/
balancer.cpp
1292 lines (1102 loc) · 52.8 KB
/
balancer.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.
*/
#include "mongo/db/s/balancer/balancer.h"
#include <absl/container/node_hash_set.h>
#include <boost/cstdint.hpp>
#include <boost/move/utility_core.hpp>
#include <boost/none.hpp>
#include <boost/none_t.hpp>
#include <boost/optional/optional.hpp>
#include <boost/preprocessor/control/iif.hpp>
#include <boost/smart_ptr.hpp>
// IWYU pragma: no_include "cxxabi.h"
#include <algorithm>
#include <cmath>
#include <iterator>
#include <map>
#include <memory>
#include <mutex>
#include <ratio>
#include <string>
#include <tuple>
#include <type_traits>
#include <utility>
#include <variant>
#include <vector>
#include "mongo/base/error_codes.h"
#include "mongo/base/status_with.h"
#include "mongo/base/string_data.h"
#include "mongo/bson/bsonelement.h"
#include "mongo/bson/bsonmisc.h"
#include "mongo/bson/bsonobj.h"
#include "mongo/bson/bsonobjbuilder.h"
#include "mongo/client/read_preference.h"
#include "mongo/db/client.h"
#include "mongo/db/database_name.h"
#include "mongo/db/namespace_string.h"
#include "mongo/db/operation_context.h"
#include "mongo/db/repl/read_concern_level.h"
#include "mongo/db/repl/repl_client_info.h"
#include "mongo/db/repl/replication_coordinator.h"
#include "mongo/db/s/balancer/actions_stream_policy.h"
#include "mongo/db/s/balancer/auto_merger_policy.h"
#include "mongo/db/s/balancer/balancer_commands_scheduler.h"
#include "mongo/db/s/balancer/balancer_commands_scheduler_impl.h"
#include "mongo/db/s/balancer/balancer_defragmentation_policy.h"
#include "mongo/db/s/balancer/cluster_statistics_impl.h"
#include "mongo/db/s/config/sharding_catalog_manager.h"
#include "mongo/db/s/sharding_config_server_parameters_gen.h"
#include "mongo/db/s/sharding_logging.h"
#include "mongo/db/server_feature_flags_gen.h"
#include "mongo/db/shard_id.h"
#include "mongo/db/write_concern_options.h"
#include "mongo/executor/scoped_task_executor.h"
#include "mongo/executor/task_executor_pool.h"
#include "mongo/logv2/log.h"
#include "mongo/logv2/log_attr.h"
#include "mongo/logv2/log_component.h"
#include "mongo/logv2/redaction.h"
#include "mongo/platform/random.h"
#include "mongo/s/balancer_configuration.h"
#include "mongo/s/catalog/sharding_catalog_client.h"
#include "mongo/s/catalog/type_chunk.h"
#include "mongo/s/catalog/type_collection.h"
#include "mongo/s/catalog/type_shard.h"
#include "mongo/s/catalog_cache.h"
#include "mongo/s/chunk.h"
#include "mongo/s/chunk_manager.h"
#include "mongo/s/chunk_version.h"
#include "mongo/s/client/shard.h"
#include "mongo/s/client/shard_registry.h"
#include "mongo/s/grid.h"
#include "mongo/s/request_types/balancer_collection_status_gen.h"
#include "mongo/s/request_types/migration_secondary_throttle_options.h"
#include "mongo/s/shard_util.h"
#include "mongo/stdx/variant.h"
#include "mongo/util/assert_util.h"
#include "mongo/util/concurrency/idle_thread_block.h"
#include "mongo/util/debug_util.h"
#include "mongo/util/decorable.h"
#include "mongo/util/fail_point.h"
#include "mongo/util/future.h"
#include "mongo/util/future_impl.h"
#include "mongo/util/pcre.h"
#include "mongo/util/scopeguard.h"
#include "mongo/util/time_support.h"
#include "mongo/util/timer.h"
#include "mongo/util/version.h"
#define MONGO_LOGV2_DEFAULT_COMPONENT ::mongo::logv2::LogComponent::kSharding
namespace mongo {
using std::map;
using std::string;
using std::vector;
namespace {
MONGO_FAIL_POINT_DEFINE(overrideBalanceRoundInterval);
const Milliseconds kBalanceRoundDefaultInterval(10 * 1000);
/**
* Balancer status response
*/
static constexpr StringData kBalancerPolicyStatusDraining = "draining"_sd;
static constexpr StringData kBalancerPolicyStatusZoneViolation = "zoneViolation"_sd;
static constexpr StringData kBalancerPolicyStatusChunksImbalance = "chunksImbalance"_sd;
static constexpr StringData kBalancerPolicyStatusDefragmentingChunks = "defragmentingChunks"_sd;
// Time interval between checks on draining shards.
constexpr Minutes kDrainingShardsCheckInterval{10};
/**
* Utility class to generate timing and statistics for a single balancer round.
