (!collectionName.isSystemDotProfile()) {
// Do not replicate system.profile modifications.
MutableOplogEntry oplogEntry;
oplogEntry.setOpType(repl::OpTypeEnum::kCommand);
oplogEntry.setNss(collectionName.getCommandNS());
oplogEntry.setUuid(uuid);
oplogEntry.setObject(BSON("drop" << collectionName.coll()));
oplogEntry.setObject2(makeObject2ForDropOrRename(numRecords));
logOperation(opCtx, &oplogEntry);
}
uassert(50715,
"dropping the server configuration collection (admin.system.version) is not allowed.",
collectionName != NamespaceString::kServerConfigurationNamespace);
if (collectionName.coll() == DurableViewCatalog::viewsCollectionName()) {
DurableViewCatalog::onSystemViewsCollectionDrop(opCtx, collectionName);
} else if (collectionName == NamespaceString::kSessionTransactionsTableNamespace) {
MongoDSessionCatalog::invalidateAllSessions(opCtx);
} else if (collectionName == NamespaceString::kConfigSettingsNamespace) {
ReadWriteConcernDefaults::get(opCtx).invalidate();
}
return {};
}
void OpObserverImpl::onDropIndex(OperationContext* opCtx,
const NamespaceString& nss,
OptionalCollectionUUID uuid,
const std::string& indexName,
const BSONObj& indexInfo) {
MutableOplogEntry oplogEntry;
oplogEntry.setOpType(repl::OpTypeEnum::kCommand);
oplogEntry.setNss(nss.getCommandNS());
oplogEntry.setUuid(uuid);
oplogEntry.setObject(BSON("dropIndexes" << nss.coll() << "index" << indexName));
oplogEntry.setObject2(indexInfo);
logOperation(opCtx, &oplogEntry);
}
repl::OpTime OpObserverImpl::preRenameCollection(OperationContext* const opCtx,
const NamespaceString& fromCollection,
const NamespaceString& toCollection,
OptionalCollectionUUID uuid,
OptionalCollectionUUID dropTargetUUID,
std::uint64_t numRecords,
bool stayTemp) {
BSONObjBuilder builder;
builder.append("renameCollection", fromCollection.ns());
builder.append("to", toCollection.ns());
builder.append("stayTemp", stayTemp);
if (dropTargetUUID) {
dropTargetUUID->appendToBuilder(&builder, "dropTarget");
}
MutableOplogEntry oplogEntry;
oplogEntry.setOpType(repl::OpTypeEnum::kCommand);
oplogEntry.setNss(fromCollection.getCommandNS());
oplogEntry.setUuid(uuid);
oplogEntry.setObject(builder.done());
if (dropTargetUUID)
oplogEntry.setObject2(makeObject2ForDropOrRename(numRecords));
logOperation(opCtx, &oplogEntry);
return {};
}
void OpObserverImpl::postRenameCollection(OperationContext* const opCtx,
const NamespaceString& fromCollection,
const NamespaceString& toCollection,
OptionalCollectionUUID uuid,
OptionalCollectionUUID dropTargetUUID,
bool stayTemp) {
if (fromCollection.isSystemDotViews())
DurableViewCatalog::onExternalChange(opCtx, fromCollection);
if (toCollection.isSystemDotViews())
DurableViewCatalog::onExternalChange(opCtx, toCollection);
}
void OpObserverImpl::onRenameCollection(OperationContext* const opCtx,
const NamespaceString& fromCollection,
const NamespaceString& toCollection,
OptionalCollectionUUID uuid,
OptionalCollectionUUID dropTargetUUID,
std::uint64_t numRecords,
bool stayTemp) {
preRenameCollection(
opCtx, fromCollection, toCollection, uuid, dropTargetUUID, numRecords, stayTemp);
postRenameCollection(opCtx, fromCollection, toCollection, uuid, dropTargetUUID, stayTemp);
}
void OpObserverImpl::onApplyOps(OperationContext* opCtx,
const std::string& dbName,
const BSONObj& applyOpCmd) {
