/
initial_syncer.cpp
2200 lines (1947 loc) · 97.4 KB
/
initial_syncer.cpp
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
/**
* 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::kReplicationInitialSync
#include "mongo/platform/basic.h"
#include "initial_syncer.h"
#include <algorithm>
#include <memory>
#include <utility>
#include "mongo/base/counter.h"
#include "mongo/base/status.h"
#include "mongo/bson/simple_bsonobj_comparator.h"
#include "mongo/bson/util/bson_extract.h"
#include "mongo/client/fetcher.h"
#include "mongo/client/remote_command_retry_scheduler.h"
#include "mongo/db/commands/feature_compatibility_version_parser.h"
#include "mongo/db/commands/server_status_metric.h"
#include "mongo/db/concurrency/d_concurrency.h"
#include "mongo/db/index_builds_coordinator.h"
#include "mongo/db/jsobj.h"
#include "mongo/db/namespace_string.h"
#include "mongo/db/repl/all_database_cloner.h"
#include "mongo/db/repl/initial_sync_state.h"
#include "mongo/db/repl/member_state.h"
#include "mongo/db/repl/oplog_buffer.h"
#include "mongo/db/repl/oplog_fetcher.h"
#include "mongo/db/repl/optime.h"
#include "mongo/db/repl/repl_server_parameters_gen.h"
#include "mongo/db/repl/replication_consistency_markers.h"
#include "mongo/db/repl/replication_process.h"
#include "mongo/db/repl/storage_interface.h"
#include "mongo/db/repl/sync_source_selector.h"
#include "mongo/db/repl/tenant_migration_donor_util.h"
#include "mongo/db/repl/transaction_oplog_application.h"
#include "mongo/db/session_txn_record_gen.h"
#include "mongo/executor/task_executor.h"
#include "mongo/executor/thread_pool_task_executor.h"
#include "mongo/logv2/log.h"
#include "mongo/rpc/metadata/repl_set_metadata.h"
#include "mongo/util/assert_util.h"
#include "mongo/util/destructor_guard.h"
#include "mongo/util/fail_point.h"
#include "mongo/util/scopeguard.h"
#include "mongo/util/str.h"
#include "mongo/util/system_clock_source.h"
#include "mongo/util/time_support.h"
#include "mongo/util/timer.h"
namespace mongo {
namespace repl {
// Failpoint for initial sync
MONGO_FAIL_POINT_DEFINE(failInitialSyncWithBadHost);
// Failpoint which fails initial sync and leaves an oplog entry in the buffer.
MONGO_FAIL_POINT_DEFINE(failInitSyncWithBufferedEntriesLeft);
// Failpoint which causes the initial sync function to hang after getting the oldest active
// transaction timestamp from the sync source.
MONGO_FAIL_POINT_DEFINE(initialSyncHangAfterGettingBeginFetchingTimestamp);
// Failpoint which causes the initial sync function to hang before creating shared data and
// splitting control flow between the oplog fetcher and the cloners.
MONGO_FAIL_POINT_DEFINE(initialSyncHangBeforeSplittingControlFlow);
// Failpoint which causes the initial sync function to hang before copying databases.
MONGO_FAIL_POINT_DEFINE(initialSyncHangBeforeCopyingDatabases);
// Failpoint which causes the initial sync function to hang before finishing.
MONGO_FAIL_POINT_DEFINE(initialSyncHangBeforeFinish);
// Failpoint which causes the initial sync function to hang before creating the oplog.
MONGO_FAIL_POINT_DEFINE(initialSyncHangBeforeCreatingOplog);
// Failpoint which stops the applier.
MONGO_FAIL_POINT_DEFINE(rsSyncApplyStop);
// Failpoint which causes the initial sync function to hang after cloning all databases.
MONGO_FAIL_POINT_DEFINE(initialSyncHangAfterDataCloning);
// Failpoint which skips clearing _initialSyncState after a successful initial sync attempt.
MONGO_FAIL_POINT_DEFINE(skipClearInitialSyncState);
// Failpoint which causes the initial sync function to fail and hang before starting a new attempt.
MONGO_FAIL_POINT_DEFINE(failAndHangInitialSync);
// Failpoint which fails initial sync before it applies the next batch of oplog entries.
MONGO_FAIL_POINT_DEFINE(failInitialSyncBeforeApplyingBatch);
// Failpoint which fasserts if applying a batch fails.
MONGO_FAIL_POINT_DEFINE(initialSyncFassertIfApplyingBatchFails);
// Failpoint which causes the initial sync function to hang before stopping the oplog fetcher.
MONGO_FAIL_POINT_DEFINE(initialSyncHangBeforeCompletingOplogFetching);
// Failpoints for synchronization, shared with cloners.
extern FailPoint initialSyncFuzzerSynchronizationPoint1;
extern FailPoint initialSyncFuzzerSynchronizationPoint2;
namespace {
using namespace executor;
using CallbackArgs = executor::TaskExecutor::CallbackArgs;
using Event = executor::TaskExecutor::EventHandle;
using Handle = executor::TaskExecutor::CallbackHandle;
using QueryResponseStatus = StatusWith<Fetcher::QueryResponse>;
using UniqueLock = stdx::unique_lock<Latch>;
using LockGuard = stdx::lock_guard<Latch>;
// Used to reset the oldest timestamp during initial sync to a non-null timestamp.
const Timestamp kTimestampOne(0, 1);
// The number of initial sync attempts that have failed since server startup. Each instance of
// InitialSyncer may run multiple attempts to fulfill an initial sync request that is triggered
// when InitialSyncer::startup() is called.