*/
class BalanceRoundDetails {
public:
BalanceRoundDetails() : _executionTimer() {}
void setSucceeded(int numCandidateChunks,
int numChunksMoved,
int numImbalancedCachedCollections) {
invariant(!_errMsg);
_numCandidateChunks = numCandidateChunks;
_numChunksMoved = numChunksMoved;
_numImbalancedCachedCollections = numImbalancedCachedCollections;
}
void setFailed(const string& errMsg) {
_errMsg = errMsg;
}
BSONObj toBSON() const {
BSONObjBuilder builder;
builder.append("executionTimeMillis", _executionTimer.millis());
builder.append("errorOccurred", _errMsg.has_value());
if (_errMsg) {
builder.append("errmsg", *_errMsg);
} else {
builder.append("candidateChunks", _numCandidateChunks);
builder.append("chunksMoved", _numChunksMoved);
builder.append("imbalancedCachedCollections", _numImbalancedCachedCollections);
}
return builder.obj();
}
private:
const Timer _executionTimer;
// Set only on success
int _numCandidateChunks{0};
int _numChunksMoved{0};
int _numImbalancedCachedCollections{0};
// Set only on failure
boost::optional<string> _errMsg;
};
/**
* Occasionally prints a log message with shard versions if the versions are not the same
* in the cluster.
*/
void warnOnMultiVersion(const vector<ClusterStatistics::ShardStatistics>& clusterStats) {
static const auto& majorMinorRE = *new pcre::Regex(R"re(^(\d+)\.(\d+)\.)re");
auto&& vii = VersionInfoInterface::instance();
auto hasMyVersion = [&](auto&& stat) {
auto m = majorMinorRE.match(stat.mongoVersion);
return m && std::stoi(std::string{m[1]}) == vii.majorVersion() &&
std::stoi(std::string{m[2]}) == vii.minorVersion();
};
// If we're all the same version, don't message
if (std::all_of(clusterStats.begin(), clusterStats.end(), hasMyVersion))
return;
BSONObjBuilder shardVersions;
for (const auto& stat : clusterStats) {
shardVersions << stat.shardId << stat.mongoVersion;
}
LOGV2_WARNING(21875,
"Multiversion cluster detected",
"localVersion"_attr = vii.version(),
"shardVersions"_attr = shardVersions.done());
}
Chunk getChunkForMaxBound(const ChunkManager& cm, const BSONObj& max) {
boost::optional<Chunk> chunkWithMaxBound;
cm.forEachChunk([&](const auto& chunk) {
if (chunk.getMax().woCompare(max) == 0) {
chunkWithMaxBound.emplace(chunk);
return false;
}
return true;
});
if (chunkWithMaxBound) {
return *chunkWithMaxBound;
}
return cm.findIntersectingChunkWithSimpleCollation(max);
}
Status processManualMigrationOutcome(OperationContext* opCtx,
const boost::optional<BSONObj>& min,
const boost::optional<BSONObj>& max,
const NamespaceString& nss,
const ShardId& destination,
Status outcome) {
// Since the commands scheduler uses a separate thread to remotely execute a
// request, the resulting clusterTime needs to be explicitly retrieved and set on the
// original context of the requestor to ensure it will be propagated back to the router.
auto& replClient = repl::ReplClientInfo::forClient(opCtx->getClient());
replClient.setLastOpToSystemLastOpTime(opCtx);
if (outcome.isOK()) {
return outcome;
}
auto swCM =
Grid::get(opCtx)->catalogCache()->getShardedCollectionRoutingInfoWithPlacementRefresh(opCtx,
nss);
if (!swCM.isOK()) {
return swCM.getStatus();
}
const auto& cm = swCM.getValue().cm;
const auto currentChunkInfo =
min ? cm.findIntersectingChunkWithSimpleCollation(*min) : getChunkForMaxBound(cm, *max);
if (currentChunkInfo.getShardId() == destination &&
outcome != ErrorCodes::BalancerInterrupted) {
// Migration calls can be interrupted after the metadata is committed but before the command
// finishes the waitForDelete stage. Any failovers, therefore, must always cause the
// moveChunk command to be retried so as to assure that the waitForDelete promise of a
// successful command has been fulfilled.
