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
|
/**
* Copyright (C) 2012 10gen Inc.
*
* This program is free software: you can redistribute it and/or modify
* it under the terms of the GNU Affero General Public License, version 3,
* as published by the Free Software Foundation.
*
* 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
* GNU Affero General Public License for more details.
*
* You should have received a copy of the GNU Affero General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*
* 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 GNU Affero General 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_LOG_DEFAULT_COMPONENT ::mongo::logger::LogComponent::kReplication
#include "mongo/platform/basic.h"
#include "mongo/db/repl/bgsync.h"
#include "mongo/base/counter.h"
#include "mongo/base/string_data.h"
#include "mongo/bson/util/bson_extract.h"
#include "mongo/client/connection_pool.h"
#include "mongo/db/auth/authorization_session.h"
#include "mongo/db/client.h"
#include "mongo/db/commands/server_status_metric.h"
#include "mongo/db/concurrency/write_conflict_exception.h"
#include "mongo/db/dbhelpers.h"
#include "mongo/db/repl/data_replicator_external_state_impl.h"
#include "mongo/db/repl/oplog.h"
#include "mongo/db/repl/oplog_interface_local.h"
#include "mongo/db/repl/replication_coordinator_global.h"
#include "mongo/db/repl/replication_coordinator_impl.h"
#include "mongo/db/repl/replication_process.h"
#include "mongo/db/repl/rollback_source_impl.h"
#include "mongo/db/repl/rs_rollback.h"
#include "mongo/db/repl/rs_rollback_no_uuid.h"
#include "mongo/db/repl/rs_sync.h"
#include "mongo/db/repl/storage_interface.h"
#include "mongo/db/s/shard_identity_rollback_notifier.h"
#include "mongo/db/server_parameters.h"
#include "mongo/rpc/get_status_from_command_result.h"
#include "mongo/rpc/metadata/repl_set_metadata.h"
#include "mongo/stdx/memory.h"
#include "mongo/util/log.h"
#include "mongo/util/mongoutils/str.h"
#include "mongo/util/time_support.h"
namespace mongo {
using std::string;
namespace repl {
namespace {
const char kHashFieldName[] = "h";
const int kSleepToAllowBatchingMillis = 2;
const int kSmallBatchLimitBytes = 40000;
const Milliseconds kRollbackOplogSocketTimeout(10 * 60 * 1000);
// 16MB max batch size / 12 byte min doc size * 10 (for good measure) = defaultBatchSize to use.
const auto defaultBatchSize = (16 * 1024 * 1024) / 12 * 10;
// Valid arguments to the rollbackMethod server parameter.
constexpr StringData kRollbackViaRefetchNoUUID = "rollbackViaRefetchNoUUID"_sd;
constexpr StringData kRollbackViaRefetch = "rollbackViaRefetch"_sd;
constexpr StringData kRollbackToCheckpoint = "rollbackToCheckpoint"_sd;
// Set this to force rollback to use a particular implementation.
MONGO_EXPORT_STARTUP_SERVER_PARAMETER(rollbackMethod,
std::string,
kRollbackViaRefetchNoUUID.toString());
// Checks that only the valid strings above can be used as a rollbackMethod
// parameter. Throws an error if an invalid string is passed to the server parameter.
MONGO_INITIALIZER(rollbackMethod)(InitializerContext*) {
if ((rollbackMethod != kRollbackViaRefetchNoUUID) && (rollbackMethod != kRollbackViaRefetch) &&
(rollbackMethod != kRollbackToCheckpoint)) {
return Status(ErrorCodes::BadValue, "Unsupported rollback method: " + rollbackMethod);
}
return Status::OK();
}
// The batchSize to use for the find/getMore queries called by the OplogFetcher
MONGO_EXPORT_STARTUP_SERVER_PARAMETER(bgSyncOplogFetcherBatchSize, int, defaultBatchSize);
/**
* Extends DataReplicatorExternalStateImpl to be member state aware.
