summaryrefslogtreecommitdiff
path: root/src/mongo/s/client/parallel.cpp
blob: 7903b2c48e5c6ce394829181d57db74a505872a3 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
/**
 *    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_LOG_DEFAULT_COMPONENT ::mongo::logger::LogComponent::kNetwork

#include "mongo/platform/basic.h"

#include "mongo/s/client/parallel.h"

#include "mongo/client/constants.h"
#include "mongo/client/dbclient_cursor.h"
#include "mongo/client/dbclient_rs.h"
#include "mongo/client/replica_set_monitor.h"
#include "mongo/db/bson/dotted_path_support.h"
#include "mongo/db/query/query_request.h"
#include "mongo/rpc/get_status_from_command_result.h"
#include "mongo/s/catalog_cache.h"
#include "mongo/s/client/shard_connection.h"
#include "mongo/s/client/shard_registry.h"
#include "mongo/s/grid.h"
#include "mongo/util/log.h"
#include "mongo/util/net/socket_exception.h"

namespace mongo {

using std::shared_ptr;
using std::map;
using std::set;
using std::string;
using std::vector;

namespace dps = ::mongo::dotted_path_support;

namespace {

/**
 * Throws an exception wrapping the error document in this cursor when the error flag is set.
 */
void throwCursorError(DBClientCursor* cursor) {
    verify(cursor);

    if (cursor->hasResultFlag(ResultFlag_ErrSet)) {
        uassertStatusOK(getStatusFromCommandResult(cursor->next()));
    }
}

}  // namespace

struct ParallelConnectionState {
    ParallelConnectionState() : count(0), done(false) {}

    std::string toString() const;

    BSONObj toBSON() const;

    // Please do not reorder. cursor destructor can use conn.
    // On a related note, never attempt to cleanup these pointers manually.
    std::shared_ptr<ShardConnection> conn;
    std::shared_ptr<DBClientCursor> cursor;

    // Version information
    std::shared_ptr<ChunkManager> manager;
    std::shared_ptr<Shard> primary;

    // Cursor status information
    long long count;
    bool done;
};

struct ParallelConnectionMetadata {
    ParallelConnectionMetadata()
        : retryNext(false), initialized(false), finished(false), completed(false), errored(false) {}

    ~ParallelConnectionMetadata() {
        cleanup(true);
    }

    void cleanup(bool full = true);

    std::shared_ptr<ParallelConnectionState> pcState;

    bool retryNext;

    bool initialized;
    bool finished;
    bool completed;

    bool errored;

    BSONObj toBSON() const;

    std::string toString() const {
        return str::stream() << "PCMData : " << toBSON();
    }
};

/**
 * Helper class to manage ownership of opened cursors while merging results.
 *
 * TODO:  Choose one set of ownership semantics so that this isn't needed - merge sort via mapreduce
 * is the main issue since it has no metadata and this holder owns the cursors.
 */
class DBClientCursorHolder {
public:
    DBClientCursorHolder() = default;

    void reset(DBClientCursor* cursor, ParallelConnectionMetadata* pcmData) {
        _cursor.reset(cursor);
        _pcmData.reset(pcmData);
    }

    DBClientCursor* get() {
        return _cursor.get();
    }

    ParallelConnectionMetadata* getMData() {
        return _pcmData.get();
    }

    void release() {
        _cursor.release();
        _pcmData.release();
    }

private:
    std::unique_ptr<DBClientCursor> _cursor;
    std::unique_ptr<ParallelConnectionMetadata> _pcmData;
};

// --------  ParallelSortClusteredCursor -----------

ParallelSortClusteredCursor::ParallelSortClusteredCursor(const QuerySpec& qSpec,
                                                         const CommandInfo& cInfo)
    : _qSpec(qSpec), _cInfo(cInfo), _totalTries(0) {
    _done = false;
    _didInit = false;

    _finishCons();
}

ParallelSortClusteredCursor::ParallelSortClusteredCursor(const set<string>& servers,
                                                         const string& ns,
                                                         const Query& q,
                                                         int options,
                                                         const BSONObj& fields)
    : _servers(servers) {
    _sortKey = q.getSort().copy();
    _needToSkip = 0;