LOGV2(6036622,
"Migration outcome is not OK, but the transaction was committed. Returning success");
outcome = Status::OK();
}
return outcome;
}
uint64_t getMaxChunkSizeBytes(OperationContext* opCtx, const CollectionType& coll) {
const auto balancerConfig = Grid::get(opCtx)->getBalancerConfiguration();
uassertStatusOK(balancerConfig->refreshAndCheck(opCtx));
return coll.getMaxChunkSizeBytes().value_or(balancerConfig->getMaxChunkSizeBytes());
}
int64_t getMaxChunkSizeMB(OperationContext* opCtx, const CollectionType& coll) {
return getMaxChunkSizeBytes(opCtx, coll) / (1024 * 1024);
}
// Returns a boolean flag indicating whether secondary throttling is enabled and the write concern
// to apply for migrations
std::tuple<bool, WriteConcernOptions> getSecondaryThrottleAndWriteConcern(
const boost::optional<MigrationSecondaryThrottleOptions>& secondaryThrottle) {
if (secondaryThrottle &&
secondaryThrottle->getSecondaryThrottle() == MigrationSecondaryThrottleOptions::kOn) {
if (secondaryThrottle->isWriteConcernSpecified()) {
return {true, secondaryThrottle->getWriteConcern()};
}
return {true, WriteConcernOptions()};
}
return {false, WriteConcernOptions()};
}
const auto _balancerDecoration = ServiceContext::declareDecoration<Balancer>();
const ReplicaSetAwareServiceRegistry::Registerer<Balancer> _balancerRegisterer("Balancer");
/**
* Returns the names of shards that are currently draining. When the balancer is disabled, draining
* shards are stuck in this state as chunks cannot be migrated.
*/
std::vector<std::string> getDrainingShardNames(OperationContext* opCtx) {
// Find the shards that are currently draining.
const auto& configShard = ShardingCatalogManager::get(opCtx)->localConfigShard();
const auto drainingShardsDocs{
uassertStatusOK(
configShard->exhaustiveFindOnConfig(opCtx,
ReadPreferenceSetting{ReadPreference::Nearest},
repl::ReadConcernLevel::kMajorityReadConcern,
NamespaceString::kConfigsvrShardsNamespace,
BSON(ShardType::draining << true),
BSONObj() /* No sorting */,
boost::none /* No limit */))
.docs};
// Build the list of the draining shard names.
std::vector<std::string> drainingShardNames;
std::transform(drainingShardsDocs.begin(),
drainingShardsDocs.end(),
std::back_inserter(drainingShardNames),
[](const auto& shardDoc) {
const auto shardEntry{uassertStatusOK(ShardType::fromBSON(shardDoc))};
return shardEntry.getName();
});
return drainingShardNames;
}
} // namespace
Balancer* Balancer::get(ServiceContext* serviceContext) {
return &_balancerDecoration(serviceContext);
}
Balancer* Balancer::get(OperationContext* operationContext) {
return get(operationContext->getServiceContext());
}
Balancer::Balancer()
: _balancedLastTime(0),
_clusterStats(std::make_unique<ClusterStatisticsImpl>()),
_chunkSelectionPolicy(std::make_unique<BalancerChunkSelectionPolicy>(_clusterStats.get())),
_commandScheduler(std::make_unique<BalancerCommandsSchedulerImpl>()),
_defragmentationPolicy(std::make_unique<BalancerDefragmentationPolicy>(
_clusterStats.get(), [this]() { _onActionsStreamPolicyStateUpdate(); })),
_autoMergerPolicy(
std::make_unique<AutoMergerPolicy>([this]() { _onActionsStreamPolicyStateUpdate(); })),
_imbalancedCollectionsCache(std::make_unique<stdx::unordered_set<NamespaceString>>()) {}
Balancer::~Balancer() {
onShutdown();
}
void Balancer::onStepUpBegin(OperationContext* opCtx, long long term) {
// Before starting step-up, ensure the balancer is ready to start. Specifically, that there is
// not an outstanding termination sequence requested during a previous step down of this node.
joinTermination();
}
void Balancer::onStepUpComplete(OperationContext* opCtx, long long term) {
initiate(opCtx);
}
void Balancer::onStepDown() {
// Asynchronously request to terminate all the worker threads and allow the stepdown sequence to
// continue.
requestTermination();
}
void Balancer::onShutdown() {
// Terminate the balancer thread so it doesn't leak memory.
requestTermination();
joinTermination();
}
void Balancer::onBecomeArbiter() {
// The Balancer is only active on config servers, and arbiters are not permitted in config
// server replica sets.