*/
class DataReplicatorExternalStateBackgroundSync : public DataReplicatorExternalStateImpl {
public:
DataReplicatorExternalStateBackgroundSync(
ReplicationCoordinator* replicationCoordinator,
ReplicationCoordinatorExternalState* replicationCoordinatorExternalState,
BackgroundSync* bgsync);
bool shouldStopFetching(const HostAndPort& source,
const rpc::ReplSetMetadata& replMetadata,
boost::optional<rpc::OplogQueryMetadata> oqMetadata) override;
private:
BackgroundSync* _bgsync;
};
DataReplicatorExternalStateBackgroundSync::DataReplicatorExternalStateBackgroundSync(
ReplicationCoordinator* replicationCoordinator,
ReplicationCoordinatorExternalState* replicationCoordinatorExternalState,
BackgroundSync* bgsync)
: DataReplicatorExternalStateImpl(replicationCoordinator, replicationCoordinatorExternalState),
_bgsync(bgsync) {}
bool DataReplicatorExternalStateBackgroundSync::shouldStopFetching(
const HostAndPort& source,
const rpc::ReplSetMetadata& replMetadata,
boost::optional<rpc::OplogQueryMetadata> oqMetadata) {
if (_bgsync->shouldStopFetching()) {
return true;
}
return DataReplicatorExternalStateImpl::shouldStopFetching(source, replMetadata, oqMetadata);
}
size_t getSize(const BSONObj& o) {
// SERVER-9808 Avoid Fortify complaint about implicit signed->unsigned conversion
return static_cast<size_t>(o.objsize());
}
} // namespace
// Failpoint which causes rollback to hang before starting.
MONGO_FP_DECLARE(rollbackHangBeforeStart);
// The count of items in the buffer
static Counter64 bufferCountGauge;
static ServerStatusMetricField<Counter64> displayBufferCount("repl.buffer.count",
&bufferCountGauge);
// The size (bytes) of items in the buffer
static Counter64 bufferSizeGauge;
static ServerStatusMetricField<Counter64> displayBufferSize("repl.buffer.sizeBytes",
&bufferSizeGauge);
// The max size (bytes) of the buffer. If the buffer does not have a size constraint, this is
// set to 0.
static Counter64 bufferMaxSizeGauge;
static ServerStatusMetricField<Counter64> displayBufferMaxSize("repl.buffer.maxSizeBytes",
&bufferMaxSizeGauge);
BackgroundSync::BackgroundSync(
ReplicationCoordinatorExternalState* replicationCoordinatorExternalState,
ReplicationProcess* replicationProcess,
std::unique_ptr<OplogBuffer> oplogBuffer)
: _oplogBuffer(std::move(oplogBuffer)),
_replCoord(getGlobalReplicationCoordinator()),
_replicationCoordinatorExternalState(replicationCoordinatorExternalState),
_replicationProcess(replicationProcess) {
// Update "repl.buffer.maxSizeBytes" server status metric to reflect the current oplog buffer's
// max size.
bufferMaxSizeGauge.increment(_oplogBuffer->getMaxSize() - bufferMaxSizeGauge.get());
}
void BackgroundSync::startup(OperationContext* opCtx) {
_oplogBuffer->startup(opCtx);
invariant(!_producerThread);
_producerThread.reset(new stdx::thread(stdx::bind(&BackgroundSync::_run, this)));
}
void BackgroundSync::shutdown(OperationContext* opCtx) {
stdx::lock_guard<stdx::mutex> lock(_mutex);
// Clear the buffer. This unblocks the OplogFetcher if it is blocked with a full queue, but
// ensures that it won't add anything. It will also unblock the OpApplier pipeline if it is
// waiting for an operation to be past the slaveDelay point.
clearBuffer(opCtx);
_state = ProducerState::Stopped;
if (_syncSourceResolver) {
_syncSourceResolver->shutdown();
}
if (_oplogFetcher) {
_oplogFetcher->shutdown();
}
if (_rollback) {
_rollback->shutdown();
}
_inShutdown = true;
}
void BackgroundSync::join(OperationContext* opCtx) {
_producerThread->join();
_oplogBuffer->shutdown(opCtx);
}
bool BackgroundSync::inShutdown() const {
stdx::lock_guard<stdx::mutex> lock(_mutex);
return _inShutdown_inlock();
}
bool BackgroundSync::_inShutdown_inlock() const {
return _inShutdown;
}
void BackgroundSync::_run() {
Client::initThread("rsBackgroundSync");
AuthorizationSession::get(cc())->grantInternalAuthorization();
while (!inShutdown()) {
try {
_runProducer();
} catch (const DBException& e) {
std::string msg(str::stream() << "sync producer problem: " << redact(e));
error() << msg;
_replCoord->setMyHeartbeatMessage(msg);
sleepmillis(100); // sleep a bit to keep from hammering this thread with temp. errors.