MutableOplogEntry oplogEntry;
oplogEntry.setOpType(repl::OpTypeEnum::kCommand);
oplogEntry.setNss({dbName, "$cmd"});
oplogEntry.setObject(applyOpCmd);
logOperation(opCtx, &oplogEntry);
}
void OpObserverImpl::onEmptyCapped(OperationContext* opCtx,
const NamespaceString& collectionName,
OptionalCollectionUUID uuid) {
if (!collectionName.isSystemDotProfile()) {
// Do not replicate system.profile modifications
MutableOplogEntry oplogEntry;
oplogEntry.setOpType(repl::OpTypeEnum::kCommand);
oplogEntry.setNss(collectionName.getCommandNS());
oplogEntry.setUuid(uuid);
oplogEntry.setObject(BSON("emptycapped" << collectionName.coll()));
logOperation(opCtx, &oplogEntry);
}
}
namespace {
// Accepts an empty BSON builder and appends the given transaction statements to an 'applyOps' array
// field. Appends as many operations as possible until either the constructed object exceeds the
// 16MB limit or the maximum number of transaction statements allowed in one entry.
//
// Returns an iterator to the first statement that wasn't packed into the applyOps object.
std::vector<repl::ReplOperation>::iterator packTransactionStatementsForApplyOps(
BSONObjBuilder* applyOpsBuilder,
std::vector<repl::ReplOperation>::iterator stmtBegin,
std::vector<repl::ReplOperation>::iterator stmtEnd) {
std::vector<repl::ReplOperation>::iterator stmtIter;
BSONArrayBuilder opsArray(applyOpsBuilder->subarrayStart("applyOps"_sd));
for (stmtIter = stmtBegin; stmtIter != stmtEnd; stmtIter++) {
const auto& stmt = *stmtIter;
// Stop packing when either number of transaction operations is reached, or when the next
// one would put the array over the maximum BSON Object User Size. We rely on the
// head room between BSONObjMaxUserSize and BSONObjMaxInternalSize to cover the
// BSON overhead and the other applyOps fields. But if the array with a single operation
// exceeds BSONObjMaxUserSize, we still log it, as a single max-length operation
// should be able to be applied.
if (opsArray.arrSize() == gMaxNumberOfTransactionOperationsInSingleOplogEntry ||
(opsArray.arrSize() > 0 &&
(opsArray.len() + OplogEntry::getDurableReplOperationSize(stmt) > BSONObjMaxUserSize)))
break;
opsArray.append(stmt.toBSON());
}
try {
// BSONArrayBuilder will throw a BSONObjectTooLarge exception if we exceeded the max BSON
// size.
opsArray.done();
} catch (const AssertionException& e) {
// Change the error code to TransactionTooLarge if it is BSONObjectTooLarge.
uassert(ErrorCodes::TransactionTooLarge,
e.reason(),
e.code() != ErrorCodes::BSONObjectTooLarge);
throw;
}
return stmtIter;
}
// Logs one applyOps entry and may update the transactions table. Assumes that the given BSON
// builder object already has an 'applyOps' field appended pointing to the desired array of ops
// i.e. { "applyOps" : [op1, op2, ...] }
//
// @param txnState the 'state' field of the transaction table entry update.
// @param startOpTime the optime of the 'startOpTime' field of the transaction table entry update.
// If boost::none, no 'startOpTime' field will be included in the new transaction table entry. Only
// meaningful if 'updateTxnTable' is true.
// @param updateTxnTable determines whether the transactions table will updated after the oplog
// entry is written.
//
// Returns the optime of the written oplog entry.