MONGO_UNREACHABLE;
}
void Balancer::initiate(OperationContext* opCtx) {
stdx::lock_guard<Latch> scopedLock(_mutex);
_imbalancedCollectionsCache->clear();
invariant(_threadSetState == ThreadSetState::Terminated);
_threadSetState = ThreadSetState::Running;
invariant(!_thread.joinable());
invariant(!_actionStreamConsumerThread.joinable());
invariant(!_threadOperationContext);
_thread = stdx::thread([this] { _mainThread(); });
}
void Balancer::requestTermination() {
stdx::lock_guard<Latch> scopedLock(_mutex);
if (_threadSetState != ThreadSetState::Running) {
return;
}
_threadSetState = ThreadSetState::Terminating;
// Interrupt the balancer thread if it has been started. We are guaranteed that the operation
// context of that thread is still alive, because we hold the balancer mutex.
if (_threadOperationContext) {
stdx::lock_guard<Client> scopedClientLock(*_threadOperationContext->getClient());
_threadOperationContext->markKilled(ErrorCodes::InterruptedDueToReplStateChange);
}
_condVar.notify_all();
_actionStreamCondVar.notify_all();
}
void Balancer::joinTermination() {
stdx::unique_lock<Latch> scopedLock(_mutex);
_joinCond.wait(scopedLock, [this] { return _threadSetState == ThreadSetState::Terminated; });
if (_thread.joinable()) {
_thread.join();
}
}
void Balancer::joinCurrentRound(OperationContext* opCtx) {
stdx::unique_lock<Latch> scopedLock(_mutex);
const auto numRoundsAtStart = _numBalancerRounds;
opCtx->waitForConditionOrInterrupt(_condVar, scopedLock, [&] {
return !_inBalancerRound || _numBalancerRounds != numRoundsAtStart;
});
}
Status Balancer::moveRange(OperationContext* opCtx,
const NamespaceString& nss,
const ConfigsvrMoveRange& request,
bool issuedByRemoteUser) {
const auto catalogClient = ShardingCatalogManager::get(opCtx)->localCatalogClient();
auto coll =
catalogClient->getCollection(opCtx, nss, repl::ReadConcernLevel::kMajorityReadConcern);
const auto maxChunkSize = getMaxChunkSizeBytes(opCtx, coll);
const auto fromShardId = [&]() {
const auto [cm, _] = uassertStatusOK(
Grid::get(opCtx)->catalogCache()->getShardedCollectionRoutingInfoWithPlacementRefresh(
opCtx, nss));
if (request.getMin()) {
const auto& chunk = cm.findIntersectingChunkWithSimpleCollation(*request.getMin());
return chunk.getShardId();
} else {
return getChunkForMaxBound(cm, *request.getMax()).getShardId();
}
}();
ShardsvrMoveRange shardSvrRequest(nss);
shardSvrRequest.setDbName(DatabaseName::kAdmin);
shardSvrRequest.setMoveRangeRequestBase(request.getMoveRangeRequestBase());
shardSvrRequest.setMaxChunkSizeBytes(maxChunkSize);
shardSvrRequest.setFromShard(fromShardId);
shardSvrRequest.setCollectionTimestamp(coll.getTimestamp());
shardSvrRequest.setEpoch(coll.getEpoch());
const auto [secondaryThrottle, wc] =
getSecondaryThrottleAndWriteConcern(request.getSecondaryThrottle());
shardSvrRequest.setSecondaryThrottle(secondaryThrottle);
shardSvrRequest.setForceJumbo(request.getForceJumbo());
auto response =
_commandScheduler->requestMoveRange(opCtx, shardSvrRequest, wc, issuedByRemoteUser)
.getNoThrow(opCtx);
return processManualMigrationOutcome(opCtx,
request.getMin(),
request.getMax(),
nss,
shardSvrRequest.getToShard(),
std::move(response));
}
void Balancer::report(OperationContext* opCtx, BSONObjBuilder* builder) {
auto balancerConfig = Grid::get(opCtx)->getBalancerConfiguration();
balancerConfig->refreshAndCheck(opCtx).ignore();
const auto mode = balancerConfig->getBalancerMode();
stdx::lock_guard<Latch> scopedLock(_mutex);
builder->append("mode", BalancerSettingsType::kBalancerModes[mode]);
builder->append("inBalancerRound", _inBalancerRound);
builder->append("numBalancerRounds", _numBalancerRounds);
builder->append("term", repl::ReplicationCoordinator::get(opCtx)->getTerm());
}
void Balancer::_consumeActionStreamLoop() {
Client::initThread("BalancerSecondary");
// TODO(SERVER-74658): Please revisit if this thread could be made killable.