Counter64 initialSyncFailedAttempts;
// The number of initial sync requests that have been requested and failed. Each instance of
// InitialSyncer (upon successful startup()) corresponds to a single initial sync request.
// This value does not include the number of times where a InitialSyncer is created successfully
// but failed in startup().
Counter64 initialSyncFailures;
// The number of initial sync requests that have been requested and completed successfully. Each
// instance of InitialSyncer corresponds to a single initial sync request.
Counter64 initialSyncCompletes;
ServerStatusMetricField<Counter64> displaySSInitialSyncFailedAttempts(
"repl.initialSync.failedAttempts", &initialSyncFailedAttempts);
ServerStatusMetricField<Counter64> displaySSInitialSyncFailures("repl.initialSync.failures",
&initialSyncFailures);
ServerStatusMetricField<Counter64> displaySSInitialSyncCompleted("repl.initialSync.completed",
&initialSyncCompletes);
ServiceContext::UniqueOperationContext makeOpCtx() {
return cc().makeOperationContext();
}
StatusWith<OpTimeAndWallTime> parseOpTimeAndWallTime(const QueryResponseStatus& fetchResult) {
if (!fetchResult.isOK()) {
return fetchResult.getStatus();
}
const auto docs = fetchResult.getValue().documents;
const auto hasDoc = docs.begin() != docs.end();
if (!hasDoc) {
return StatusWith<OpTimeAndWallTime>{ErrorCodes::NoMatchingDocument,
"no oplog entry found"};
}
return OpTimeAndWallTime::parseOpTimeAndWallTimeFromOplogEntry(docs.front());
}
void pauseAtInitialSyncFuzzerSyncronizationPoints(std::string msg) {
// Set and unset by the InitialSyncTest fixture to cause initial sync to pause so that the
// Initial Sync Fuzzer can run commands on the sync source.
if (MONGO_unlikely(initialSyncFuzzerSynchronizationPoint1.shouldFail())) {
LOGV2(21158,
"initialSyncFuzzerSynchronizationPoint1 fail point enabled",
"failpointMessage"_attr = msg);
initialSyncFuzzerSynchronizationPoint1.pauseWhileSet();
}
if (MONGO_unlikely(initialSyncFuzzerSynchronizationPoint2.shouldFail())) {
LOGV2(21160, "initialSyncFuzzerSynchronizationPoint2 fail point enabled");
initialSyncFuzzerSynchronizationPoint2.pauseWhileSet();
}
}
} // namespace
InitialSyncer::InitialSyncer(
InitialSyncerOptions opts,
std::unique_ptr<DataReplicatorExternalState> dataReplicatorExternalState,
ThreadPool* writerPool,
StorageInterface* storage,
ReplicationProcess* replicationProcess,
const OnCompletionFn& onCompletion)
: _fetchCount(0),
_opts(opts),
_dataReplicatorExternalState(std::move(dataReplicatorExternalState)),
_exec(_dataReplicatorExternalState->getSharedTaskExecutor()),
_clonerExec(_exec),
_writerPool(writerPool),
_storage(storage),
_replicationProcess(replicationProcess),
_onCompletion(onCompletion),
_createClientFn(
[] { return std::make_unique<DBClientConnection>(true /* autoReconnect */); }),
_createOplogFetcherFn(CreateOplogFetcherFn::get()) {
uassert(ErrorCodes::BadValue, "task executor cannot be null", _exec);
uassert(ErrorCodes::BadValue, "invalid storage interface", _storage);
uassert(ErrorCodes::BadValue, "invalid replication process", _replicationProcess);
uassert(ErrorCodes::BadValue, "invalid getMyLastOptime function", _opts.getMyLastOptime);
uassert(ErrorCodes::BadValue, "invalid setMyLastOptime function", _opts.setMyLastOptime);
uassert(ErrorCodes::BadValue, "invalid resetOptimes function", _opts.resetOptimes);
uassert(ErrorCodes::BadValue, "invalid sync source selector", _opts.syncSourceSelector);
uassert(ErrorCodes::BadValue, "callback function cannot be null", _onCompletion);
}
InitialSyncer::~InitialSyncer() {
DESTRUCTOR_GUARD({
shutdown().transitional_ignore();
join();
});
}
bool InitialSyncer::isActive() const {
stdx::lock_guard<Latch> lock(_mutex);
return _isActive_inlock();
}
bool InitialSyncer::_isActive_inlock() const {
return State::kRunning == _state || State::kShuttingDown == _state;
}
Status InitialSyncer::startup(OperationContext* opCtx,
std::uint32_t initialSyncMaxAttempts) noexcept {
invariant(opCtx);
invariant(initialSyncMaxAttempts >= 1U);
stdx::lock_guard<Latch> lock(_mutex);
switch (_state) {
case State::kPreStart:
_state = State::kRunning;
break;
case State::kRunning:
return Status(ErrorCodes::IllegalOperation, "initial syncer already started");
case State::kShuttingDown:
return Status(ErrorCodes::ShutdownInProgress, "initial syncer shutting down");
case State::kComplete:
return Status(ErrorCodes::ShutdownInProgress, "initial syncer completed");
}
_setUp_inlock(opCtx, initialSyncMaxAttempts);
// Start first initial sync attempt.