OpTimeBundle logApplyOpsForTransaction(OperationContext* opCtx,
MutableOplogEntry* oplogEntry,
boost::optional<DurableTxnStateEnum> txnState,
boost::optional<repl::OpTime> startOpTime,
const bool updateTxnTable) {
oplogEntry->setOpType(repl::OpTypeEnum::kCommand);
oplogEntry->setNss({"admin", "$cmd"});
oplogEntry->setSessionId(opCtx->getLogicalSessionId());
oplogEntry->setTxnNumber(opCtx->getTxnNumber());
try {
OpTimeBundle times;
times.writeOpTime = logOperation(opCtx, oplogEntry);
times.wallClockTime = oplogEntry->getWallClockTime();
if (updateTxnTable) {
SessionTxnRecord sessionTxnRecord;
sessionTxnRecord.setLastWriteOpTime(times.writeOpTime);
sessionTxnRecord.setLastWriteDate(times.wallClockTime);
sessionTxnRecord.setState(txnState);
sessionTxnRecord.setStartOpTime(startOpTime);
onWriteOpCompleted(opCtx, {}, sessionTxnRecord);
}
return times;
} catch (const AssertionException& e) {
// Change the error code to TransactionTooLarge if it is BSONObjectTooLarge.
uassert(ErrorCodes::TransactionTooLarge,
e.reason(),
e.code() != ErrorCodes::BSONObjectTooLarge);
throw;
}
MONGO_UNREACHABLE;
}
// Logs transaction oplog entries for preparing a transaction or committing an unprepared
// transaction. This includes the in-progress 'partialTxn' oplog entries followed by the implicit
// prepare or commit entry. If the 'prepare' argument is true, it will log entries for a prepared
// transaction. Otherwise, it logs entries for an unprepared transaction. The total number of oplog
// entries written will be <= the size of the given 'stmts' vector, and will depend on how many
// transaction statements are given, the data size of each statement, and the
// 'maxNumberOfTransactionOperationsInSingleOplogEntry' server parameter.
//
// This function expects that the size of 'oplogSlots' be at least as big as the size of 'stmts' in
// the worst case, where each operation requires an applyOps entry of its own. If there are more
// oplog slots than applyOps operations are written, the number of oplog slots corresponding to the
// number of applyOps written will be used. It also expects that the vector of given statements is
// non-empty.
//
// In the case of writing entries for a prepared transaction, the last oplog entry (i.e. the
// implicit prepare) will always be written using the last oplog slot given, even if this means
// skipping over some reserved slots.
//
// The number of oplog entries written is returned.
int logOplogEntriesForTransaction(OperationContext* opCtx,
std::vector<repl::ReplOperation>* stmts,
const std::vector<OplogSlot>& oplogSlots,
size_t numberOfPreImagesToWrite,
bool prepare) {
invariant(!stmts->empty());
invariant(stmts->size() <= oplogSlots.size());
// Storage transaction commit is the last place inside a transaction that can throw an
// exception. In order to safely allow exceptions to be thrown at that point, this function must
// be called from an outer WriteUnitOfWork in order to be rolled back upon reaching the
// exception.
invariant(opCtx->lockState()->inAWriteUnitOfWork());
const auto txnParticipant = TransactionParticipant::get(opCtx);
OpTimeBundle prevWriteOpTime;
auto numEntriesWritten = 0;
// Writes to the oplog only require a Global intent lock. Guaranteed by
// OplogSlotReserver.
invariant(opCtx->lockState()->isWriteLocked());
prevWriteOpTime.writeOpTime = txnParticipant.getLastWriteOpTime();
auto currOplogSlot = oplogSlots.begin();
if (numberOfPreImagesToWrite > 0) {
for (auto& statement : *stmts) {
if (statement.getPreImage().isEmpty()) {
continue;
}
auto slot = *currOplogSlot;
++currOplogSlot;
MutableOplogEntry preImageEntry;
preImageEntry.setOpType(repl::OpTypeEnum::kNoop);
preImageEntry.setObject(statement.getPreImage());
preImageEntry.setNss(statement.getNss());
preImageEntry.setUuid(statement.getUuid());
preImageEntry.setOpTime(slot);
auto opTime = logOperation(opCtx, &preImageEntry);
statement.setPreImageOpTime(opTime);
}
}
// At the beginning of each loop iteration below, 'stmtsIter' will always point to the
// first statement of the sequence of remaining, unpacked transaction statements. If all
// statements have been packed, it should point to stmts.end(), which is the loop's
// termination condition.
