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
|
/**
* 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::kQuery
#include "mongo/platform/basic.h"
#include "mongo/s/query/cluster_find.h"
#include <fmt/format.h>
#include <memory>
#include <set>
#include <vector>
#include "mongo/base/status_with.h"
#include "mongo/bson/util/bson_extract.h"
#include "mongo/client/connpool.h"
#include "mongo/client/read_preference.h"
#include "mongo/db/auth/authorization_session.h"
#include "mongo/db/commands.h"
#include "mongo/db/curop.h"
#include "mongo/db/curop_failpoint_helpers.h"
#include "mongo/db/pipeline/change_stream_invalidation_info.h"
#include "mongo/db/query/canonical_query.h"
#include "mongo/db/query/find_common.h"
#include "mongo/db/query/getmore_request.h"
#include "mongo/db/query/query_planner_common.h"
#include "mongo/executor/task_executor_pool.h"
#include "mongo/logv2/log.h"
#include "mongo/platform/overflow_arithmetic.h"
#include "mongo/s/catalog_cache.h"
#include "mongo/s/client/num_hosts_targeted_metrics.h"
#include "mongo/s/client/shard_registry.h"
#include "mongo/s/cluster_commands_helpers.h"
#include "mongo/s/grid.h"
#include "mongo/s/query/async_results_merger.h"
#include "mongo/s/query/cluster_client_cursor_impl.h"
#include "mongo/s/query/cluster_cursor_manager.h"
#include "mongo/s/query/establish_cursors.h"
#include "mongo/s/query/store_possible_cursor.h"
#include "mongo/s/stale_exception.h"
#include "mongo/s/transaction_router.h"
#include "mongo/util/fail_point.h"
#include "mongo/util/scopeguard.h"
namespace mongo {
namespace {
using namespace fmt::literals;
static const BSONObj kSortKeyMetaProjection = BSON("$meta"
<< "sortKey");
static const BSONObj kGeoNearDistanceMetaProjection = BSON("$meta"
<< "geoNearDistance");
// We must allow some amount of overhead per result document, since when we make a cursor response
// the documents are elements of a BSONArray. The overhead is 1 byte/doc for the type + 1 byte/doc
// for the field name's null terminator + 1 byte per digit in the array index. The index can be no
// more than 8 decimal digits since the response is at most 16MB, and 16 * 1024 * 1024 < 1 * 10^8.
static const int kPerDocumentOverheadBytesUpperBound = 10;
const char kFindCmdName[] = "find";
/**
* Given the FindCommandRequest 'findCommand' being executed by mongos, returns a copy of the query
* which is suitable for forwarding to the targeted hosts.
*/
StatusWith<std::unique_ptr<FindCommandRequest>> transformQueryForShards(
const FindCommandRequest& findCommand, bool appendGeoNearDistanceProjection) {
// If there is a limit, we forward the sum of the limit and the skip.
boost::optional<int64_t> newLimit;
if (findCommand.getLimit()) {
long long newLimitValue;
if (overflow::add(
*findCommand.getLimit(), findCommand.getSkip().value_or(0), &newLimitValue)) {
return Status(
ErrorCodes::Overflow,
str::stream()
<< "sum of limit and skip cannot be represented as a 64-bit integer, limit: "
<< *findCommand.getLimit() << ", skip: " << findCommand.getSkip().value_or(0));
}
newLimit = newLimitValue;
}
// Similarly, if nToReturn is set, we forward the sum of nToReturn and the skip.
boost::optional<int64_t> newNToReturn;
if (findCommand.getNtoreturn()) {
// 'singleBatch' and ntoreturn mean the same as 'singleBatch' and limit, so perform the
// conversion.
if (findCommand.getSingleBatch()) {
int64_t newLimitValue;
if (overflow::add(*findCommand.getNtoreturn(),
findCommand.getSkip().value_or(0),
&newLimitValue)) {
return Status(ErrorCodes::Overflow,
str::stream()
<< "sum of ntoreturn and skip cannot be represented as a 64-bit "
"integer, ntoreturn: "
<< *findCommand.getNtoreturn()
<< ", skip: " << findCommand.getSkip().value_or(0));
}
newLimit = newLimitValue;
} else {
int64_t newNToReturnValue;
if (overflow::add(*findCommand.getNtoreturn(),
findCommand.getSkip().value_or(0),
&newNToReturnValue)) {
return Status(ErrorCodes::Overflow,
str::stream()
<< "sum of ntoreturn and skip cannot be represented as a 64-bit "
"integer, ntoreturn: "
<< *findCommand.getNtoreturn()
<< ", skip: " << findCommand.getSkip().value_or(0));
}
newNToReturn = newNToReturnValue;
}
}
// If there is a sort other than $natural, we send a sortKey meta-projection to the remote node.