{
stdx::lock_guard<Client> lk(cc());
cc().setSystemOperationUnkillableByStepdown(lk);
}
auto opCtx = cc().makeOperationContext();
executor::ScopedTaskExecutor executor(
Grid::get(opCtx.get())->getExecutorPool()->getFixedExecutor());
ScopeGuard onExitCleanup([this, &executor] {
_defragmentationPolicy->interruptAllDefragmentations();
_autoMergerPolicy->disable();
// Explicitly cancel and drain any outstanding streaming action already dispatched to the
// task executor.
executor->shutdown();
executor->join();
// When shutting down, the task executor may or may not invoke the
// applyActionResponseTo()callback for canceled streaming actions: to ensure a consistent
// state of the balancer after a step down, _outstandingStreamingOps needs then to be reset
// to 0 once all the tasks have been drained.
_outstandingStreamingOps.store(0);
});
// Lambda function to sleep for throttling
auto applyThrottling =
[lastActionTime = Date_t::fromMillisSinceEpoch(0)](const Milliseconds throttle) mutable {
auto timeSinceLastAction = Date_t::now() - lastActionTime;
if (throttle > timeSinceLastAction) {
auto sleepingTime = throttle - timeSinceLastAction;
LOGV2_DEBUG(6443700,
2,
"Applying throttling on balancer secondary thread",
"sleepingTime"_attr = sleepingTime);
sleepFor(sleepingTime);
}
lastActionTime = Date_t::now();
};
auto backOff = Backoff(Seconds(1), Milliseconds::max());
bool errorOccurred = false;
while (true) {
{
stdx::unique_lock<Latch> ul(_mutex);
// Keep asking for more actions if we meet all these conditions:
// - Balancer is in kRunning state
// - There are less than kMaxOutstandingStreamingOperations
// - There were actions to schedule on the previous iteration or there is an update on
// the streams state
auto stopWaitingCondition = [&] {
return _threadSetState != ThreadSetState::Running ||
(_outstandingStreamingOps.load() <= kMaxOutstandingStreamingOperations &&
_actionStreamsStateUpdated.load());
};
if (!errorOccurred) {
_actionStreamCondVar.wait(ul, stopWaitingCondition);
} else {
// Enable retries in case of error by performing a backoff
_actionStreamCondVar.wait_for(
ul, backOff.nextSleep().toSystemDuration(), stopWaitingCondition);
}
if (_threadSetState != ThreadSetState::Running) {
break;
}
}
// Clear flags
errorOccurred = false;
_actionStreamsStateUpdated.store(false);
// Get active streams
auto activeStreams = [&]() -> std::vector<ActionsStreamPolicy*> {
auto balancerConfig = Grid::get(opCtx.get())->getBalancerConfiguration();
std::vector<ActionsStreamPolicy*> streams;
if (balancerConfig->shouldBalanceForAutoMerge() && _autoMergerPolicy->isEnabled()) {
streams.push_back(_autoMergerPolicy.get());
}
if (balancerConfig->shouldBalance()) {
streams.push_back(_defragmentationPolicy.get());
}
return streams;
}();
// Get next action from a random stream together with its stream
auto [nextAction, sourcedStream] =
[&]() -> std::tuple<boost::optional<BalancerStreamAction>, ActionsStreamPolicy*> {
auto client = opCtx->getClient();
std::shuffle(activeStreams.begin(), activeStreams.end(), client->getPrng().urbg());
for (auto stream : activeStreams) {
try {
auto action = stream->getNextStreamingAction(opCtx.get());
if (action.has_value()) {
return std::make_tuple(std::move(action), stream);
}
} catch (const DBException& e) {
LOGV2_WARNING(7435001,
"Failed to get next action from action stream",
"error"_attr = redact(e),
"stream"_attr = stream->getName());
errorOccurred = true;
}
}
return std::make_tuple(boost::none, nullptr);
}();
if (!nextAction.has_value()) {
continue;
}
// Signal there are still actions to be consumed by next iteration
_actionStreamsStateUpdated.store(true);
_outstandingStreamingOps.fetchAndAdd(1);
stdx::visit(
OverloadedVisitor{
[&, stream = sourcedStream](MergeInfo&& mergeAction) {
applyThrottling(Milliseconds(chunkDefragmentationThrottlingMS.load()));
auto result =
_commandScheduler
->requestMergeChunks(opCtx.get(),
mergeAction.nss,