    _done = false;
    _didInit = false;

    // Populate legacy fields
    _ns = ns;
    _query = q.obj.getOwned();
    _options = options;
    _fields = fields.getOwned();
    _batchSize = 0;

    _finishCons();
}

ParallelSortClusteredCursor::~ParallelSortClusteredCursor() {
    // WARNING: Commands (in particular M/R) connect via _oldInit() directly to shards
    bool isDirectShardCursor = _cursorMap.empty();

    // Non-direct shard cursors are owned by the _cursorMap, so we release
    // them in the array here.  Direct shard cursors clean themselves.
    if (!isDirectShardCursor) {
        for (int i = 0; i < _numServers; i++)
            _cursors[i].release();
    }

    delete[] _cursors;
    _cursors = 0;

    // Clear out our metadata after removing legacy cursor data
    _cursorMap.clear();

    // Just to be sure
    _done = true;
}

void ParallelSortClusteredCursor::init(OperationContext* opCtx) {
    if (_didInit)
        return;
    _didInit = true;

    if (!_qSpec.isEmpty()) {
        fullInit(opCtx);
    } else {
        // You can only get here by using the legacy constructor
        // TODO: Eliminate this
        _oldInit(opCtx);
    }
}

void ParallelSortClusteredCursor::_finishCons() {
    _numServers = _servers.size();
    _lastFrom = 0;
    _cursors = 0;

    if (!_qSpec.isEmpty()) {
        _needToSkip = _qSpec.ntoskip();
        _cursors = 0;
        _sortKey = _qSpec.sort();
        _fields = _qSpec.fields();
    }

    // Partition sort key fields into (a) text meta fields and (b) all other fields.
    set<string> textMetaSortKeyFields;
    set<string> normalSortKeyFields;

    // Transform _sortKey fields {a:{$meta:"textScore"}} into {a:-1}, in order to apply the
    // merge sort for text metadata in the correct direction.
    BSONObjBuilder transformedSortKeyBuilder;

    BSONObjIterator sortKeyIt(_sortKey);
    while (sortKeyIt.more()) {
        BSONElement e = sortKeyIt.next();
        if (QueryRequest::isTextScoreMeta(e)) {
            textMetaSortKeyFields.insert(e.fieldName());
            transformedSortKeyBuilder.append(e.fieldName(), -1);
        } else {
            normalSortKeyFields.insert(e.fieldName());
            transformedSortKeyBuilder.append(e);
        }
    }
    _sortKey = transformedSortKeyBuilder.obj();

    // Verify that that all text metadata sort fields are in the projection.  For all other sort
    // fields, copy them into the projection if they are missing (and if projection is
    // negative).
    if (!_sortKey.isEmpty() && !_fields.isEmpty()) {
        BSONObjBuilder b;
        bool isNegative = false;
        {
            BSONObjIterator i(_fields);
            while (i.more()) {
                BSONElement e = i.next();
                b.append(e);

                string fieldName = e.fieldName();

                if (QueryRequest::isTextScoreMeta(e)) {
                    textMetaSortKeyFields.erase(fieldName);
                } else {
                    // exact field
                    bool found = normalSortKeyFields.erase(fieldName);

                    // subfields
                    set<string>::const_iterator begin =
                        normalSortKeyFields.lower_bound(fieldName + ".\x00");
                    set<string>::const_iterator end =
                        normalSortKeyFields.lower_bound(fieldName + ".\xFF");
                    normalSortKeyFields.erase(begin, end);

                    if (!e.trueValue()) {
                        uassert(13431,
                                "have to have sort key in projection and removing it",
                                !found && begin == end);
                    } else if (!e.isABSONObj()) {
                        isNegative = true;
                    }
                }
            }
        }

        if (isNegative) {
            for (set<string>::const_iterator it(normalSortKeyFields.begin()),
                 end(normalSortKeyFields.end());
                 it != end;
                 ++it) {
                b.append(*it, 1);
            }
        }