BSONObj newProjection = findCommand.getProjection();
if (!findCommand.getSort().isEmpty() &&
!findCommand.getSort()[query_request_helper::kNaturalSortField]) {
BSONObjBuilder projectionBuilder;
projectionBuilder.appendElements(findCommand.getProjection());
projectionBuilder.append(AsyncResultsMerger::kSortKeyField, kSortKeyMetaProjection);
newProjection = projectionBuilder.obj();
}
if (appendGeoNearDistanceProjection) {
invariant(findCommand.getSort().isEmpty());
BSONObjBuilder projectionBuilder;
projectionBuilder.appendElements(findCommand.getProjection());
projectionBuilder.append(AsyncResultsMerger::kSortKeyField, kGeoNearDistanceMetaProjection);
newProjection = projectionBuilder.obj();
}
auto newQR = std::make_unique<FindCommandRequest>(findCommand);
newQR->setProjection(newProjection);
newQR->setSkip(boost::none);
newQR->setLimit(newLimit);
newQR->setNtoreturn(newNToReturn);
// Even if the client sends us singleBatch=true, we may need to retrieve
// multiple batches from a shard in order to return the single requested batch to the client.
// Therefore, we must always send singleBatch=false to the shards.
newQR->setSingleBatch(false);
// Any expansion of the 'showRecordId' flag should have already happened on mongos.
if (newQR->getShowRecordId())
newQR->setShowRecordId(false);
uassertStatusOK(query_request_helper::validateFindCommandRequest(*newQR));
return std::move(newQR);
}
/**
* Constructs the find commands sent to each targeted shard to establish cursors, attaching the
* shardVersion and txnNumber, if necessary.
*/
std::vector<std::pair<ShardId, BSONObj>> constructRequestsForShards(
OperationContext* opCtx,
const ChunkManager& cm,
const std::set<ShardId>& shardIds,
const CanonicalQuery& query,
bool appendGeoNearDistanceProjection) {
std::unique_ptr<FindCommandRequest> findCommandToForward;
if (shardIds.size() > 1) {
findCommandToForward = uassertStatusOK(transformQueryForShards(
query.getFindCommandRequest(), appendGeoNearDistanceProjection));
} else {
// Forwards the FindCommandRequest as is to a single shard so that limit and skip can
// be applied on mongod.
findCommandToForward = std::make_unique<FindCommandRequest>(query.getFindCommandRequest());
}
auto& readConcernArgs = repl::ReadConcernArgs::get(opCtx);
if (readConcernArgs.wasAtClusterTimeSelected()) {
// If mongos selected atClusterTime or received it from client, transmit it to shard.
findCommandToForward->setReadConcern(readConcernArgs.toBSONInner());
}
auto shardRegistry = Grid::get(opCtx)->shardRegistry();
std::vector<std::pair<ShardId, BSONObj>> requests;
for (const auto& shardId : shardIds) {
const auto shard = uassertStatusOK(shardRegistry->getShard(opCtx, shardId));
invariant(!shard->isConfig() || shard->getConnString());
BSONObjBuilder cmdBuilder;
findCommandToForward->serialize(BSONObj(), &cmdBuilder);
if (cm.isSharded()) {
cm.getVersion(shardId).appendToCommand(&cmdBuilder);
} else if (!query.nss().isOnInternalDb()) {
ChunkVersion::UNSHARDED().appendToCommand(&cmdBuilder);
cmdBuilder.append("databaseVersion", cm.dbVersion().toBSON());
}
if (opCtx->getTxnNumber()) {
cmdBuilder.append(OperationSessionInfo::kTxnNumberFieldName, *opCtx->getTxnNumber());
}
requests.emplace_back(shardId, cmdBuilder.obj());
}
return requests;
}
void updateNumHostsTargetedMetrics(OperationContext* opCtx,
const ChunkManager& cm,
int nTargetedShards) {
int nShardsOwningChunks = 0;
if (cm.isSharded()) {
nShardsOwningChunks = cm.getNShardsOwningChunks();
}
auto targetType = NumHostsTargetedMetrics::get(opCtx).parseTargetType(
opCtx, nTargetedShards, nShardsOwningChunks);
NumHostsTargetedMetrics::get(opCtx).addNumHostsTargeted(
NumHostsTargetedMetrics::QueryType::kFindCmd, targetType);
}
CursorId runQueryWithoutRetrying(OperationContext* opCtx,
const CanonicalQuery& query,
const ReadPreferenceSetting& readPref,
const ChunkManager& cm,
std::vector<BSONObj>* results,
bool* partialResultsReturned) {
auto findCommand = query.getFindCommandRequest();
// Get the set of shards on which we will run the query.
auto shardIds = getTargetedShardsForQuery(
query.getExpCtx(), cm, findCommand.getFilter(), findCommand.getCollation());
// Construct the query and parameters. Defer setting skip and limit here until
// we determine if the query is targeting multi-shards or a single shard below.