std::uint32_t initialSyncAttempt = 0;
_attemptExec = std::make_unique<executor::ScopedTaskExecutor>(
_exec, Status(ErrorCodes::CallbackCanceled, "Initial Sync Attempt Canceled"));
_clonerAttemptExec = std::make_unique<executor::ScopedTaskExecutor>(
_clonerExec, Status(ErrorCodes::CallbackCanceled, "Initial Sync Attempt Canceled"));
auto status = _scheduleWorkAndSaveHandle_inlock(
[=](const executor::TaskExecutor::CallbackArgs& args) {
_startInitialSyncAttemptCallback(args, initialSyncAttempt, initialSyncMaxAttempts);
},
&_startInitialSyncAttemptHandle,
str::stream() << "_startInitialSyncAttemptCallback-" << initialSyncAttempt);
if (!status.isOK()) {
_state = State::kComplete;
return status;
}
return Status::OK();
}
Status InitialSyncer::shutdown() {
stdx::lock_guard<Latch> lock(_mutex);
switch (_state) {
case State::kPreStart:
// Transition directly from PreStart to Complete if not started yet.
_state = State::kComplete;
return Status::OK();
case State::kRunning:
_state = State::kShuttingDown;
break;
case State::kShuttingDown:
case State::kComplete:
// Nothing to do if we are already in ShuttingDown or Complete state.
return Status::OK();
}
_cancelRemainingWork_inlock();
return Status::OK();
}
void InitialSyncer::cancelCurrentAttempt() {
stdx::lock_guard lk(_mutex);
if (_isActive_inlock()) {
LOGV2_DEBUG(4427201,
1,
"Cancelling the current initial sync attempt.",
"currentAttempt"_attr = _stats.failedInitialSyncAttempts + 1);
_cancelRemainingWork_inlock();
} else {
LOGV2_DEBUG(4427202,
1,
"There is no initial sync attempt to cancel because the initial syncer is not "
"currently active.");
}
}
void InitialSyncer::_cancelRemainingWork_inlock() {
_cancelHandle_inlock(_startInitialSyncAttemptHandle);
_cancelHandle_inlock(_chooseSyncSourceHandle);
_cancelHandle_inlock(_getBaseRollbackIdHandle);
_cancelHandle_inlock(_getLastRollbackIdHandle);
_cancelHandle_inlock(_getNextApplierBatchHandle);
_shutdownComponent_inlock(_oplogFetcher);
if (_sharedData) {
// We actually hold the required lock, but the lock object itself is not passed through.
_clearRetriableError(WithLock::withoutLock());
stdx::lock_guard<InitialSyncSharedData> lock(*_sharedData);
_sharedData->setStatusIfOK(
lock, Status{ErrorCodes::CallbackCanceled, "Initial sync attempt canceled"});
}
if (_client) {
_client->shutdownAndDisallowReconnect();
}
_shutdownComponent_inlock(_applier);
_shutdownComponent_inlock(_fCVFetcher);
_shutdownComponent_inlock(_lastOplogEntryFetcher);
_shutdownComponent_inlock(_beginFetchingOpTimeFetcher);
(*_attemptExec)->shutdown();
(*_clonerAttemptExec)->shutdown();
_attemptCanceled = true;
}
void InitialSyncer::join() {
stdx::unique_lock<Latch> lk(_mutex);
_stateCondition.wait(lk, [this]() { return !_isActive_inlock(); });
}
InitialSyncer::State InitialSyncer::getState_forTest() const {
stdx::lock_guard<Latch> lk(_mutex);
return _state;
}
Date_t InitialSyncer::getWallClockTime_forTest() const {
stdx::lock_guard<Latch> lk(_mutex);
return _lastApplied.wallTime;
}
void InitialSyncer::setAllowedOutageDuration_forTest(Milliseconds allowedOutageDuration) {
stdx::lock_guard<Latch> lk(_mutex);
_allowedOutageDuration = allowedOutageDuration;
if (_sharedData) {
stdx::lock_guard<InitialSyncSharedData> lk(*_sharedData);
_sharedData->setAllowedOutageDuration_forTest(lk, allowedOutageDuration);
}
}
bool InitialSyncer::_isShuttingDown() const {
stdx::lock_guard<Latch> lock(_mutex);
return _isShuttingDown_inlock();
}
bool InitialSyncer::_isShuttingDown_inlock() const {
return State::kShuttingDown == _state;
}
std::string InitialSyncer::getDiagnosticString() const {
LockGuard lk(_mutex);
str::stream out;
out << "InitialSyncer -"
<< " opts: " << _opts.toString() << " oplogFetcher: " << _oplogFetcher->toString()
<< " opsBuffered: " << _oplogBuffer->getSize() << " active: " << _isActive_inlock()
<< " shutting down: " << _isShuttingDown_inlock();
if (_initialSyncState) {
out << " opsAppied: " << _initialSyncState->appliedOps;
}
return out;
}
BSONObj InitialSyncer::getInitialSyncProgress() const {
LockGuard lk(_mutex);
// We return an empty BSON object after an initial sync attempt has been successfully
// completed. When an initial sync attempt completes successfully, initialSyncCompletes is
// incremented and then _initialSyncState is cleared. We check that _initialSyncState has been
// cleared because an initial sync attempt can fail even after initialSyncCompletes is
// incremented, and we also check that initialSyncCompletes is positive because an initial sync
// attempt can also fail before _initialSyncState is initialized.