mergeAction.shardId,
mergeAction.chunkRange,
mergeAction.collectionPlacementVersion)
.thenRunOn(*executor)
.onCompletion([this, stream, action = std::move(mergeAction)](
const Status& status) {
_applyStreamingActionResponseToPolicy(action, status, stream);
});
},
[&, stream = sourcedStream](DataSizeInfo&& dataSizeAction) {
auto result =
_commandScheduler
->requestDataSize(opCtx.get(),
dataSizeAction.nss,
dataSizeAction.shardId,
dataSizeAction.chunkRange,
dataSizeAction.version,
dataSizeAction.keyPattern,
dataSizeAction.estimatedValue,
dataSizeAction.maxSize)
.thenRunOn(*executor)
.onCompletion([this, stream, action = std::move(dataSizeAction)](
const StatusWith<DataSizeResponse>& swDataSize) {
_applyStreamingActionResponseToPolicy(action, swDataSize, stream);
});
},
[&, stream = sourcedStream](MergeAllChunksOnShardInfo&& mergeAllChunksAction) {
if (mergeAllChunksAction.applyThrottling) {
applyThrottling(Milliseconds(autoMergerThrottlingMS.load()));
}
auto result =
_commandScheduler
->requestMergeAllChunksOnShard(
opCtx.get(), mergeAllChunksAction.nss, mergeAllChunksAction.shardId)
.thenRunOn(*executor)
.onCompletion(
[this, stream, action = mergeAllChunksAction](
const StatusWith<NumMergedChunks>& swNumMergedChunks) {
_applyStreamingActionResponseToPolicy(
action, swNumMergedChunks, stream);
});
},
[](MigrateInfo&& _) {
uasserted(ErrorCodes::BadValue,
"Migrations cannot be processed as Streaming Actions");
}},
std::move(nextAction.value()));
}
}
void Balancer::_mainThread() {
ON_BLOCK_EXIT([this] {
{
stdx::lock_guard<Latch> scopedLock(_mutex);
_threadSetState = ThreadSetState::Terminated;
LOGV2_DEBUG(21855, 1, "Balancer thread set terminated");
}
_joinCond.notify_all();
});
Client::initThread("Balancer");
// TODO(SERVER-74658): Please revisit if this thread could be made killable.
{
stdx::lock_guard<Client> lk(cc());
cc().setSystemOperationUnkillableByStepdown(lk);
}
auto opCtx = cc().makeOperationContext();
auto shardingContext = Grid::get(opCtx.get());
LOGV2(21856, "CSRS balancer is starting");
{
stdx::lock_guard<Latch> scopedLock(_mutex);
_threadOperationContext = opCtx.get();
}
const Seconds kInitBackoffInterval(10);
auto balancerConfig = shardingContext->getBalancerConfiguration();
while (!_terminationRequested()) {
Status refreshStatus = balancerConfig->refreshAndCheck(opCtx.get());
if (!refreshStatus.isOK()) {
LOGV2_WARNING(
21876,
"Balancer settings could not be loaded because of {error} and will be retried in "
"{backoffInterval}",
"Got error while refreshing balancer settings, will retry with a backoff",
"backoffInterval"_attr = Milliseconds(kInitBackoffInterval),
"error"_attr = refreshStatus);
_sleepFor(opCtx.get(), kInitBackoffInterval);
continue;
}
break;
}
LOGV2(6036605, "Starting command scheduler");
_commandScheduler->start(opCtx.get());
_actionStreamConsumerThread = stdx::thread([&] { _consumeActionStreamLoop(); });
LOGV2(6036606, "Balancer worker thread initialised. Entering main loop.");
// Main balancer loop
auto lastMigrationTime = Date_t::fromMillisSinceEpoch(0);
auto lastDrainingShardsCheckTime{Date_t::fromMillisSinceEpoch(0)};
while (!_terminationRequested()) {
BalanceRoundDetails roundDetails;
_beginRound(opCtx.get());
try {
shardingContext->shardRegistry()->reload(opCtx.get());
uassert(13258, "oids broken after resetting!", _checkOIDs(opCtx.get()));
Status refreshStatus = balancerConfig->refreshAndCheck(opCtx.get());
if (!refreshStatus.isOK()) {
LOGV2_WARNING(21877,
"Skipping balancing round due to {error}",
"Skipping balancing round",
"error"_attr = refreshStatus);
_endRound(opCtx.get(), kBalanceRoundDefaultInterval);
continue;
}
if (!balancerConfig->shouldBalance() || _terminationRequested()) {
if (balancerConfig->getBalancerMode() == BalancerSettingsType::BalancerMode::kOff &&
Date_t::now() - lastDrainingShardsCheckTime >= kDrainingShardsCheckInterval) {
const auto drainingShardNames{getDrainingShardNames(opCtx.get())};