ClusterClientCursorParams params(
query.nss(), APIParameters::get(opCtx), readPref, ReadConcernArgs::get(opCtx));
params.originatingCommandObj = CurOp::get(opCtx)->opDescription().getOwned();
params.batchSize =
findCommand.getBatchSize() ? findCommand.getBatchSize() : findCommand.getNtoreturn();
params.tailableMode = query_request_helper::getTailableMode(findCommand);
params.isAllowPartialResults = findCommand.getAllowPartialResults();
params.lsid = opCtx->getLogicalSessionId();
params.txnNumber = opCtx->getTxnNumber();
params.originatingPrivileges = {
Privilege(ResourcePattern::forExactNamespace(query.nss()), ActionType::find)};
if (TransactionRouter::get(opCtx)) {
params.isAutoCommit = false;
}
// This is the batchSize passed to each subsequent getMore command issued by the cursor. We
// usually use the batchSize associated with the initial find, but as it is illegal to send a
// getMore with a batchSize of 0, we set it to use the default batchSize logic.
if (params.batchSize && *params.batchSize == 0) {
params.batchSize = boost::none;
}
// $natural sort is actually a hint to use a collection scan, and shouldn't be treated like a
// sort on mongos. Including a $natural anywhere in the sort spec results in the whole sort
// being considered a hint to use a collection scan.
BSONObj sortComparatorObj;
if (query.getSortPattern() && !findCommand.getSort()[query_request_helper::kNaturalSortField]) {
// We have already validated the input sort object. Serialize the raw sort spec into one
// suitable for use as the ordering specification in BSONObj::woCompare(). In particular, we
// want to eliminate sorts using expressions (like $meta) and replace them with a
// placeholder. When mongos performs a merge-sort, any $meta expressions have already been
// performed on the shards. Mongos just needs to know the length of the sort pattern and
// whether each part of the sort pattern is ascending or descending.
sortComparatorObj = query.getSortPattern()
->serialize(SortPattern::SortKeySerialization::kForSortKeyMerging)
.toBson();
}
bool appendGeoNearDistanceProjection = false;
bool compareWholeSortKeyOnRouter = false;
if (!query.getSortPattern() &&
QueryPlannerCommon::hasNode(query.root(), MatchExpression::GEO_NEAR)) {
// There is no specified sort, and there is a GEO_NEAR node. This means we should merge sort
// by the geoNearDistance. Request the projection {$sortKey: <geoNearDistance>} from the
// shards. Indicate to the AsyncResultsMerger that it should extract the sort key
// {"$sortKey": <geoNearDistance>} and sort by the order {"$sortKey": 1}.
sortComparatorObj = AsyncResultsMerger::kWholeSortKeySortPattern;
compareWholeSortKeyOnRouter = true;
appendGeoNearDistanceProjection = true;
}
// Tailable cursors can't have a sort, which should have already been validated.
tassert(4457013,
"tailable cursor unexpectedly has a sort",
sortComparatorObj.isEmpty() || !findCommand.getTailable());
// Construct the requests that we will use to establish cursors on the targeted shards,
// attaching the shardVersion and txnNumber, if necessary.
auto requests =
constructRequestsForShards(opCtx, cm, shardIds, query, appendGeoNearDistanceProjection);
// Establish the cursors with a consistent shardVersion across shards.
params.remotes = establishCursors(opCtx,
Grid::get(opCtx)->getExecutorPool()->getArbitraryExecutor(),
query.nss(),
readPref,
requests,
findCommand.getAllowPartialResults());
// Determine whether the cursor we may eventually register will be single- or multi-target.
const auto cursorType = params.remotes.size() > 1
? ClusterCursorManager::CursorType::MultiTarget
: ClusterCursorManager::CursorType::SingleTarget;
// Only set skip, limit and sort to be applied to on the router for the multi-shard case. For
// the single-shard case skip/limit as well as sorts are appled on mongod.
if (cursorType == ClusterCursorManager::CursorType::MultiTarget) {
params.skipToApplyOnRouter = findCommand.getSkip();
params.limit = findCommand.getLimit();
params.sortToApplyOnRouter = sortComparatorObj;
params.compareWholeSortKeyOnRouter = compareWholeSortKeyOnRouter;
}
// Transfer the established cursors to a ClusterClientCursor.
auto ccc = ClusterClientCursorImpl::make(
opCtx, Grid::get(opCtx)->getExecutorPool()->getArbitraryExecutor(), std::move(params));
// Retrieve enough data from the ClusterClientCursor for the first batch of results.