if (!_initialSyncState && initialSyncCompletes.get() > 0) {
return BSONObj();
}
return _getInitialSyncProgress_inlock();
}
void InitialSyncer::_appendInitialSyncProgressMinimal_inlock(BSONObjBuilder* bob) const {
_stats.append(bob);
if (!_initialSyncState) {
return;
}
if (_initialSyncState->allDatabaseCloner) {
const auto allDbClonerStats = _initialSyncState->allDatabaseCloner->getStats();
const auto approxTotalDataSize = allDbClonerStats.dataSize;
bob->appendNumber("approxTotalDataSize", approxTotalDataSize);
long long approxTotalBytesCopied = 0;
for (auto dbClonerStats : allDbClonerStats.databaseStats) {
for (auto collClonerStats : dbClonerStats.collectionStats) {
approxTotalBytesCopied += collClonerStats.approxBytesCopied;
}
}
bob->appendNumber("approxTotalBytesCopied", approxTotalBytesCopied);
if (approxTotalBytesCopied > 0) {
const auto statsObj = bob->asTempObj();
auto totalInitialSyncElapsedMillis =
statsObj.getField("totalInitialSyncElapsedMillis").safeNumberLong();
const auto downloadRate =
(double)totalInitialSyncElapsedMillis / (double)approxTotalBytesCopied;
const auto remainingInitialSyncEstimatedMillis =
downloadRate * (double)(approxTotalDataSize - approxTotalBytesCopied);
bob->appendNumber("remainingInitialSyncEstimatedMillis",
(long long)remainingInitialSyncEstimatedMillis);
}
}
bob->appendNumber("appliedOps", _initialSyncState->appliedOps);
if (!_initialSyncState->beginApplyingTimestamp.isNull()) {
bob->append("initialSyncOplogStart", _initialSyncState->beginApplyingTimestamp);
}
// Only include the beginFetchingTimestamp if it's different from the beginApplyingTimestamp.
if (!_initialSyncState->beginFetchingTimestamp.isNull() &&
_initialSyncState->beginFetchingTimestamp != _initialSyncState->beginApplyingTimestamp) {
bob->append("initialSyncOplogFetchingStart", _initialSyncState->beginFetchingTimestamp);
}
if (!_initialSyncState->stopTimestamp.isNull()) {
bob->append("initialSyncOplogEnd", _initialSyncState->stopTimestamp);
}
if (_sharedData) {
stdx::lock_guard<InitialSyncSharedData> sdLock(*_sharedData);
auto unreachableSince = _sharedData->getSyncSourceUnreachableSince(sdLock);
if (unreachableSince != Date_t()) {
bob->append("syncSourceUnreachableSince", unreachableSince);
bob->append("currentOutageDurationMillis",
durationCount<Milliseconds>(_sharedData->getCurrentOutageDuration(sdLock)));
}
bob->append("totalTimeUnreachableMillis",
durationCount<Milliseconds>(_sharedData->getTotalTimeUnreachable(sdLock)));
}
}
BSONObj InitialSyncer::_getInitialSyncProgress_inlock() const {
try {
BSONObjBuilder bob;
_appendInitialSyncProgressMinimal_inlock(&bob);
if (_initialSyncState) {
if (_initialSyncState->allDatabaseCloner) {
BSONObjBuilder dbsBuilder(bob.subobjStart("databases"));
_initialSyncState->allDatabaseCloner->getStats().append(&dbsBuilder);
dbsBuilder.doneFast();
}
}
return bob.obj();
} catch (const DBException& e) {
LOGV2(21161,
"Error creating initial sync progress object: {error}",
"Error creating initial sync progress object",
"error"_attr = e.toString());
}
BSONObjBuilder bob;
_appendInitialSyncProgressMinimal_inlock(&bob);
return bob.obj();
}
void InitialSyncer::setCreateClientFn_forTest(const CreateClientFn& createClientFn) {
LockGuard lk(_mutex);
_createClientFn = createClientFn;
}
void InitialSyncer::setCreateOplogFetcherFn_forTest(
std::unique_ptr<OplogFetcherFactory> createOplogFetcherFn) {
LockGuard lk(_mutex);
_createOplogFetcherFn = std::move(createOplogFetcherFn);
}
OplogFetcher* InitialSyncer::getOplogFetcher_forTest() const {
// Wait up to 10 seconds.
for (auto i = 0; i < 100; i++) {
{
LockGuard lk(_mutex);
if (_oplogFetcher) {
return _oplogFetcher.get();
}
}
sleepmillis(100);
}
invariant(false, "Timed out getting OplogFetcher pointer for test");
return nullptr;
}
void InitialSyncer::setClonerExecutor_forTest(std::shared_ptr<executor::TaskExecutor> clonerExec) {
_clonerExec = clonerExec;
}
void InitialSyncer::waitForCloner_forTest() {
_initialSyncState->allDatabaseClonerFuture.wait();
}
void InitialSyncer::_setUp_inlock(OperationContext* opCtx, std::uint32_t initialSyncMaxAttempts) {
// 'opCtx' is passed through from startup().