if (!drainingShardNames.empty()) {
LOGV2_WARNING(6434000,
"Draining of removed shards cannot be completed because the "
"balancer is disabled",
"shards"_attr = drainingShardNames);
}
lastDrainingShardsCheckTime = Date_t::now();
}
_autoMergerPolicy->disable();
LOGV2_DEBUG(21859, 1, "Skipping balancing round because balancing is disabled");
_endRound(opCtx.get(), kBalanceRoundDefaultInterval);
continue;
}
if (!_autoMergerPolicy->isEnabled() && balancerConfig->shouldBalanceForAutoMerge()) {
_autoMergerPolicy->enable();
}
boost::optional<Milliseconds> forcedBalancerRoundInterval(boost::none);
overrideBalanceRoundInterval.execute([&](const BSONObj& data) {
forcedBalancerRoundInterval = Milliseconds(data["intervalMs"].numberInt());
LOGV2(21864,
"overrideBalanceRoundInterval: using customized balancing interval",
"balancerInterval"_attr = *forcedBalancerRoundInterval);
});
LOGV2_DEBUG(21860,
1,
"Start balancing round. waitForDelete: {waitForDelete}, "
"secondaryThrottle: {secondaryThrottle}",
"Start balancing round",
"waitForDelete"_attr = balancerConfig->waitForDelete(),
"secondaryThrottle"_attr = balancerConfig->getSecondaryThrottle().toBSON());
static Occasionally sampler;
if (sampler.tick()) {
warnOnMultiVersion(uassertStatusOK(_clusterStats->getStats(opCtx.get())));
}
// Collect and apply up-to-date configuration values on the cluster collections.
_defragmentationPolicy->startCollectionDefragmentations(opCtx.get());
// Reactivate the Automerger if needed.
_autoMergerPolicy->checkInternalUpdates();
// The current configuration is allowing the balancer to perform operations.
// Unblock the secondary thread if needed.
_actionStreamCondVar.notify_all();
{
Status status = _splitChunksIfNeeded(opCtx.get());
if (!status.isOK()) {
LOGV2_WARNING(21878,
"Failed to split chunks due to {error}",
"Failed to split chunks",
"error"_attr = status);
} else {
LOGV2_DEBUG(21861, 1, "Done enforcing zone range boundaries.");
}
const std::vector<ClusterStatistics::ShardStatistics> shardStats =
uassertStatusOK(_clusterStats->getStats(opCtx.get()));
stdx::unordered_set<ShardId> availableShards;
availableShards.reserve(shardStats.size());
std::transform(
shardStats.begin(),
shardStats.end(),
std::inserter(availableShards, availableShards.end()),
[](const ClusterStatistics::ShardStatistics& shardStatistics) -> ShardId {
return shardStatistics.shardId;
});
const auto chunksToDefragment =
_defragmentationPolicy->selectChunksToMove(opCtx.get(), &availableShards);
const auto chunksToRebalance = uassertStatusOK(
_chunkSelectionPolicy->selectChunksToMove(opCtx.get(),
shardStats,
&availableShards,
_imbalancedCollectionsCache.get()));
if (chunksToRebalance.empty() && chunksToDefragment.empty()) {
LOGV2_DEBUG(21862, 1, "No need to move any chunk");
_balancedLastTime = 0;
LOGV2_DEBUG(21863, 1, "End balancing round");
_endRound(opCtx.get(),
forcedBalancerRoundInterval ? *forcedBalancerRoundInterval
: kBalanceRoundDefaultInterval);
} else {
auto timeSinceLastMigration = Date_t::now() - lastMigrationTime;
_sleepFor(opCtx.get(),
forcedBalancerRoundInterval
? *forcedBalancerRoundInterval - timeSinceLastMigration
: Milliseconds(balancerMigrationsThrottlingMs.load()) -
timeSinceLastMigration);
_balancedLastTime =
_moveChunks(opCtx.get(), chunksToRebalance, chunksToDefragment);
lastMigrationTime = Date_t::now();
roundDetails.setSucceeded(
static_cast<int>(chunksToRebalance.size() + chunksToDefragment.size()),
_balancedLastTime,
_imbalancedCollectionsCache->size());
auto catalogManager = ShardingCatalogManager::get(opCtx.get());
ShardingLogging::get(opCtx.get())
->logAction(opCtx.get(),
"balancer.round",
"",
roundDetails.toBSON(),
catalogManager->localConfigShard(),
catalogManager->localCatalogClient())
.ignore();
LOGV2_DEBUG(6679500, 1, "End balancing round");
// Migration throttling of `balancerMigrationsThrottlingMs` will be applied
// before the next call to _moveChunks, so don't sleep here.