FindCommon::waitInFindBeforeMakingBatch(opCtx, query);
auto cursorState = ClusterCursorManager::CursorState::NotExhausted;
int bytesBuffered = 0;
// This loop will not result in actually calling getMore against shards, but just loading
// results from the initial batches (that were obtained while establishing cursors) into
// 'results'.
while (!FindCommon::enoughForFirstBatch(findCommand, results->size())) {
auto next = uassertStatusOK(ccc->next(RouterExecStage::ExecContext::kInitialFind));
if (next.isEOF()) {
// We reached end-of-stream. If the cursor is not tailable, then we mark it as
// exhausted. If it is tailable, usually we keep it open (i.e. "NotExhausted") even
// when we reach end-of-stream. However, if all the remote cursors are exhausted, there
// is no hope of returning data and thus we need to close the mongos cursor as well.
if (!ccc->isTailable() || ccc->remotesExhausted()) {
cursorState = ClusterCursorManager::CursorState::Exhausted;
}
break;
}
auto nextObj = *next.getResult();
// If adding this object will cause us to exceed the message size limit, then we stash it
// for later.
if (!FindCommon::haveSpaceForNext(nextObj, results->size(), bytesBuffered)) {
ccc->queueResult(nextObj);
break;
}
// Add doc to the batch. Account for the space overhead associated with returning this doc
// inside a BSON array.
bytesBuffered += (nextObj.objsize() + kPerDocumentOverheadBytesUpperBound);
results->push_back(std::move(nextObj));
}
ccc->detachFromOperationContext();
if (findCommand.getSingleBatch() && !ccc->isTailable()) {
cursorState = ClusterCursorManager::CursorState::Exhausted;
}
// Fill out query exec properties.
CurOp::get(opCtx)->debug().nShards = ccc->getNumRemotes();
CurOp::get(opCtx)->debug().nreturned = results->size();
// If the caller wants to know whether the cursor returned partial results, set it here.
if (partialResultsReturned) {
*partialResultsReturned = ccc->partialResultsReturned();
}
// If the cursor is exhausted, then there are no more results to return and we don't need to
// allocate a cursor id.
if (cursorState == ClusterCursorManager::CursorState::Exhausted) {
CurOp::get(opCtx)->debug().cursorExhausted = true;
if (shardIds.size() > 0) {
updateNumHostsTargetedMetrics(opCtx, cm, shardIds.size());
}
return CursorId(0);
}
// Register the cursor with the cursor manager for subsequent getMore's.
auto cursorManager = Grid::get(opCtx)->getCursorManager();
const auto cursorLifetime = findCommand.getNoCursorTimeout()
? ClusterCursorManager::CursorLifetime::Immortal
: ClusterCursorManager::CursorLifetime::Mortal;
auto authUsers = AuthorizationSession::get(opCtx->getClient())->getAuthenticatedUserNames();
ccc->incNBatches();
auto cursorId = uassertStatusOK(cursorManager->registerCursor(
opCtx, ccc.releaseCursor(), query.nss(), cursorType, cursorLifetime, authUsers));
// Record the cursorID in CurOp.
CurOp::get(opCtx)->debug().cursorid = cursorId;
if (shardIds.size() > 0) {
updateNumHostsTargetedMetrics(opCtx, cm, shardIds.size());
}
return cursorId;
}
/**
* Populates or re-populates some state of the OperationContext from what's stored on the cursor
* and/or what's specified on the request.
*/
Status setUpOperationContextStateForGetMore(OperationContext* opCtx,
const GetMoreCommandRequest& cmd,
const ClusterCursorManager::PinnedCursor& cursor) {
if (auto readPref = cursor->getReadPreference()) {
ReadPreferenceSetting::get(opCtx) = *readPref;
}
if (auto readConcern = cursor->getReadConcern()) {
// Used to return "atClusterTime" in cursor replies to clients for snapshot reads.