        _fields = b.obj();
    }

    if (!_qSpec.isEmpty()) {
        _qSpec.setFields(_fields);
    }

    uassert(
        17306, "have to have all text meta sort keys in projection", textMetaSortKeyFields.empty());
}

void ParallelSortClusteredCursor::fullInit(OperationContext* opCtx) {
    startInit(opCtx);
    finishInit(opCtx);
}

void ParallelSortClusteredCursor::_markStaleNS(const NamespaceString& staleNS,
                                               const StaleConfigException& e) {
    if (_staleNSMap.find(staleNS.ns()) == _staleNSMap.end()) {
        _staleNSMap[staleNS.ns()] = 1;
    }

    const int tries = ++_staleNSMap[staleNS.ns()];

    if (tries >= 5) {
        uassertStatusOK(e.toStatus("too many retries of stale version info"));
    }
}

void ParallelSortClusteredCursor::setupVersionAndHandleSlaveOk(
    OperationContext* opCtx,
    std::shared_ptr<ParallelConnectionState> state,
    const ShardId& shardId,
    std::shared_ptr<Shard> primary,
    const NamespaceString& ns,
    const string& vinfo,
    std::shared_ptr<ChunkManager> manager) {
    if (manager) {
        state->manager = manager;
    } else if (primary) {
        state->primary = primary;
    }

    verify(!primary || shardId == primary->getId());

    // Setup conn
    if (!state->conn) {
        const auto shard =
            uassertStatusOK(Grid::get(opCtx)->shardRegistry()->getShard(opCtx, shardId));
        state->conn.reset(new ShardConnection(opCtx, shard->getConnString(), ns.ns(), manager));
    }

    const DBClientBase* rawConn = state->conn->getRawConn();
    bool allowShardVersionFailure = rawConn->type() == ConnectionString::SET &&
        DBClientReplicaSet::isSecondaryQuery(_qSpec.ns(), _qSpec.query(), _qSpec.options());

    // Skip shard version checking if primary is known to be down.
    if (allowShardVersionFailure) {
        const DBClientReplicaSet* replConn = dynamic_cast<const DBClientReplicaSet*>(rawConn);
        invariant(replConn);
        ReplicaSetMonitorPtr rsMonitor = ReplicaSetMonitor::get(replConn->getSetName());
        uassert(16388,
                str::stream() << "cannot access unknown replica set: " << replConn->getSetName(),
                rsMonitor != nullptr);
        if (!rsMonitor->isKnownToHaveGoodPrimary()) {
            state->conn->donotCheckVersion();

            // A side effect of this short circuiting is the mongos will not be able figure out
            // that the primary is now up on it's own and has to rely on other threads to refresh
            // node states.

            static Occasionally sampler;
            if (sampler.tick()) {
                const DBClientReplicaSet* repl = dynamic_cast<const DBClientReplicaSet*>(rawConn);
                dassert(repl);
                warning() << "Primary for " << repl->getServerAddress()
                          << " was down before, bypassing setShardVersion."
                          << " The local replica set view and targeting may be stale.";
            }

            return;
        }
    }

    try {
        if (state->conn->setVersion()) {
            LOG(2) << "pcursor: needed to set remote version on connection to value "
                   << "compatible with " << vinfo;
        }
    } catch (const DBException& dbExcep) {
        auto errCode = dbExcep.code();
        if (allowShardVersionFailure &&
            (ErrorCodes::isNotMasterError(errCode) || ErrorCodes::isNetworkError(errCode) ||
             errCode == ErrorCodes::FailedToSatisfyReadPreference)) {
            // It's okay if we don't set the version when talking to a secondary, we can
            // be stale in any case.