} catch (const std::exception& e2) {
// redact(std::exception&) doesn't work
severe() << "sync producer exception: " << redact(e2.what());
fassertFailed(28546);
}
}
stop(true);
}
void BackgroundSync::_runProducer() {
if (getState() == ProducerState::Stopped) {
sleepsecs(1);
return;
}
// TODO(spencer): Use a condition variable to await loading a config.
// TODO(siyuan): Control bgsync with producer state.
if (_replCoord->getMemberState().startup()) {
// Wait for a config to be loaded
sleepsecs(1);
return;
}
invariant(!_replCoord->getMemberState().rollback());
// We need to wait until initial sync has started.
if (_replCoord->getMyLastAppliedOpTime().isNull()) {
sleepsecs(1);
return;
}
// we want to start when we're no longer primary
// start() also loads _lastOpTimeFetched, which we know is set from the "if"
auto opCtx = cc().makeOperationContext();
if (getState() == ProducerState::Starting) {
start(opCtx.get());
}
_produce(opCtx.get());
}
void BackgroundSync::_produce(OperationContext* opCtx) {
if (MONGO_FAIL_POINT(stopReplProducer)) {
// This log output is used in js tests so please leave it.
log() << "bgsync - stopReplProducer fail point "
"enabled. Blocking until fail point is disabled.";
// TODO(SERVER-27120): Remove the return statement and uncomment the while loop.
// Currently we cannot block here or we prevent primaries from being fully elected since
// we'll never call _signalNoNewDataForApplier.
// while (MONGO_FAIL_POINT(stopReplProducer) && !inShutdown()) {
// mongo::sleepsecs(1);
// }
mongo::sleepsecs(1);
return;
}
// this oplog reader does not do a handshake because we don't want the server it's syncing
// from to track how far it has synced
{
stdx::unique_lock<stdx::mutex> lock(_mutex);
if (_lastOpTimeFetched.isNull()) {
// then we're initial syncing and we're still waiting for this to be set
lock.unlock();
sleepsecs(1);
// if there is no one to sync from
return;
}
if (_state != ProducerState::Running) {
return;
}
}
// find a target to sync from the last optime fetched
OpTime lastOpTimeFetched;
HostAndPort source;
HostAndPort oldSource = _syncSourceHost;
SyncSourceResolverResponse syncSourceResp;
{
const OpTime minValidSaved =
_replicationProcess->getConsistencyMarkers()->getMinValid(opCtx);
stdx::lock_guard<stdx::mutex> lock(_mutex);
if (_state != ProducerState::Running) {
return;
}
const auto requiredOpTime = (minValidSaved > _lastOpTimeFetched) ? minValidSaved : OpTime();
lastOpTimeFetched = _lastOpTimeFetched;
_syncSourceHost = HostAndPort();
_syncSourceResolver = stdx::make_unique<SyncSourceResolver>(
_replicationCoordinatorExternalState->getTaskExecutor(),
_replCoord,
lastOpTimeFetched,
requiredOpTime,
[&syncSourceResp](const SyncSourceResolverResponse& resp) { syncSourceResp = resp; });
}
// This may deadlock if called inside the mutex because SyncSourceResolver::startup() calls
// ReplicationCoordinator::chooseNewSyncSource(). ReplicationCoordinatorImpl's mutex has to
// acquired before BackgroundSync's.
// It is safe to call startup() outside the mutex on this instance of SyncSourceResolver because
// we do not destroy this instance outside of this function which is only called from a single
// thread.
auto status = _syncSourceResolver->startup();
if (ErrorCodes::CallbackCanceled == status || ErrorCodes::isShutdownError(status.code())) {
return;
}
fassertStatusOK(40349, status);
_syncSourceResolver->join();
{
stdx::lock_guard<stdx::mutex> lock(_mutex);
_syncSourceResolver.reset();
}
if (syncSourceResp.syncSourceStatus == ErrorCodes::OplogStartMissing) {
// All (accessible) sync sources were too stale.
if (_replCoord->getMemberState().primary()) {
warning() << "Too stale to catch up.";
log() << "Our newest OpTime : " << lastOpTimeFetched;
log() << "Earliest OpTime available is " << syncSourceResp.earliestOpTimeSeen
<< " from " << syncSourceResp.getSyncSource();
_replCoord->abortCatchupIfNeeded().transitional_ignore();
return;
}
// We only need to mark ourselves as too stale once.
if (_tooStale) {
return;
}
// Mark yourself as too stale.