auto stmtsIter = stmts->begin();
while (stmtsIter != stmts->end()) {
BSONObjBuilder applyOpsBuilder;
auto nextStmt =
packTransactionStatementsForApplyOps(&applyOpsBuilder, stmtsIter, stmts->end());
// If we packed the last op, then the next oplog entry we log should be the implicit
// commit or implicit prepare, i.e. we omit the 'partialTxn' field.
auto firstOp = stmtsIter == stmts->begin();
auto lastOp = nextStmt == stmts->end();
auto implicitCommit = lastOp && !prepare;
auto implicitPrepare = lastOp && prepare;
auto isPartialTxn = !lastOp;
// A 'prepare' oplog entry should never include a 'partialTxn' field.
invariant(!(isPartialTxn && implicitPrepare));
if (implicitPrepare) {
applyOpsBuilder.append("prepare", true);
}
if (isPartialTxn) {
applyOpsBuilder.append("partialTxn", true);
}
// The 'count' field gives the total number of individual operations in the
// transaction, and is included on a non-initial implicit commit or prepare entry.
if (lastOp && !firstOp) {
applyOpsBuilder.append("count", static_cast<long long>(stmts->size()));
}
// For both prepared and unprepared transactions, update the transactions table on
// the first and last op.
auto updateTxnTable = firstOp || lastOp;
// Use the next reserved oplog slot. In the special case of writing the implicit
// 'prepare' oplog entry, we use the last reserved oplog slot, since callers of this
// function will expect that timestamp to be used as the 'prepare' timestamp. This
// may mean we skipped over some reserved slots, but there's no harm in that.
auto oplogSlot = implicitPrepare ? oplogSlots.back() : *currOplogSlot++;
// The first optime of the transaction is always the first oplog slot, except in the
// case of a single prepare oplog entry.
auto firstOpTimeOfTxn =
(implicitPrepare && firstOp) ? oplogSlots.back() : oplogSlots.front();
// We always write the startOpTime field, which is the first optime of the
// transaction, except when transitioning to 'committed' state, in which it should
// no longer be set.
auto startOpTime = boost::make_optional(!implicitCommit, firstOpTimeOfTxn);
MutableOplogEntry oplogEntry;
oplogEntry.setOpTime(oplogSlot);
oplogEntry.setPrevWriteOpTimeInTransaction(prevWriteOpTime.writeOpTime);
oplogEntry.setObject(applyOpsBuilder.done());
auto txnState = isPartialTxn
? DurableTxnStateEnum::kInProgress
: (implicitPrepare ? DurableTxnStateEnum::kPrepared : DurableTxnStateEnum::kCommitted);
prevWriteOpTime =
logApplyOpsForTransaction(opCtx, &oplogEntry, txnState, startOpTime, updateTxnTable);
// Advance the iterator to the beginning of the remaining unpacked statements.
stmtsIter = nextStmt;
numEntriesWritten++;
}
return numEntriesWritten;
}
void logCommitOrAbortForPreparedTransaction(OperationContext* opCtx,
MutableOplogEntry* oplogEntry,
DurableTxnStateEnum durableState) {
oplogEntry->setOpType(repl::OpTypeEnum::kCommand);
oplogEntry->setNss({"admin", "$cmd"});
oplogEntry->setSessionId(opCtx->getLogicalSessionId());
oplogEntry->setTxnNumber(opCtx->getTxnNumber());
oplogEntry->setPrevWriteOpTimeInTransaction(
TransactionParticipant::get(opCtx).getLastWriteOpTime());
// There should not be a parent WUOW outside of this one. This guarantees the safety of the
// write conflict retry loop.