ReadConcernArgs::get(opCtx) = *readConcern;
}
auto apiParamsFromClient = APIParameters::get(opCtx);
uassert(ErrorCodes::APIMismatchError,
"API parameter mismatch: getMore used params {}, the cursor-creating command "
"used {}"_format(apiParamsFromClient.toBSON().toString(),
cursor->getAPIParameters().toBSON().toString()),
apiParamsFromClient == cursor->getAPIParameters());
// If the originating command had a 'comment' field, we extract it and set it on opCtx. Note
// that if the 'getMore' command itself has a 'comment' field, we give precedence to it.
auto comment = cursor->getOriginatingCommand()["comment"];
if (!opCtx->getComment() && comment) {
opCtx->setComment(comment.wrap());
}
if (cursor->isTailableAndAwaitData()) {
// For tailable + awaitData cursors, the request may have indicated a maximum amount of time
// to wait for new data. If not, default it to 1 second. We track the deadline instead via
// the 'waitForInsertsDeadline' decoration.
auto timeout = Milliseconds{cmd.getMaxTimeMS().value_or(1000)};
awaitDataState(opCtx).waitForInsertsDeadline =
opCtx->getServiceContext()->getPreciseClockSource()->now() + timeout;
invariant(cursor->setAwaitDataTimeout(timeout).isOK());
} else if (cmd.getMaxTimeMS()) {
return {ErrorCodes::BadValue,
"maxTimeMS can only be used with getMore for tailable, awaitData cursors"};
} else if (cursor->getLeftoverMaxTimeMicros() < Microseconds::max()) {
// Be sure to do this only for non-tailable cursors.
opCtx->setDeadlineAfterNowBy(cursor->getLeftoverMaxTimeMicros(),
ErrorCodes::MaxTimeMSExpired);
}
return Status::OK();
}
} // namespace
const size_t ClusterFind::kMaxRetries = 10;
CursorId ClusterFind::runQuery(OperationContext* opCtx,
const CanonicalQuery& query,
const ReadPreferenceSetting& readPref,
std::vector<BSONObj>* results,
bool* partialResultsReturned) {
// If the user supplied a 'partialResultsReturned' out-parameter, default it to false here.
if (partialResultsReturned) {
*partialResultsReturned = false;
}
// We must always have a BSONObj vector into which to output our results.
invariant(results);
auto findCommand = query.getFindCommandRequest();
// Projection on the reserved sort key field is illegal in mongos.
if (findCommand.getProjection().hasField(AsyncResultsMerger::kSortKeyField)) {
uasserted(ErrorCodes::BadValue,
str::stream() << "Projection contains illegal field '"
<< AsyncResultsMerger::kSortKeyField
<< "': " << findCommand.getProjection());
}
// Attempting to establish a resumable query through mongoS is illegal.
uassert(ErrorCodes::BadValue,
"Queries on mongoS may not request or provide a resume token",
!findCommand.getRequestResumeToken() && findCommand.getResumeAfter().isEmpty());
auto const catalogCache = Grid::get(opCtx)->catalogCache();
// Re-target and re-send the initial find command to the shards until we have established the
// shard version.
for (size_t retries = 1; retries <= kMaxRetries; ++retries) {
auto swCM = getCollectionRoutingInfoForTxnCmd(opCtx, query.nss());
if (swCM == ErrorCodes::NamespaceNotFound) {
// If the database doesn't exist, we successfully return an empty result set without
// creating a cursor.
return CursorId(0);
}
const auto cm = uassertStatusOK(std::move(swCM));
try {
return runQueryWithoutRetrying(
opCtx, query, readPref, cm, results, partialResultsReturned);
} catch (ExceptionFor<ErrorCodes::StaleDbVersion>& ex) {
if (retries >= kMaxRetries) {
// Check if there are no retries remaining, so the last received error can be
// propagated to the caller.
ex.addContext(str::stream()
<< "Failed to run query after " << kMaxRetries << " retries");
throw;
}
LOGV2_DEBUG(22839,
1,
"Received error status for query",
"query"_attr = redact(query.toStringShort()),
"attemptNumber"_attr = retries,
"maxRetries"_attr = kMaxRetries,
"error"_attr = redact(ex));
// Mark database entry in cache as stale.
Grid::get(opCtx)->catalogCache()->onStaleDatabaseVersion(ex->getDb(),
ex->getVersionWanted());
if (auto txnRouter = TransactionRouter::get(opCtx)) {
if (!txnRouter.canContinueOnStaleShardOrDbError(kFindCmdName, ex.toStatus())) {
throw;
}
// Reset the default global read timestamp so the retry's routing table reflects the
// chunk placement after the refresh (no-op if the transaction is not running with
// snapshot read concern).
txnRouter.onStaleShardOrDbError(opCtx, kFindCmdName, ex.toStatus());
txnRouter.setDefaultAtClusterTime(opCtx);
}
} catch (DBException& ex) {
if (retries >= kMaxRetries) {
// Check if there are no retries remaining, so the last received error can be
// propagated to the caller.