_tooStale = true;
error() << "too stale to catch up -- entering maintenance mode";
log() << "Our newest OpTime : " << lastOpTimeFetched;
log() << "Earliest OpTime available is " << syncSourceResp.earliestOpTimeSeen;
log() << "See http://dochub.mongodb.org/core/resyncingaverystalereplicasetmember";
// Activate maintenance mode and transition to RECOVERING.
auto status = _replCoord->setMaintenanceMode(true);
if (!status.isOK()) {
warning() << "Failed to transition into maintenance mode: " << status;
}
status = _replCoord->setFollowerMode(MemberState::RS_RECOVERING);
if (!status.isOK()) {
warning() << "Failed to transition into " << MemberState(MemberState::RS_RECOVERING)
<< ". Current state: " << _replCoord->getMemberState() << causedBy(status);
}
return;
} else if (syncSourceResp.isOK() && !syncSourceResp.getSyncSource().empty()) {
{
stdx::lock_guard<stdx::mutex> lock(_mutex);
_syncSourceHost = syncSourceResp.getSyncSource();
source = _syncSourceHost;
}
// If our sync source has not changed, it is likely caused by our heartbeat data map being
// out of date. In that case we sleep for 1 second to reduce the amount we spin waiting
// for our map to update.
if (oldSource == source) {
log() << "Chose same sync source candidate as last time, " << source
<< ". Sleeping for 1 second to avoid immediately choosing a new sync source for "
"the same reason as last time.";
sleepsecs(1);
}
} else {
if (!syncSourceResp.isOK()) {
log() << "failed to find sync source, received error "
<< syncSourceResp.syncSourceStatus.getStatus();
}
// No sync source found.
sleepsecs(1);
return;
}
// If we find a good sync source after having gone too stale, disable maintenance mode so we can
// transition to SECONDARY.
if (_tooStale) {
_tooStale = false;
log() << "No longer too stale. Able to sync from " << _syncSourceHost;
auto status = _replCoord->setMaintenanceMode(false);
if (!status.isOK()) {
warning() << "Failed to leave maintenance mode: " << status;
}
}
long long lastHashFetched;
{
stdx::lock_guard<stdx::mutex> lock(_mutex);
if (_state != ProducerState::Running) {
return;
}
lastOpTimeFetched = _lastOpTimeFetched;
lastHashFetched = _lastFetchedHash;
}
if (!_replCoord->getMemberState().primary()) {
_replCoord->signalUpstreamUpdater();
}
// Set the applied point if unset. This is most likely the first time we've established a sync
// source since stepping down or otherwise clearing the applied point. We need to set this here,
// before the OplogWriter gets a chance to append to the oplog.
if (_replicationProcess->getConsistencyMarkers()->getAppliedThrough(opCtx).isNull()) {
_replicationProcess->getConsistencyMarkers()->setAppliedThrough(
opCtx, _replCoord->getMyLastAppliedOpTime());
}
// "lastFetched" not used. Already set in _enqueueDocuments.
Status fetcherReturnStatus = Status::OK();
DataReplicatorExternalStateBackgroundSync dataReplicatorExternalState(
_replCoord, _replicationCoordinatorExternalState, this);
OplogFetcher* oplogFetcher;
try {
auto onOplogFetcherShutdownCallbackFn = [&fetcherReturnStatus](const Status& status) {
fetcherReturnStatus = status;
};
// The construction of OplogFetcher has to be outside bgsync mutex, because it calls
// replication coordinator.