            static Occasionally sampler;
            if (sampler.tick()) {
                const DBClientReplicaSet* repl =
                    dynamic_cast<const DBClientReplicaSet*>(state->conn->getRawConn());
                dassert(repl);
                warning() << "Cannot contact primary for " << repl->getServerAddress()
                          << " to check shard version."
                          << " The local replica set view and targeting may be stale.";
            }
        } else {
            throw;
        }
    }
}

void ParallelSortClusteredCursor::startInit(OperationContext* opCtx) {
    const bool returnPartial = (_qSpec.options() & QueryOption_PartialResults);
    const NamespaceString nss(!_cInfo.isEmpty() ? _cInfo.versionedNS : _qSpec.ns());

    string prefix;
    if (MONGO_unlikely(shouldLog(logger::LogSeverity::Debug(2)))) {
        if (_totalTries > 0) {
            prefix = str::stream() << "retrying (" << _totalTries << " tries)";
        } else {
            prefix = "creating";
        }
    }
    LOG(2) << "pcursor: " << prefix << " pcursor over " << _qSpec << " and " << _cInfo;

    shared_ptr<ChunkManager> manager;
    shared_ptr<Shard> primary;

    {
        auto routingInfoStatus =
            Grid::get(opCtx)->catalogCache()->getCollectionRoutingInfo(opCtx, nss);
        if (routingInfoStatus != ErrorCodes::NamespaceNotFound) {
            auto routingInfo = uassertStatusOK(std::move(routingInfoStatus));
            manager = routingInfo.cm();
            // ParallelSortClusteredCursor has two states - either !cm && primary, which means
            // unsharded collection, or cm && !primary, which means sharded collection.
            if (!manager) {
                primary = routingInfo.db().primary();
            }
        }
    }

    set<ShardId> shardIds;
    string vinfo;

    if (manager) {
        if (MONGO_unlikely(shouldLog(logger::LogSeverity::Debug(2)))) {
            vinfo = str::stream() << "[" << manager->getns().ns() << " @ "
                                  << manager->getVersion().toString() << "]";
        }

        manager->getShardIdsForQuery(opCtx,
                                     !_cInfo.isEmpty() ? _cInfo.cmdFilter : _qSpec.filter(),
                                     !_cInfo.isEmpty() ? _cInfo.cmdCollation : BSONObj(),
                                     &shardIds);
    } else if (primary) {
        if (MONGO_unlikely(shouldLog(logger::LogSeverity::Debug(2)))) {
            vinfo = str::stream() << "[unsharded @ " << primary->toString() << "]";
        }

        shardIds.insert(primary->getId());
    }

    // Close all cursors on extra shards first, as these will be invalid
    for (auto& cmEntry : _cursorMap) {
        const auto& shardId = cmEntry.first;

        if (shardIds.find(shardId) == shardIds.end()) {
            LOG(2) << "pcursor: closing cursor on shard " << shardId
                   << " as the connection is no longer required by " << vinfo;

            cmEntry.second.cleanup(true);
        }
    }

    LOG(2) << "pcursor: initializing over " << shardIds.size() << " shards required by " << vinfo;

    // Don't retry indefinitely for whatever reason
    _totalTries++;
    uassert(15986, "too many retries in total", _totalTries < 10);

    for (const ShardId& shardId : shardIds) {
        auto& mdata = _cursorMap[shardId];

        LOG(2) << "pcursor: initializing on shard " << shardId << ", current connection state is "
               << mdata.toBSON();

        // This may be the first time connecting to this shard, if so we can get an error here
        try {
            if (mdata.initialized) {
                invariant(mdata.pcState);

                auto state = mdata.pcState;

                bool compatiblePrimary = true;
                bool compatibleManager = true;

                if (primary && !state->primary)
                    warning() << "Collection becoming unsharded detected";
                if (manager && !state->manager)
                    warning() << "Collection becoming sharded detected";
                if (primary && state->primary && primary != state->primary)
                    warning() << "Weird shift of primary detected";

                compatiblePrimary = primary && state->primary && primary == state->primary;
                compatibleManager =
                    manager && state->manager && manager->compatibleWith(*state->manager, shardId);

                if (compatiblePrimary || compatibleManager) {
                    // If we're compatible, don't need to retry unless forced
                    if (!mdata.retryNext)
                        continue;
                    // Do partial cleanup
                    mdata.cleanup(false);
                } else {
                    // Force total cleanup of connection if no longer compatible
                    mdata.cleanup(true);
                }
            } else {
                // Cleanup connection if we're not yet initialized
                mdata.cleanup(false);
            }

            mdata.pcState = std::make_shared<ParallelConnectionState>();
            auto state = mdata.pcState;

            setupVersionAndHandleSlaveOk(opCtx, state, shardId, primary, nss, vinfo, manager);

            const string& ns = _qSpec.ns();