ex.addContext(str::stream()
<< "Failed to run query after " << kMaxRetries << " retries");
throw;
} else if (!ErrorCodes::isStaleShardVersionError(ex.code()) &&
ex.code() != ErrorCodes::ShardInvalidatedForTargeting &&
ex.code() != ErrorCodes::ShardNotFound) {
if (ErrorCodes::isRetriableError(ex.code())) {
ex.addContext("Encountered retryable error during query");
} else {
// Errors other than stale metadata or from trying to reach a non existent shard
// are fatal to the operation. Network errors and replication retries happen at
// the level of the AsyncResultsMerger.
ex.addContext("Encountered non-retryable error during query");
}
throw;
}
LOGV2_DEBUG(22840,
1,
"Received error status for query",
"query"_attr = redact(query.toStringShort()),
"attemptNumber"_attr = retries,
"maxRetries"_attr = kMaxRetries,
"error"_attr = redact(ex));
if (ex.code() != ErrorCodes::ShardInvalidatedForTargeting) {
if (auto staleInfo = ex.extraInfo<StaleConfigInfo>()) {
catalogCache->invalidateShardOrEntireCollectionEntryForShardedCollection(
query.nss(), staleInfo->getVersionWanted(), staleInfo->getShardId());
} else {
catalogCache->invalidateCollectionEntry_LINEARIZABLE(query.nss());
}
}
catalogCache->setOperationShouldBlockBehindCatalogCacheRefresh(opCtx, true);
if (auto txnRouter = TransactionRouter::get(opCtx)) {
if (!txnRouter.canContinueOnStaleShardOrDbError(kFindCmdName, ex.toStatus())) {
if (ex.code() == ErrorCodes::ShardInvalidatedForTargeting) {
(void)catalogCache->getCollectionRoutingInfoWithRefresh(opCtx, query.nss());
}
throw;
}
// Reset the default global read timestamp so the retry's routing table reflects the
// chunk placement after the refresh (no-op if the transaction is not running with
// snapshot read concern).
txnRouter.onStaleShardOrDbError(opCtx, kFindCmdName, ex.toStatus());
txnRouter.setDefaultAtClusterTime(opCtx);
}
}
}
MONGO_UNREACHABLE
}
/**
* Validates that the lsid on the OperationContext matches that on the cursor, returning it to the
* ClusterClusterCursor manager if it does not.
*/
void validateLSID(OperationContext* opCtx,
int64_t cursorId,
const ClusterCursorManager::PinnedCursor& cursor) {
if (opCtx->getLogicalSessionId() && !cursor->getLsid()) {
uasserted(50799,
str::stream() << "Cannot run getMore on cursor " << cursorId
<< ", which was not created in a session, in session "
<< *opCtx->getLogicalSessionId());
}
if (!opCtx->getLogicalSessionId() && cursor->getLsid()) {
uasserted(50800,
str::stream() << "Cannot run getMore on cursor " << cursorId
<< ", which was created in session " << *cursor->getLsid()
<< ", without an lsid");
}
if (opCtx->getLogicalSessionId() && cursor->getLsid() &&
(*opCtx->getLogicalSessionId() != *cursor->getLsid())) {
uasserted(50801,
str::stream() << "Cannot run getMore on cursor " << cursorId
<< ", which was created in session " << *cursor->getLsid()
<< ", in session " << *opCtx->getLogicalSessionId());
}
}
/**
* Validates that the txnNumber on the OperationContext matches that on the cursor, returning it to
* the ClusterClusterCursor manager if it does not.
*/
void validateTxnNumber(OperationContext* opCtx,
int64_t cursorId,
const ClusterCursorManager::PinnedCursor& cursor) {
if (opCtx->getTxnNumber() && !cursor->getTxnNumber()) {
uasserted(50802,
str::stream() << "Cannot run getMore on cursor " << cursorId
<< ", which was not created in a transaction, in transaction "
<< *opCtx->getTxnNumber());
}
if (!opCtx->getTxnNumber() && cursor->getTxnNumber()) {
uasserted(50803,
str::stream() << "Cannot run getMore on cursor " << cursorId
<< ", which was created in transaction " << *cursor->getTxnNumber()
<< ", without a txnNumber");
}
if (opCtx->getTxnNumber() && cursor->getTxnNumber() &&
(*opCtx->getTxnNumber() != *cursor->getTxnNumber())) {
uasserted(50804,
str::stream() << "Cannot run getMore on cursor " << cursorId
<< ", which was created in transaction " << *cursor->getTxnNumber()
<< ", in transaction " << *opCtx->getTxnNumber());
}
}
/**
* Validates that the OperationSessionInfo (i.e. txnNumber and lsid) on the OperationContext match
* that stored on the cursor. The cursor is returned to the ClusterCursorManager if it does not.