auto oplogFetcherPtr = stdx::make_unique<OplogFetcher>(
_replicationCoordinatorExternalState->getTaskExecutor(),
OpTimeWithHash(lastHashFetched, lastOpTimeFetched),
source,
NamespaceString::kRsOplogNamespace,
_replCoord->getConfig(),
_replicationCoordinatorExternalState->getOplogFetcherMaxFetcherRestarts(),
syncSourceResp.rbid,
true /* requireFresherSyncSource */,
&dataReplicatorExternalState,
stdx::bind(&BackgroundSync::_enqueueDocuments,
this,
stdx::placeholders::_1,
stdx::placeholders::_2,
stdx::placeholders::_3),
onOplogFetcherShutdownCallbackFn,
bgSyncOplogFetcherBatchSize);
stdx::lock_guard<stdx::mutex> lock(_mutex);
if (_state != ProducerState::Running) {
return;
}
_oplogFetcher = std::move(oplogFetcherPtr);
oplogFetcher = _oplogFetcher.get();
} catch (const mongo::DBException& ex) {
fassertFailedWithStatus(34440, exceptionToStatus());
}
const auto logLevel = Command::testCommandsEnabled ? 0 : 1;
LOG(logLevel) << "scheduling fetcher to read remote oplog on " << _syncSourceHost
<< " starting at " << oplogFetcher->getFindQuery_forTest()["filter"];
auto scheduleStatus = oplogFetcher->startup();
if (!scheduleStatus.isOK()) {
warning() << "unable to schedule fetcher to read remote oplog on " << source << ": "
<< scheduleStatus;
return;
}
oplogFetcher->join();
LOG(1) << "fetcher stopped reading remote oplog on " << source;
// If the background sync is stopped after the fetcher is started, we need to
// re-evaluate our sync source and oplog common point.
if (getState() != ProducerState::Running) {
log() << "Replication producer stopped after oplog fetcher finished returning a batch from "
"our sync source. Abandoning this batch of oplog entries and re-evaluating our "
"sync source.";
return;
}
if (fetcherReturnStatus.code() == ErrorCodes::OplogOutOfOrder) {
// This is bad because it means that our source
// has not returned oplog entries in ascending ts order, and they need to be.
warning() << redact(fetcherReturnStatus);
// Do not blacklist the server here, it will be blacklisted when we try to reuse it,
// if it can't return a matching oplog start from the last fetch oplog ts field.
return;
} else if (fetcherReturnStatus.code() == ErrorCodes::OplogStartMissing) {
auto storageInterface = StorageInterface::get(opCtx);
_runRollback(opCtx, fetcherReturnStatus, source, syncSourceResp.rbid, storageInterface);
} else if (fetcherReturnStatus == ErrorCodes::InvalidBSON) {
Seconds blacklistDuration(60);
warning() << "Fetcher got invalid BSON while querying oplog. Blacklisting sync source "
<< source << " for " << blacklistDuration << ".";
_replCoord->blacklistSyncSource(source, Date_t::now() + blacklistDuration);
} else if (!fetcherReturnStatus.isOK()) {
warning() << "Fetcher stopped querying remote oplog with error: "
<< redact(fetcherReturnStatus);
}
}
Status BackgroundSync::_enqueueDocuments(Fetcher::Documents::const_iterator begin,
Fetcher::Documents::const_iterator end,
const OplogFetcher::DocumentsInfo& info) {
// If this is the first batch of operations returned from the query, "toApplyDocumentCount" will
// be one fewer than "networkDocumentCount" because the first document (which was applied
// previously) is skipped.
if (info.toApplyDocumentCount == 0) {
return Status::OK(); // Nothing to do.
}
auto opCtx = cc().makeOperationContext();
// Wait for enough space.
_oplogBuffer->waitForSpace(opCtx.get(), info.toApplyDocumentBytes);
{
// Don't add more to the buffer if we are in shutdown. Continue holding the lock until we
// are done to prevent going into shutdown. This avoids a race where shutdown() clears the
// buffer between the time we check _inShutdown and the point where we finish writing to the
// buffer.
stdx::unique_lock<stdx::mutex> lock(_mutex);
if (_state != ProducerState::Running) {
return Status::OK();
}
OCCASIONALLY {
LOG(2) << "bgsync buffer has " << _oplogBuffer->getSize() << " bytes";
}
// Buffer docs for later application.
_oplogBuffer->pushAllNonBlocking(opCtx.get(), begin, end);
// Update last fetched info.
_lastFetchedHash = info.lastDocument.value;
_lastOpTimeFetched = info.lastDocument.opTime;
LOG(3) << "batch resetting _lastOpTimeFetched: " << _lastOpTimeFetched;
}
bufferCountGauge.increment(info.toApplyDocumentCount);
bufferSizeGauge.increment(info.toApplyDocumentBytes);
// Check some things periodically (whenever we run out of items in the current cursor batch).
if (info.networkDocumentBytes > 0 && info.networkDocumentBytes < kSmallBatchLimitBytes) {
// On a very low latency network, if we don't wait a little, we'll be
// getting ops to write almost one at a time. This will both be expensive
// for the upstream server as well as potentially defeating our parallel
// application of batches on the secondary.