invariant(!opCtx->lockState()->inAWriteUnitOfWork());
// We must not have a maximum lock timeout, since writing the commit or abort oplog entry for a
// prepared transaction must always succeed.
invariant(!opCtx->lockState()->hasMaxLockTimeout());
writeConflictRetry(
opCtx, "onPreparedTransactionCommitOrAbort", NamespaceString::kRsOplogNamespace.ns(), [&] {
// Writes to the oplog only require a Global intent lock. Guaranteed by
// OplogSlotReserver.
invariant(opCtx->lockState()->isWriteLocked());
WriteUnitOfWork wuow(opCtx);
const auto oplogOpTime = logOperation(opCtx, oplogEntry);
invariant(oplogEntry->getOpTime().isNull() || oplogEntry->getOpTime() == oplogOpTime);
SessionTxnRecord sessionTxnRecord;
sessionTxnRecord.setLastWriteOpTime(oplogOpTime);
sessionTxnRecord.setLastWriteDate(oplogEntry->getWallClockTime());
sessionTxnRecord.setState(durableState);
onWriteOpCompleted(opCtx, {}, sessionTxnRecord);
wuow.commit();
});
}
} // namespace
void OpObserverImpl::onUnpreparedTransactionCommit(OperationContext* opCtx,
std::vector<repl::ReplOperation>* statements,
size_t numberOfPreImagesToWrite) {
invariant(opCtx->getTxnNumber());
if (!opCtx->writesAreReplicated()) {
return;
}
// It is possible that the transaction resulted in no changes. In that case, we should
// not write an empty applyOps entry.
if (statements->empty())
return;
repl::OpTime commitOpTime;
// Reserve all the optimes in advance, so we only need to get the optime mutex once. We
// reserve enough entries for all statements in the transaction.
auto oplogSlots = repl::getNextOpTimes(opCtx, statements->size() + numberOfPreImagesToWrite);
if (MONGO_unlikely(hangAndFailUnpreparedCommitAfterReservingOplogSlot.shouldFail())) {
hangAndFailUnpreparedCommitAfterReservingOplogSlot.pauseWhileSet(opCtx);
uasserted(51268, "hangAndFailUnpreparedCommitAfterReservingOplogSlot fail point enabled");
}
// Log in-progress entries for the transaction along with the implicit commit.
int numOplogEntries = logOplogEntriesForTransaction(
opCtx, statements, oplogSlots, numberOfPreImagesToWrite, false);
commitOpTime = oplogSlots[numOplogEntries - 1];
invariant(!commitOpTime.isNull());
shardObserveTransactionPrepareOrUnpreparedCommit(opCtx, *statements, commitOpTime);
}
void OpObserverImpl::onPreparedTransactionCommit(
OperationContext* opCtx,
OplogSlot commitOplogEntryOpTime,
Timestamp commitTimestamp,
const std::vector<repl::ReplOperation>& statements) noexcept {
invariant(opCtx->getTxnNumber());
if (!opCtx->writesAreReplicated()) {
return;
}
invariant(!commitTimestamp.isNull());
MutableOplogEntry oplogEntry;
oplogEntry.setOpTime(commitOplogEntryOpTime);
CommitTransactionOplogObject cmdObj;
cmdObj.setCommitTimestamp(commitTimestamp);
oplogEntry.setObject(cmdObj.toBSON());
logCommitOrAbortForPreparedTransaction(opCtx, &oplogEntry, DurableTxnStateEnum::kCommitted);
}
void OpObserverImpl::onTransactionPrepare(OperationContext* opCtx,
const std::vector<OplogSlot>& reservedSlots,
std::vector<repl::ReplOperation>* statements,
size_t numberOfPreImagesToWrite) {
invariant(!reservedSlots.empty());
const auto prepareOpTime = reservedSlots.back();
invariant(opCtx->getTxnNumber());
invariant(!prepareOpTime.isNull());
// Don't write oplog entry on secondaries.