            // Setup cursor
            if (!state->cursor) {
                //
                // Here we decide whether to split the query limits up for multiple shards.
                // NOTE: There's a subtle issue here, in that it's possible we target a single
                // shard first, but are stale, and then target multiple shards, or vice-versa.
                // In both these cases, we won't re-use the old cursor created here, since the
                // shard version must have changed on the single shard between queries.
                //

                if (shardIds.size() > 1) {
                    // Query limits split for multiple shards

                    state->cursor.reset(new DBClientCursor(
                        state->conn->get(),
                        NamespaceString(ns),
                        _qSpec.query(),
                        isCommand() ? 1 : 0,  // nToReturn (0 if query indicates multi)
                        0,                    // nToSkip
                        // Does this need to be a ptr?
                        _qSpec.fields().isEmpty() ? 0 : _qSpec.fieldsData(),  // fieldsToReturn
                        _qSpec.options(),                                     // options
                        // NtoReturn is weird.
                        // If zero, it means use default size, so we do that for all cursors
                        // If positive, it's the batch size (we don't want this cursor limiting
                        // results), that's done at a higher level
                        // If negative, it's the batch size, but we don't create a cursor - so we
                        // don't want to create a child cursor either.
                        // Either way, if non-zero, we want to pull back the batch size + the skip
                        // amount as quickly as possible.  Potentially, for a cursor on a single
                        // shard or if we keep better track of chunks, we can actually add the skip
                        // value into the cursor and/or make some assumptions about the return value
                        // size ( (batch size + skip amount) / num_servers ).
                        _qSpec.ntoreturn() == 0 ? 0 : (_qSpec.ntoreturn() > 0
                                                           ? _qSpec.ntoreturn() + _qSpec.ntoskip()
                                                           : _qSpec.ntoreturn() -
                                                               _qSpec.ntoskip())));  // batchSize
                } else {
                    // Single shard query

                    state->cursor.reset(new DBClientCursor(
                        state->conn->get(),
                        NamespaceString(ns),
                        _qSpec.query(),
                        _qSpec.ntoreturn(),  // nToReturn
                        _qSpec.ntoskip(),    // nToSkip
                        // Does this need to be a ptr?
                        _qSpec.fields().isEmpty() ? 0 : _qSpec.fieldsData(),  // fieldsToReturn
                        _qSpec.options(),                                     // options
                        0));                                                  // batchSize
                }
            }

            bool lazyInit = state->conn->get()->lazySupported();
            if (lazyInit) {
                // Need to keep track if this is a second or third try for replica sets
                state->cursor->initLazy(mdata.retryNext);
                mdata.retryNext = false;
                mdata.initialized = true;
            } else {
                bool success = state->cursor->init();

                // Without full initialization, throw an exception
                uassert(15987,
                        str::stream() << "could not fully initialize cursor on shard " << shardId
                                      << ", current connection state is "
                                      << mdata.toBSON().toString(),
                        success);

                mdata.retryNext = false;
                mdata.initialized = true;
                mdata.finished = true;
            }

            LOG(2) << "pcursor: initialized " << (isCommand() ? "command " : "query ")
                   << (lazyInit ? "(lazily) " : "(full) ") << "on shard " << shardId
                   << ", current connection state is " << mdata.toBSON();
        } catch (StaleConfigException& e) {
            // Our version isn't compatible with the current version anymore on at least one shard,
            // need to retry immediately
            NamespaceString staleNS(e->getNss());
            _markStaleNS(staleNS, e);

            Grid::get(opCtx)->catalogCache()->invalidateShardedCollection(staleNS);