//
// The inference here is basically if the batch is really small, we are "caught up".
sleepmillis(kSleepToAllowBatchingMillis);
}
return Status::OK();
}
bool BackgroundSync::peek(OperationContext* opCtx, BSONObj* op) {
return _oplogBuffer->peek(opCtx, op);
}
void BackgroundSync::waitForMore() {
// Block for one second before timing out.
_oplogBuffer->waitForData(Seconds(1));
}
void BackgroundSync::consume(OperationContext* opCtx) {
// this is just to get the op off the queue, it's been peeked at
// and queued for application already
BSONObj op;
if (_oplogBuffer->tryPop(opCtx, &op)) {
bufferCountGauge.decrement(1);
bufferSizeGauge.decrement(getSize(op));
} else {
invariant(inShutdown());
// This means that shutdown() was called between the consumer's calls to peek() and
// consume(). shutdown() cleared the buffer so there is nothing for us to consume here.
// Since our postcondition is already met, it is safe to return successfully.
}
}
void BackgroundSync::_runRollback(OperationContext* opCtx,
const Status& fetcherReturnStatus,
const HostAndPort& source,
int requiredRBID,
StorageInterface* storageInterface) {
if (_replCoord->getMemberState().primary()) {
warning() << "Rollback situation detected in catch-up mode. Aborting catch-up mode.";
_replCoord->abortCatchupIfNeeded().transitional_ignore();
return;
}
// Rollback is a synchronous operation that uses the task executor and may not be
// executed inside the fetcher callback.
OpTime lastOpTimeFetched;
{
stdx::lock_guard<stdx::mutex> lock(_mutex);
lastOpTimeFetched = _lastOpTimeFetched;
}
log() << "Starting rollback due to " << redact(fetcherReturnStatus);
// TODO: change this to call into the Applier directly to block until the applier is
// drained.
//
// Wait till all buffered oplog entries have drained and been applied.
auto lastApplied = _replCoord->getMyLastAppliedOpTime();
if (lastApplied != lastOpTimeFetched) {
log() << "Waiting for all operations from " << lastApplied << " until " << lastOpTimeFetched
<< " to be applied before starting rollback.";
while (lastOpTimeFetched > (lastApplied = _replCoord->getMyLastAppliedOpTime())) {
sleepmillis(10);
if (getState() != ProducerState::Running) {
return;
}
}
}
if (MONGO_FAIL_POINT(rollbackHangBeforeStart)) {
// This log output is used in js tests so please leave it.
log() << "rollback - rollbackHangBeforeStart fail point "
"enabled. Blocking until fail point is disabled.";
while (MONGO_FAIL_POINT(rollbackHangBeforeStart) && !inShutdown()) {
mongo::sleepsecs(1);
}
}
OplogInterfaceLocal localOplog(opCtx, NamespaceString::kRsOplogNamespace.ns());
// TODO: We will have to check what storage engine is being used eventually, since
// "rollback to a checkpoint" only works on certain storage engines that have this
// functionality enabled.
bool checkpointEnabled = false;
// If the featureCompatibilityVersion (FCV) of the server is 3.4, but the user
// attempts to set the server parameter to "rollback via refetch" with UUIDs or
// "rollback to a checkpoint," we default to using "rollback via refetch" without UUIDs,
// since in FCV 3.4, oplogs with UUIDs and storage engines that support "rollback to a
// checkpoint" are not supported.
if (rollbackMethod == kRollbackViaRefetchNoUUID ||
(serverGlobalParams.featureCompatibility.version.load() ==
ServerGlobalParams::FeatureCompatibility::Version::k34)) {
if ((rollbackMethod == kRollbackViaRefetch) || (rollbackMethod == kRollbackToCheckpoint)) {
log() << "Rollback falling back onto \"rollback via refetch\" without UUID support. "
<< rollbackMethod
<< " is not feature compatible with featureCompatabilityVersion 3.4.";
} else {
log()
<< "Rollback falling back on \"rollback via refetch\" without UUID support due to "
"startup server parameter.";
}
_fallBackOnRollbackViaRefetch(opCtx, source, requiredRBID, &localOplog, false);
} else if (!checkpointEnabled || rollbackMethod == kRollbackViaRefetch) {
// In this block, we are guaranteed to be in FCV 3.6. If we are not running on a storage
// engine that supports "rollback to a checkpoint," if the rollback server parameter was set
// to "rollback to a checkpoint," we default to "rollback via refetch" instead.