if (!opCtx->writesAreReplicated()) {
return;
}
{
// We should have reserved enough slots.
invariant(reservedSlots.size() >= statements->size());
TransactionParticipant::SideTransactionBlock sideTxn(opCtx);
writeConflictRetry(
opCtx, "onTransactionPrepare", NamespaceString::kRsOplogNamespace.ns(), [&] {
// Writes to the oplog only require a Global intent lock. Guaranteed by
// OplogSlotReserver.
invariant(opCtx->lockState()->isWriteLocked());
WriteUnitOfWork wuow(opCtx);
// It is possible that the transaction resulted in no changes, In that case, we
// should not write any operations other than the prepare oplog entry.
if (!statements->empty()) {
// We had reserved enough oplog slots for the worst case where each operation
// produced one oplog entry. When operations are smaller and can be packed, we
// will waste the extra slots. The implicit prepare oplog entry will still use
// the last reserved slot, because the transaction participant has already used
// that as the prepare time.
logOplogEntriesForTransaction(opCtx,
statements,
reservedSlots,
numberOfPreImagesToWrite,
true /* prepare */);
} else {
// Log an empty 'prepare' oplog entry.
// We need to have at least one reserved slot.
invariant(reservedSlots.size() > 0);
BSONObjBuilder applyOpsBuilder;
BSONArrayBuilder opsArray(applyOpsBuilder.subarrayStart("applyOps"_sd));
opsArray.done();
applyOpsBuilder.append("prepare", true);
auto oplogSlot = reservedSlots.front();
MutableOplogEntry oplogEntry;
oplogEntry.setOpTime(oplogSlot);
oplogEntry.setPrevWriteOpTimeInTransaction(repl::OpTime());
oplogEntry.setObject(applyOpsBuilder.done());
logApplyOpsForTransaction(opCtx,
&oplogEntry,
DurableTxnStateEnum::kPrepared,
oplogSlot,
true /* updateTxnTable */);
}
wuow.commit();
});
}
shardObserveTransactionPrepareOrUnpreparedCommit(opCtx, *statements, prepareOpTime);
}
void OpObserverImpl::onTransactionAbort(OperationContext* opCtx,
boost::optional<OplogSlot> abortOplogEntryOpTime) {
invariant(opCtx->getTxnNumber());
if (!opCtx->writesAreReplicated()) {
return;
}
auto txnParticipant = TransactionParticipant::get(opCtx);
invariant(txnParticipant);
if (!abortOplogEntryOpTime) {
invariant(!txnParticipant.transactionIsCommitted());
return;
}
MutableOplogEntry oplogEntry;
oplogEntry.setOpTime(*abortOplogEntryOpTime);
AbortTransactionOplogObject cmdObj;
oplogEntry.setObject(cmdObj.toBSON());
logCommitOrAbortForPreparedTransaction(opCtx, &oplogEntry, DurableTxnStateEnum::kAborted);
}
void OpObserverImpl::onReplicationRollback(OperationContext* opCtx,
const RollbackObserverInfo& rbInfo) {
// Reset the key manager cache.
auto validator = LogicalTimeValidator::get(opCtx);
if (validator) {
validator->resetKeyManagerCache();
}
// Check if the shard identity document rolled back.
if (rbInfo.shardIdentityRolledBack) {
fassertFailedNoTrace(50712);
}
// Force the config server to update its shard registry on next access. Otherwise it may have
// the stale data that has been just rolled back.
if (serverGlobalParams.clusterRole == ClusterRole::ConfigServer) {
if (auto shardRegistry = Grid::get(opCtx)->shardRegistry()) {
shardRegistry->clearEntries();
}
}
// Force the default read/write concern cache to reload on next access in case the defaults
// document was rolled back.
ReadWriteConcernDefaults::get(opCtx).invalidate();
// Make sure the in-memory FCV matches the on-disk FCV.
FeatureCompatibilityVersion::onReplicationRollback(opCtx);
}
} // namespace mongo
|