*/
void validateOperationSessionInfo(OperationContext* opCtx,
int64_t cursorId,
ClusterCursorManager::PinnedCursor* cursor) {
auto returnCursorGuard = makeGuard(
[cursor] { cursor->returnCursor(ClusterCursorManager::CursorState::NotExhausted); });
validateLSID(opCtx, cursorId, *cursor);
validateTxnNumber(opCtx, cursorId, *cursor);
returnCursorGuard.dismiss();
}
StatusWith<CursorResponse> ClusterFind::runGetMore(OperationContext* opCtx,
const GetMoreCommandRequest& cmd) {
auto cursorManager = Grid::get(opCtx)->getCursorManager();
auto authzSession = AuthorizationSession::get(opCtx->getClient());
auto authChecker = [&authzSession](UserNameIterator userNames) -> Status {
return authzSession->isCoauthorizedWith(userNames)
? Status::OK()
: Status(ErrorCodes::Unauthorized, "User not authorized to access cursor");
};
NamespaceString nss(cmd.getDbName(), cmd.getCollection());
int64_t cursorId = cmd.getCommandParameter();
auto pinnedCursor = cursorManager->checkOutCursor(nss, cursorId, opCtx, authChecker);
if (!pinnedCursor.isOK()) {
return pinnedCursor.getStatus();
}
invariant(cursorId == pinnedCursor.getValue().getCursorId());
validateOperationSessionInfo(opCtx, cursorId, &pinnedCursor.getValue());
// Ensure that the client still has the privileges to run the originating command.
if (!authzSession->isAuthorizedForPrivileges(
pinnedCursor.getValue()->getOriginatingPrivileges())) {
uasserted(ErrorCodes::Unauthorized,
str::stream() << "not authorized for getMore with cursor id " << cursorId);
}
// Set the originatingCommand object and the cursorID in CurOp.
{
CurOp::get(opCtx)->debug().nShards = pinnedCursor.getValue()->getNumRemotes();
CurOp::get(opCtx)->debug().cursorid = cursorId;
stdx::lock_guard<Client> lk(*opCtx->getClient());
CurOp::get(opCtx)->setOriginatingCommand_inlock(
pinnedCursor.getValue()->getOriginatingCommand());
CurOp::get(opCtx)->setGenericCursor_inlock(pinnedCursor.getValue().toGenericCursor());
}
// If the 'failGetMoreAfterCursorCheckout' failpoint is enabled, throw an exception with the
// specified 'errorCode' value, or ErrorCodes::InternalError if 'errorCode' is omitted.
failGetMoreAfterCursorCheckout.executeIf(
[](const BSONObj& data) {
auto errorCode = (data["errorCode"] ? data["errorCode"].safeNumberLong()
: ErrorCodes::InternalError);
uasserted(errorCode, "Hit the 'failGetMoreAfterCursorCheckout' failpoint");
},
[&opCtx, nss](const BSONObj& data) {
auto dataForFailCommand =
data.addField(BSON("failCommands" << BSON_ARRAY("getMore")).firstElement());
auto* getMoreCommand = CommandHelpers::findCommand("getMore");
return CommandHelpers::shouldActivateFailCommandFailPoint(
dataForFailCommand, nss, getMoreCommand, opCtx->getClient());
});
// If the 'waitAfterPinningCursorBeforeGetMoreBatch' fail point is enabled, set the 'msg'
// field of this operation's CurOp to signal that we've hit this point.
if (MONGO_unlikely(waitAfterPinningCursorBeforeGetMoreBatch.shouldFail())) {
CurOpFailpointHelpers::waitWhileFailPointEnabled(
&waitAfterPinningCursorBeforeGetMoreBatch,
opCtx,
"waitAfterPinningCursorBeforeGetMoreBatch");
}
auto opCtxSetupStatus =
setUpOperationContextStateForGetMore(opCtx, cmd, pinnedCursor.getValue());
if (!opCtxSetupStatus.isOK()) {
return opCtxSetupStatus;
}
std::vector<BSONObj> batch;
int bytesBuffered = 0;
long long batchSize = cmd.getBatchSize().value_or(0);
long long startingFrom = pinnedCursor.getValue()->getNumReturnedSoFar();
auto cursorState = ClusterCursorManager::CursorState::NotExhausted;
BSONObj postBatchResumeToken;
bool stashedResult = false;
// If the 'waitWithPinnedCursorDuringGetMoreBatch' fail point is enabled, set the 'msg'
// field of this operation's CurOp to signal that we've hit this point.