if (rollbackMethod == kRollbackToCheckpoint) {
log() << "Rollback falling back on \"rollback via refetch\" because this storage "
"engine does not support \"rollback to a checkpoint\".";
} else {
log() << "Rollback falling back on \"rollback via refetch\" due to startup "
"server parameter.";
}
_fallBackOnRollbackViaRefetch(opCtx, source, requiredRBID, &localOplog, true);
} else {
// If we are running on a storage engine that does support "rollback to a
// checkpoint" and are in FCV 3.6, then we default to the "rollback to a
// checkpoint" algorithm.
AbstractAsyncComponent* rollback;
StatusWith<OpTime> onRollbackShutdownResult =
Status(ErrorCodes::InternalError, "Rollback failed but didn’t return an error message");
try {
auto executor = _replicationCoordinatorExternalState->getTaskExecutor();
auto onRollbackShutdownCallbackFn = [&onRollbackShutdownResult](
const StatusWith<OpTime>& lastApplied) noexcept {
onRollbackShutdownResult = lastApplied;
};
stdx::lock_guard<stdx::mutex> lock(_mutex);
if (_state != ProducerState::Running) {
return;
}
_rollback = stdx::make_unique<RollbackImpl>(
executor,
&localOplog,
source,
NamespaceString::kRsOplogNamespace,
_replicationCoordinatorExternalState->getOplogFetcherMaxFetcherRestarts(),
requiredRBID,
_replCoord,
storageInterface,
onRollbackShutdownCallbackFn);
rollback = _rollback.get();
} catch (...) {
fassertFailedWithStatus(40401, exceptionToStatus());
}
log() << "Scheduling rollback (sync source: " << source << ")";
auto scheduleStatus = rollback->startup();
if (!scheduleStatus.isOK()) {
warning() << "Unable to schedule rollback: " << scheduleStatus;
} else {
rollback->join();
auto status = onRollbackShutdownResult.getStatus();
if (status.isOK()) {
log() << "Rollback successful. Last applied optime: "
<< onRollbackShutdownResult.getValue();
} else if (ErrorCodes::IncompatibleRollbackAlgorithm == status) {
log() << "Rollback falling back on \"rollback via refetch\" due to " << status;
_fallBackOnRollbackViaRefetch(opCtx, source, requiredRBID, &localOplog, true);
} else {
warning() << "Rollback failed with error: " << status;
}
}
}
// Reset the producer to clear the sync source and the last optime fetched.
stop(true);
startProducerIfStopped();
}
void BackgroundSync::_fallBackOnRollbackViaRefetch(OperationContext* opCtx,
const HostAndPort& source,
int requiredRBID,
OplogInterface* localOplog,
bool useUUID) {
const int messagingPortTags = 0;
ConnectionPool connectionPool(messagingPortTags);
std::unique_ptr<ConnectionPool::ConnectionPtr> connection;
auto getConnection = [&connection, &connectionPool, source]() -> DBClientBase* {
if (!connection.get()) {
connection.reset(new ConnectionPool::ConnectionPtr(
&connectionPool, source, Date_t::now(), kRollbackOplogSocketTimeout));
};
return connection->get();
};
RollbackSourceImpl rollbackSource(
getConnection, source, NamespaceString::kRsOplogNamespace.ns());
if (useUUID) {
rollback(opCtx, *localOplog, rollbackSource, requiredRBID, _replCoord, _replicationProcess);
} else {
rollbackNoUUID(
opCtx, *localOplog, rollbackSource, requiredRBID, _replCoord, _replicationProcess);
}
}
HostAndPort BackgroundSync::getSyncTarget() const {
stdx::unique_lock<stdx::mutex> lock(_mutex);
return _syncSourceHost;
}
void BackgroundSync::clearSyncTarget() {
stdx::unique_lock<stdx::mutex> lock(_mutex);
_syncSourceHost = HostAndPort();
}
void BackgroundSync::stop(bool resetLastFetchedOptime) {
stdx::lock_guard<stdx::mutex> lock(_mutex);
_state = ProducerState::Stopped;
_syncSourceHost = HostAndPort();
if (resetLastFetchedOptime) {
invariant(_oplogBuffer->isEmpty());
_lastOpTimeFetched = OpTime();
_lastFetchedHash = 0;
}
if (_syncSourceResolver) {
_syncSourceResolver->shutdown();
}
if (_oplogFetcher) {
_oplogFetcher->shutdown();
}
}
void BackgroundSync::start(OperationContext* opCtx) {
OpTimeWithHash lastAppliedOpTimeWithHash;
do {
lastAppliedOpTimeWithHash = _readLastAppliedOpTimeWithHash(opCtx);
stdx::lock_guard<stdx::mutex> lk(_mutex);
// Double check the state after acquiring the mutex.