_replicationProcess->getConsistencyMarkers()->setInitialSyncFlag(opCtx);
_replicationProcess->getConsistencyMarkers()->clearInitialSyncId(opCtx);
auto serviceCtx = opCtx->getServiceContext();
_storage->setInitialDataTimestamp(serviceCtx, Timestamp::kAllowUnstableCheckpointsSentinel);
_storage->setStableTimestamp(serviceCtx, Timestamp::min());
LOGV2_DEBUG(21162, 1, "Creating oplogBuffer");
_oplogBuffer = _dataReplicatorExternalState->makeInitialSyncOplogBuffer(opCtx);
_oplogBuffer->startup(opCtx);
_stats.initialSyncStart = _exec->now();
_stats.maxFailedInitialSyncAttempts = initialSyncMaxAttempts;
_stats.failedInitialSyncAttempts = 0;
_allowedOutageDuration = Seconds(initialSyncTransientErrorRetryPeriodSeconds.load());
}
void InitialSyncer::_tearDown_inlock(OperationContext* opCtx,
const StatusWith<OpTimeAndWallTime>& lastApplied) {
_stats.initialSyncEnd = _exec->now();
// This might not be necessary if we failed initial sync.
invariant(_oplogBuffer);
_oplogBuffer->shutdown(opCtx);
if (!lastApplied.isOK()) {
return;
}
const auto lastAppliedOpTime = lastApplied.getValue().opTime;
auto initialDataTimestamp = lastAppliedOpTime.getTimestamp();
// A node coming out of initial sync must guarantee at least one oplog document is visible
// such that others can sync from this node. Oplog visibility is only advanced when applying
// oplog entries during initial sync. Correct the visibility to match the initial sync time
// before transitioning to steady state replication.
const bool orderedCommit = true;
_storage->oplogDiskLocRegister(opCtx, initialDataTimestamp, orderedCommit);
tenant_migration_donor::recoverTenantMigrationAccessBlockers(opCtx);
reconstructPreparedTransactions(opCtx, repl::OplogApplication::Mode::kInitialSync);
_replicationProcess->getConsistencyMarkers()->setInitialSyncIdIfNotSet(opCtx);
_replicationProcess->getConsistencyMarkers()->clearInitialSyncFlag(opCtx);
// All updates that represent initial sync must be completed before setting the initial data
// timestamp.
_storage->setInitialDataTimestamp(opCtx->getServiceContext(), initialDataTimestamp);
auto currentLastAppliedOpTime = _opts.getMyLastOptime();
if (currentLastAppliedOpTime.isNull()) {
_opts.setMyLastOptime(lastApplied.getValue());
} else {
invariant(currentLastAppliedOpTime == lastAppliedOpTime);
}
LOGV2(21163,
"initial sync done; took "
"{duration}.",
"Initial sync done",
"duration"_attr =
duration_cast<Seconds>(_stats.initialSyncEnd - _stats.initialSyncStart));
initialSyncCompletes.increment();
}
void InitialSyncer::_startInitialSyncAttemptCallback(
const executor::TaskExecutor::CallbackArgs& callbackArgs,
std::uint32_t initialSyncAttempt,
std::uint32_t initialSyncMaxAttempts) noexcept {
auto status = [&] {
stdx::lock_guard<Latch> lock(_mutex);
return _checkForShutdownAndConvertStatus_inlock(
callbackArgs,
str::stream() << "error while starting initial sync attempt "
<< (initialSyncAttempt + 1) << " of " << initialSyncMaxAttempts);
}();
if (!status.isOK()) {
_finishInitialSyncAttempt(status);
return;
}
LOGV2(21164,
"Starting initial sync (attempt {initialSyncAttempt} of {initialSyncMaxAttempts})",
"Starting initial sync attempt",
"initialSyncAttempt"_attr = (initialSyncAttempt + 1),
"initialSyncMaxAttempts"_attr = initialSyncMaxAttempts);
// This completion guard invokes _finishInitialSyncAttempt on destruction.
auto cancelRemainingWorkInLock = [this]() { _cancelRemainingWork_inlock(); };
auto finishInitialSyncAttemptFn = [this](const StatusWith<OpTimeAndWallTime>& lastApplied) {
_finishInitialSyncAttempt(lastApplied);
};
auto onCompletionGuard =
std::make_shared<OnCompletionGuard>(cancelRemainingWorkInLock, finishInitialSyncAttemptFn);
// Lock guard must be declared after completion guard because completion guard destructor
// has to run outside lock.
stdx::lock_guard<Latch> lock(_mutex);
_oplogApplier = {};
LOGV2_DEBUG(
21165, 2, "Resetting sync source so a new one can be chosen for this initial sync attempt");
_syncSource = HostAndPort();
LOGV2_DEBUG(21166, 2, "Resetting all optimes before starting this initial sync attempt");
_opts.resetOptimes();
_lastApplied = {OpTime(), Date_t()};
_lastFetched = {};
LOGV2_DEBUG(
21167, 2, "Resetting the oldest timestamp before starting this initial sync attempt");
auto storageEngine = getGlobalServiceContext()->getStorageEngine();
if (storageEngine) {
// Set the oldestTimestamp to one because WiredTiger does not allow us to set it to zero
// since that would also set the all_durable point to zero. We specifically don't set
// the stable timestamp here because that will trigger taking a first stable checkpoint even
// though the initialDataTimestamp is still set to kAllowUnstableCheckpointsSentinel.
storageEngine->setOldestTimestamp(kTimestampOne);
}
LOGV2_DEBUG(21168,
2,
"Resetting feature compatibility version to last-lts. If the sync source is in "
"latest feature compatibility version, we will find out when we clone the "
"server configuration collection (admin.system.version)");
serverGlobalParams.mutableFeatureCompatibility.reset();
// Clear the oplog buffer.