if (MONGO_unlikely(waitWithPinnedCursorDuringGetMoreBatch.shouldFail())) {
CurOpFailpointHelpers::waitWhileFailPointEnabled(&waitWithPinnedCursorDuringGetMoreBatch,
opCtx,
"waitWithPinnedCursorDuringGetMoreBatch");
}
while (!FindCommon::enoughForGetMore(batchSize, batch.size())) {
auto context = batch.empty()
? RouterExecStage::ExecContext::kGetMoreNoResultsYet
: RouterExecStage::ExecContext::kGetMoreWithAtLeastOneResultInBatch;
StatusWith<ClusterQueryResult> next =
Status{ErrorCodes::InternalError, "uninitialized cluster query result"};
try {
next = pinnedCursor.getValue()->next(context);
} catch (const ExceptionFor<ErrorCodes::CloseChangeStream>&) {
// This exception is thrown when a $changeStream stage encounters an event
// that invalidates the cursor. We should close the cursor and return without
// error.
cursorState = ClusterCursorManager::CursorState::Exhausted;
break;
} catch (const ExceptionFor<ErrorCodes::ChangeStreamInvalidated>& ex) {
// This exception is thrown when a change-stream cursor is invalidated. Set the PBRT
// to the resume token of the invalidating event, and mark the cursor response as
// invalidated. We always expect to have ExtraInfo for this error code.
const auto extraInfo = ex.extraInfo<ChangeStreamInvalidationInfo>();
tassert(
5493707, "Missing ChangeStreamInvalidationInfo on exception", extraInfo != nullptr);
postBatchResumeToken = extraInfo->getInvalidateResumeToken();
cursorState = ClusterCursorManager::CursorState::Exhausted;
break;
}
if (!next.isOK()) {
return next.getStatus();
}
if (next.getValue().isEOF()) {
// We reached end-of-stream. If the cursor is not tailable, then we mark it as
// exhausted. If it is tailable, usually we keep it open (i.e. "NotExhausted") even when
// we reach end-of-stream. However, if all the remote cursors are exhausted, there is no
// hope of returning data and thus we need to close the mongos cursor as well.
if (!pinnedCursor.getValue()->isTailable() ||
pinnedCursor.getValue()->remotesExhausted()) {
cursorState = ClusterCursorManager::CursorState::Exhausted;
}
break;
}
if (!FindCommon::haveSpaceForNext(
*next.getValue().getResult(), batch.size(), bytesBuffered)) {
pinnedCursor.getValue()->queueResult(*next.getValue().getResult());
stashedResult = true;
break;
}
// Add doc to the batch. Account for the space overhead associated with returning this doc
// inside a BSON array.
bytesBuffered +=
(next.getValue().getResult()->objsize() + kPerDocumentOverheadBytesUpperBound);
batch.push_back(std::move(*next.getValue().getResult()));
// Update the postBatchResumeToken. For non-$changeStream aggregations, this will be empty.
postBatchResumeToken = pinnedCursor.getValue()->getPostBatchResumeToken();
}
// If the cursor has been exhausted, we will communicate this by returning a CursorId of zero.
auto idToReturn =
(cursorState == ClusterCursorManager::CursorState::Exhausted ? CursorId(0) : cursorId);
// For empty batches, or in the case where the final result was added to the batch rather than
// being stashed, we update the PBRT here to ensure that it is the most recent available.
if (idToReturn && !stashedResult) {
postBatchResumeToken = pinnedCursor.getValue()->getPostBatchResumeToken();
}
const bool partialResultsReturned = pinnedCursor.getValue()->partialResultsReturned();
pinnedCursor.getValue()->setLeftoverMaxTimeMicros(opCtx->getRemainingMaxTimeMicros());
pinnedCursor.getValue()->incNBatches();
// Upon successful completion, transfer ownership of the cursor back to the cursor manager. If
// the cursor has been exhausted, the cursor manager will clean it up for us.
pinnedCursor.getValue().returnCursor(cursorState);
// Set nReturned and whether the cursor has been exhausted.
CurOp::get(opCtx)->debug().cursorExhausted = (idToReturn == 0);
CurOp::get(opCtx)->debug().nreturned = batch.size();
if (MONGO_unlikely(waitBeforeUnpinningOrDeletingCursorAfterGetMoreBatch.shouldFail())) {
CurOpFailpointHelpers::waitWhileFailPointEnabled(
&waitBeforeUnpinningOrDeletingCursorAfterGetMoreBatch,
opCtx,
"waitBeforeUnpinningOrDeletingCursorAfterGetMoreBatch");
}
auto atClusterTime = !opCtx->inMultiDocumentTransaction()
? repl::ReadConcernArgs::get(opCtx).getArgsAtClusterTime()
: boost::none;
return CursorResponse(nss,
idToReturn,
std::move(batch),
atClusterTime ? atClusterTime->asTimestamp()
: boost::optional<Timestamp>{},
startingFrom,
postBatchResumeToken,
boost::none,
boost::none,
partialResultsReturned);
}
} // namespace mongo
|