if (_state != ProducerState::Starting) {
return;
}
// If a node steps down during drain mode, then the buffer may not be empty at the beginning
// of secondary state.
if (!_oplogBuffer->isEmpty()) {
log() << "going to start syncing, but buffer is not empty";
}
_state = ProducerState::Running;
// When a node steps down during drain mode, the last fetched optime would be newer than
// the last applied.
if (_lastOpTimeFetched <= lastAppliedOpTimeWithHash.opTime) {
_lastOpTimeFetched = lastAppliedOpTimeWithHash.opTime;
_lastFetchedHash = lastAppliedOpTimeWithHash.value;
}
// Reload the last applied optime from disk if it has been changed.
} while (lastAppliedOpTimeWithHash.opTime != _replCoord->getMyLastAppliedOpTime());
LOG(1) << "bgsync fetch queue set to: " << _lastOpTimeFetched << " " << _lastFetchedHash;
}
void BackgroundSync::clearBuffer(OperationContext* opCtx) {
_oplogBuffer->clear(opCtx);
const auto count = bufferCountGauge.get();
bufferCountGauge.decrement(count);
const auto size = bufferSizeGauge.get();
bufferSizeGauge.decrement(size);
}
OpTimeWithHash BackgroundSync::_readLastAppliedOpTimeWithHash(OperationContext* opCtx) {
BSONObj oplogEntry;
try {
bool success = writeConflictRetry(
opCtx, "readLastAppliedHash", NamespaceString::kRsOplogNamespace.ns(), [&] {
Lock::DBLock lk(opCtx, "local", MODE_X);
return Helpers::getLast(
opCtx, NamespaceString::kRsOplogNamespace.ns().c_str(), oplogEntry);
});
if (!success) {
// This can happen when we are to do an initial sync. lastHash will be set
// after the initial sync is complete.
return OpTimeWithHash(0);
}
} catch (const DBException& ex) {
severe() << "Problem reading " << NamespaceString::kRsOplogNamespace.ns() << ": "
<< redact(ex);
fassertFailed(18904);
}
long long hash;
auto status = bsonExtractIntegerField(oplogEntry, kHashFieldName, &hash);
if (!status.isOK()) {
severe() << "Most recent entry in " << NamespaceString::kRsOplogNamespace.ns()
<< " is missing or has invalid \"" << kHashFieldName
<< "\" field. Oplog entry: " << redact(oplogEntry) << ": " << redact(status);
fassertFailed(18902);
}
OplogEntry parsedEntry(oplogEntry);
return OpTimeWithHash(hash, parsedEntry.getOpTime());
}
bool BackgroundSync::shouldStopFetching() const {
// Check if we have been stopped.
if (getState() != ProducerState::Running) {
LOG(2) << "Stopping oplog fetcher due to stop request.";
return true;
}
// Check current sync source.
if (getSyncTarget().empty()) {
LOG(1) << "Stopping oplog fetcher; canceling oplog query because we have no valid sync "
"source.";
return true;
}
return false;
}
void BackgroundSync::pushTestOpToBuffer(OperationContext* opCtx, const BSONObj& op) {
_oplogBuffer->push(opCtx, op);
bufferCountGauge.increment();
bufferSizeGauge.increment(op.objsize());
}
BackgroundSync::ProducerState BackgroundSync::getState() const {
stdx::lock_guard<stdx::mutex> lock(_mutex);
return _state;
}
void BackgroundSync::startProducerIfStopped() {
stdx::lock_guard<stdx::mutex> lock(_mutex);
// Let producer run if it's already running.
if (_state == ProducerState::Stopped) {
_state = ProducerState::Starting;
}
}
} // namespace repl
} // namespace mongo
|