_oplogBuffer->clear(makeOpCtx().get());
// Get sync source.
std::uint32_t chooseSyncSourceAttempt = 0;
std::uint32_t chooseSyncSourceMaxAttempts =
static_cast<std::uint32_t>(numInitialSyncConnectAttempts.load());
// _scheduleWorkAndSaveHandle_inlock() is shutdown-aware.
status = _scheduleWorkAndSaveHandle_inlock(
[=](const executor::TaskExecutor::CallbackArgs& args) {
_chooseSyncSourceCallback(
args, chooseSyncSourceAttempt, chooseSyncSourceMaxAttempts, onCompletionGuard);
},
&_chooseSyncSourceHandle,
str::stream() << "_chooseSyncSourceCallback-" << chooseSyncSourceAttempt);
if (!status.isOK()) {
onCompletionGuard->setResultAndCancelRemainingWork_inlock(lock, status);
return;
}
}
void InitialSyncer::_chooseSyncSourceCallback(
const executor::TaskExecutor::CallbackArgs& callbackArgs,
std::uint32_t chooseSyncSourceAttempt,
std::uint32_t chooseSyncSourceMaxAttempts,
std::shared_ptr<OnCompletionGuard> onCompletionGuard) noexcept try {
stdx::unique_lock<Latch> lock(_mutex);
// Cancellation should be treated the same as other errors. In this case, the most likely cause
// of a failed _chooseSyncSourceCallback() task is a cancellation triggered by
// InitialSyncer::shutdown() or the task executor shutting down.
auto status =
_checkForShutdownAndConvertStatus_inlock(callbackArgs, "error while choosing sync source");
if (!status.isOK()) {
onCompletionGuard->setResultAndCancelRemainingWork_inlock(lock, status);
return;
}
if (MONGO_unlikely(failInitialSyncWithBadHost.shouldFail())) {
status = Status(ErrorCodes::InvalidSyncSource,
"initial sync failed - failInitialSyncWithBadHost failpoint is set.");
onCompletionGuard->setResultAndCancelRemainingWork_inlock(lock, status);
return;
}
auto syncSource = _chooseSyncSource_inlock();
if (!syncSource.isOK()) {
if (chooseSyncSourceAttempt + 1 >= chooseSyncSourceMaxAttempts) {
onCompletionGuard->setResultAndCancelRemainingWork_inlock(
lock,
Status(ErrorCodes::InitialSyncOplogSourceMissing,
"No valid sync source found in current replica set to do an initial sync."));
return;
}
auto when = (*_attemptExec)->now() + _opts.syncSourceRetryWait;
LOGV2_DEBUG(21169,
1,
"Error getting sync source: '{error}', trying again in "
"{syncSourceRetryWait} at {retryTime}. Attempt {chooseSyncSourceAttempt} of "
"{numInitialSyncConnectAttempts}",
"Error getting sync source. Waiting to retry",
"error"_attr = syncSource.getStatus(),
"syncSourceRetryWait"_attr = _opts.syncSourceRetryWait,
"retryTime"_attr = when.toString(),
"chooseSyncSourceAttempt"_attr = (chooseSyncSourceAttempt + 1),
"numInitialSyncConnectAttempts"_attr = numInitialSyncConnectAttempts.load());
auto status = _scheduleWorkAtAndSaveHandle_inlock(
when,
[=](const executor::TaskExecutor::CallbackArgs& args) {
_chooseSyncSourceCallback(args,
chooseSyncSourceAttempt + 1,
chooseSyncSourceMaxAttempts,
onCompletionGuard);
},
&_chooseSyncSourceHandle,
str::stream() << "_chooseSyncSourceCallback-" << (chooseSyncSourceAttempt + 1));
if (!status.isOK()) {
onCompletionGuard->setResultAndCancelRemainingWork_inlock(lock, status);
return;
}
return;
}
if (MONGO_unlikely(initialSyncHangBeforeCreatingOplog.shouldFail())) {
// This log output is used in js tests so please leave it.
LOGV2(21170,
"initial sync - initialSyncHangBeforeCreatingOplog fail point "
"enabled. Blocking until fail point is disabled.");
lock.unlock();
while (MONGO_unlikely(initialSyncHangBeforeCreatingOplog.shouldFail()) &&
!_isShuttingDown()) {
mongo::sleepsecs(1);
}
lock.lock();
}
// There is no need to schedule separate task to create oplog collection since we are already in
// a callback and we are certain there's no existing operation context (required for creating
// collections and dropping user databases) attached to the current thread.
status = _truncateOplogAndDropReplicatedDatabases();
if (!status.isOK()) {
onCompletionGuard->setResultAndCancelRemainingWork_inlock(lock, status);
return;
}
_syncSource = syncSource.getValue();
// Schedule rollback ID checker.