_endRound(opCtx.get(), Milliseconds(0));
}
}
} catch (const DBException& e) {
LOGV2(21865,
"caught exception while doing balance: {error}",
"Error while doing balance",
"error"_attr = e);
// Just to match the opening statement if in log level 1
LOGV2_DEBUG(21866, 1, "End balancing round");
// This round failed, tell the world!
roundDetails.setFailed(e.what());
auto catalogManager = ShardingCatalogManager::get(opCtx.get());
ShardingLogging::get(opCtx.get())
->logAction(opCtx.get(),
"balancer.round",
"",
roundDetails.toBSON(),
catalogManager->localConfigShard(),
catalogManager->localCatalogClient())
.ignore();
// Sleep a fair amount before retrying because of the error
_endRound(opCtx.get(), kBalanceRoundDefaultInterval);
}
}
{
stdx::lock_guard<Latch> scopedLock(_mutex);
invariant(_threadSetState == ThreadSetState::Terminating);
}
_commandScheduler->stop();
_actionStreamConsumerThread.join();
{
stdx::lock_guard<Latch> scopedLock(_mutex);
_threadOperationContext = nullptr;
}
LOGV2(21867, "CSRS balancer is now stopped");
}
void Balancer::_applyStreamingActionResponseToPolicy(const BalancerStreamAction& action,
const BalancerStreamActionResponse& response,
ActionsStreamPolicy* policy) {
invariant(_outstandingStreamingOps.addAndFetch(-1) >= 0);
ThreadClient tc("BalancerSecondaryThread::applyActionResponse", getGlobalServiceContext());
// TODO(SERVER-74658): Please revisit if this thread could be made killable.
{
stdx::lock_guard<Client> lk(*tc.get());
tc.get()->setSystemOperationUnkillableByStepdown(lk);
}
auto opCtx = tc->makeOperationContext();
policy->applyActionResult(opCtx.get(), action, response);
};
bool Balancer::_terminationRequested() {
stdx::lock_guard<Latch> scopedLock(_mutex);
return (_threadSetState != ThreadSetState::Running);
}
void Balancer::_beginRound(OperationContext* opCtx) {
stdx::unique_lock<Latch> lock(_mutex);
_inBalancerRound = true;
_condVar.notify_all();
}
void Balancer::_endRound(OperationContext* opCtx, Milliseconds waitTimeout) {
{
stdx::lock_guard<Latch> lock(_mutex);
_inBalancerRound = false;
_numBalancerRounds++;
_condVar.notify_all();
}
MONGO_IDLE_THREAD_BLOCK;
_sleepFor(opCtx, waitTimeout);
}
void Balancer::_sleepFor(OperationContext* opCtx, Milliseconds waitTimeout) {
stdx::unique_lock<Latch> lock(_mutex);
_condVar.wait_for(lock, waitTimeout.toSystemDuration(), [&] {
return _threadSetState != ThreadSetState::Running;
});
}
bool Balancer::_checkOIDs(OperationContext* opCtx) {
auto shardingContext = Grid::get(opCtx);
const auto all = shardingContext->shardRegistry()->getAllShardIds(opCtx);
// map of OID machine ID => shardId
map<int, ShardId> oids;
for (const ShardId& shardId : all) {
if (_terminationRequested()) {
return false;
}
auto shardStatus = shardingContext->shardRegistry()->getShard(opCtx, shardId);
if (!shardStatus.isOK()) {
continue;
}
const auto s = std::move(shardStatus.getValue());
auto result = uassertStatusOK(
s->runCommandWithFixedRetryAttempts(opCtx,
ReadPreferenceSetting{ReadPreference::PrimaryOnly},
DatabaseName::kAdmin,
BSON("features" << 1),
Seconds(30),
Shard::RetryPolicy::kIdempotent));
uassertStatusOK(result.commandStatus);
BSONObj f = std::move(result.response);
if (f["oidMachine"].isNumber()) {