_rollbackChecker = std::make_unique<RollbackChecker>(*_attemptExec, _syncSource);
auto scheduleResult = _rollbackChecker->reset([=](const RollbackChecker::Result& result) {
return _rollbackCheckerResetCallback(result, onCompletionGuard);
});
status = scheduleResult.getStatus();
if (!status.isOK()) {
onCompletionGuard->setResultAndCancelRemainingWork_inlock(lock, status);
return;
}
_getBaseRollbackIdHandle = scheduleResult.getValue();
} catch (const DBException&) {
// Report exception as an initial syncer failure.
stdx::unique_lock<Latch> lock(_mutex);
onCompletionGuard->setResultAndCancelRemainingWork_inlock(lock, exceptionToStatus());
}
Status InitialSyncer::_truncateOplogAndDropReplicatedDatabases() {
// truncate oplog; drop user databases.
LOGV2_DEBUG(21171,
1,
"About to truncate the oplog, if it exists, ns:{namespace}, and drop all "
"user databases (so that we can clone them).",
"About to truncate the oplog, if it exists, and drop all user databases (so that "
"we can clone them)",
"namespace"_attr = _opts.localOplogNS);
auto opCtx = makeOpCtx();
// We are not replicating nor validating these writes.
UnreplicatedWritesBlock unreplicatedWritesBlock(opCtx.get());
// 1.) Truncate the oplog.
LOGV2_DEBUG(21172,
2,
"Truncating the existing oplog: {namespace}",
"Truncating the existing oplog",
"namespace"_attr = _opts.localOplogNS);
Timer timer;
auto status = _storage->truncateCollection(opCtx.get(), _opts.localOplogNS);
LOGV2(21173,
"Initial syncer oplog truncation finished in: {durationMillis}ms",
"Initial syncer oplog truncation finished",
"durationMillis"_attr = timer.millis());
if (!status.isOK()) {
// 1a.) Create the oplog.
LOGV2_DEBUG(21174,
2,
"Creating the oplog: {namespace}",
"Creating the oplog",
"namespace"_attr = _opts.localOplogNS);
status = _storage->createOplog(opCtx.get(), _opts.localOplogNS);
if (!status.isOK()) {
return status;
}
}
// 2a.) Abort any index builds started during initial sync.
IndexBuildsCoordinator::get(opCtx.get())
->abortAllIndexBuildsForInitialSync(opCtx.get(), "Aborting index builds for initial sync");
// 2b.) Drop user databases.
LOGV2_DEBUG(21175, 2, "Dropping user databases");
return _storage->dropReplicatedDatabases(opCtx.get());
}
void InitialSyncer::_rollbackCheckerResetCallback(
const RollbackChecker::Result& result, std::shared_ptr<OnCompletionGuard> onCompletionGuard) {
stdx::lock_guard<Latch> lock(_mutex);
auto status = _checkForShutdownAndConvertStatus_inlock(result.getStatus(),
"error while getting base rollback ID");
if (!status.isOK()) {
onCompletionGuard->setResultAndCancelRemainingWork_inlock(lock, status);
return;
}
// Since the beginFetchingOpTime is retrieved before significant work is done copying
// data from the sync source, we allow the OplogEntryFetcher to use its default retry strategy
// which retries up to 'numInitialSyncOplogFindAttempts' times'. This will fail relatively
// quickly in the presence of network errors, allowing us to choose a different sync source.
status = _scheduleLastOplogEntryFetcher_inlock(
[=](const StatusWith<mongo::Fetcher::QueryResponse>& response,
mongo::Fetcher::NextAction*,
mongo::BSONObjBuilder*) mutable {
_lastOplogEntryFetcherCallbackForDefaultBeginFetchingOpTime(response,
onCompletionGuard);
},
kFetcherHandlesRetries);
if (!status.isOK()) {
onCompletionGuard->setResultAndCancelRemainingWork_inlock(lock, status);
return;
}
}
void InitialSyncer::_lastOplogEntryFetcherCallbackForDefaultBeginFetchingOpTime(
const StatusWith<Fetcher::QueryResponse>& result,
std::shared_ptr<OnCompletionGuard> onCompletionGuard) {
stdx::unique_lock<Latch> lock(_mutex);
auto status = _checkForShutdownAndConvertStatus_inlock(
result.getStatus(), "error while getting last oplog entry for begin timestamp");
if (!status.isOK()) {
onCompletionGuard->setResultAndCancelRemainingWork_inlock(lock, status);
return;
}
const auto opTimeResult = parseOpTimeAndWallTime(result);
status = opTimeResult.getStatus();
if (!status.isOK()) {
onCompletionGuard->setResultAndCancelRemainingWork_inlock(lock, status);
return;
}
// This is the top of the oplog before we query for the oldest active transaction timestamp. If
// that query returns that there are no active transactions, we will use this as the
// beginFetchingTimestamp.
const auto& defaultBeginFetchingOpTime = opTimeResult.getValue().opTime;
std::string logMsg = str::stream() << "Initial Syncer got the defaultBeginFetchingTimestamp: "
<< defaultBeginFetchingOpTime.toString();
pauseAtInitialSyncFuzzerSyncronizationPoints(logMsg);
status = _scheduleGetBeginFetchingOpTime_inlock(onCompletionGuard, defaultBeginFetchingOpTime);
if (!status.isOK()) {
onCompletionGuard->setResultAndCancelRemainingWork_inlock(lock, status);
return;
}
}
Status InitialSyncer::_scheduleGetBeginFetchingOpTime_inlock(
std::shared_ptr<OnCompletionGuard> onCompletionGuard,
const OpTime& defaultBeginFetchingOpTime) {
const auto preparedState = DurableTxnState_serializer(DurableTxnStateEnum::kPrepared);
const auto inProgressState = DurableTxnState_serializer(DurableTxnStateEnum::kInProgress);
// Obtain the oldest active transaction timestamp from the remote by querying their
// transactions table. To prevent oplog holes from causing this query to return an inaccurate
// timestamp, we specify an afterClusterTime of Timestamp(0, 1) so that we wait for all previous
// writes to be visible.
BSONObjBuilder cmd;
cmd.append("find", NamespaceString::kSessionTransactionsTableNamespace.coll().toString());
cmd.append("filter",
BSON("state" << BSON("$in" << BSON_ARRAY(preparedState << inProgressState))));
cmd.append("sort", BSON(SessionTxnRecord::kStartOpTimeFieldName << 1));
cmd.append("readConcern",
BSON("level"
<< "local"
<< "afterClusterTime" << Timestamp(0, 1)));
cmd.append("limit", 1);
_beginFetchingOpTimeFetcher = std::make_unique<Fetcher>(
*_attemptExec,
_syncSource,
NamespaceString::kSessionTransactionsTableNamespace.db().toString(),
cmd.obj(),
[=](const StatusWith<mongo::Fetcher::QueryResponse>& response,
mongo::Fetcher::NextAction*,
mongo::BSONObjBuilder*) mutable {
_getBeginFetchingOpTimeCallback(
response, onCompletionGuard, defaultBeginFetchingOpTime);
},
ReadPreferenceSetting::secondaryPreferredMetadata(),
RemoteCommandRequest::kNoTimeout /* find network timeout */,
RemoteCommandRequest::kNoTimeout /* getMore network timeout */,
RemoteCommandRetryScheduler::makeRetryPolicy<ErrorCategory::RetriableError>(
numInitialSyncOplogFindAttempts.load(), executor::RemoteCommandRequest::kNoTimeout));
Status scheduleStatus = _beginFetchingOpTimeFetcher->schedule();
if (!scheduleStatus.isOK()) {
_beginFetchingOpTimeFetcher.reset();
}
return scheduleStatus;
}
void InitialSyncer::_getBeginFetchingOpTimeCallback(
const StatusWith<Fetcher::QueryResponse>& result,
std::shared_ptr<OnCompletionGuard> onCompletionGuard,
const OpTime& defaultBeginFetchingOpTime) {
stdx::unique_lock<Latch> lock(_mutex);
auto status = _checkForShutdownAndConvertStatus_inlock(
result.getStatus(),
"error while getting oldest active transaction timestamp for begin fetching timestamp");
if (!status.isOK()) {
onCompletionGuard->setResultAndCancelRemainingWork_inlock(lock, status);
return;
}
const auto docs = result.getValue().documents;
if (docs.size() > 1) {
onCompletionGuard->setResultAndCancelRemainingWork_inlock(
lock,
Status(ErrorCodes::TooManyMatchingDocuments,
str::stream() << "Expected to receive one document for the oldest active "
"transaction entry, but received: "
<< docs.size() << ". First: " << redact(docs.front())
<< ". Last: " << redact(docs.back())));
return;
}
// Set beginFetchingOpTime if the oldest active transaction timestamp actually exists. Otherwise
// use the sync source's top of the oplog from before querying for the oldest active transaction
// timestamp. This will mean that even if a transaction is started on the sync source after
// querying for the oldest active transaction timestamp, the node will still fetch its oplog
// entries.
OpTime beginFetchingOpTime = defaultBeginFetchingOpTime;
if (docs.size() != 0) {
auto entry = SessionTxnRecord::parse(
IDLParserErrorContext("oldest active transaction optime for initial sync"),
docs.front());
auto optime = entry.getStartOpTime();
if (optime) {
beginFetchingOpTime = optime.get();
}
}
std::string logMsg = str::stream()
<< "Initial Syncer got the beginFetchingTimestamp: " << beginFetchingOpTime.toString();
pauseAtInitialSyncFuzzerSyncronizationPoints(logMsg);
if (MONGO_unlikely(initialSyncHangAfterGettingBeginFetchingTimestamp.shouldFail())) {
LOGV2(21176, "initialSyncHangAfterGettingBeginFetchingTimestamp fail point enabled");
initialSyncHangAfterGettingBeginFetchingTimestamp.pauseWhileSet();
}
// Since the beginFetchingOpTime is retrieved before significant work is done copying
// data from the sync source, we allow the OplogEntryFetcher to use its default retry strategy
// which retries up to 'numInitialSyncOplogFindAttempts' times'. This will fail relatively
// quickly in the presence of network errors, allowing us to choose a different sync source.
status = _scheduleLastOplogEntryFetcher_inlock(
[=](const StatusWith<mongo::Fetcher::QueryResponse>& response,
mongo::Fetcher::NextAction*,
mongo::BSONObjBuilder*) mutable {
_lastOplogEntryFetcherCallbackForBeginApplyingTimestamp(
response, onCompletionGuard, beginFetchingOpTime);
},
kFetcherHandlesRetries);
if (!status.isOK()) {
onCompletionGuard->setResultAndCancelRemainingWork_inlock(lock, status);
return;
}
}