            LOG(1) << "stale config of ns " << staleNS << " during initialization, will retry"
                   << causedBy(redact(e));

            // This is somewhat strange
            if (staleNS != nss) {
                warning() << "versioned ns " << nss.ns() << " doesn't match stale config namespace "
                          << staleNS;
            }

            // Restart with new chunk manager
            startInit(opCtx);
            return;
        } catch (NetworkException& e) {
            warning() << "socket exception when initializing on " << shardId
                      << ", current connection state is " << mdata.toBSON() << causedBy(redact(e));
            mdata.errored = true;
            if (returnPartial) {
                mdata.cleanup(true);
                continue;
            }
            throw;
        } catch (DBException& e) {
            warning() << "db exception when initializing on " << shardId
                      << ", current connection state is " << mdata.toBSON() << causedBy(redact(e));
            mdata.errored = true;
            if (returnPartial && e.code() == 15925 /* From above! */) {
                mdata.cleanup(true);
                continue;
            }
            throw;
        } catch (std::exception& e) {
            warning() << "exception when initializing on " << shardId
                      << ", current connection state is " << mdata.toBSON() << causedBy(e);
            mdata.errored = true;
            throw;
        } catch (...) {
            warning() << "unknown exception when initializing on " << shardId
                      << ", current connection state is " << mdata.toBSON();
            mdata.errored = true;
            throw;
        }
    }

    // Sanity check final init'ed connections
    for (const auto& cmEntry : _cursorMap) {
        const auto& shardId = cmEntry.first;
        const auto& mdata = cmEntry.second;

        if (!mdata.pcState) {
            continue;
        }

        // Make sure all state is in shards
        invariant(shardIds.find(shardId) != shardIds.end());
        invariant(mdata.initialized == true);

        if (!mdata.completed) {
            invariant(mdata.pcState->conn->ok());
        }

        invariant(mdata.pcState->cursor);
        invariant(mdata.pcState->primary || mdata.pcState->manager);
        invariant(!mdata.retryNext);

        if (mdata.completed) {
            invariant(mdata.finished);
        }

        if (mdata.finished) {
            invariant(mdata.initialized);
        }

        if (!returnPartial) {
            invariant(mdata.initialized);
        }
    }
}

void ParallelSortClusteredCursor::finishInit(OperationContext* opCtx) {
    bool returnPartial = (_qSpec.options() & QueryOption_PartialResults);
    bool specialVersion = _cInfo.versionedNS.size() > 0;
    string ns = specialVersion ? _cInfo.versionedNS : _qSpec.ns();

    bool retry = false;
    map<string, StaleConfigException> staleNSExceptions;

    LOG(2) << "pcursor: finishing over " << _cursorMap.size() << " shards";

    for (auto& cmEntry : _cursorMap) {
        const auto& shardId = cmEntry.first;
        auto& mdata = cmEntry.second;

        LOG(2) << "pcursor: finishing on shard " << shardId << ", current connection state is "
               << mdata.toBSON();

        // Ignore empty conns for now
        if (!mdata.pcState)
            continue;

        auto state = mdata.pcState;

        try {
            // Sanity checks
            if (!mdata.completed)
                verify(state->conn && state->conn->ok());
            verify(state->cursor);
            verify(state->manager || state->primary);
            verify(!state->manager || !state->primary);


            // If we weren't init'ing lazily, ignore this
            if (!mdata.finished) {
                mdata.finished = true;

                // Mark the cursor as non-retry by default
                mdata.retryNext = false;

                if (!state->cursor->initLazyFinish(mdata.retryNext)) {
                    if (!mdata.retryNext) {
                        uassert(15988, "error querying server", false);
                    } else {
                        retry = true;
                        continue;
                    }
                }

                mdata.completed = false;
            }

            if (!mdata.completed) {
                mdata.completed = true;

                // Make sure we didn't get an error we should rethrow
                // TODO : Refactor this to something better
                throwCursorStale(state->cursor.get());
                throwCursorError(state->cursor.get());

                // Finalize state
                state->cursor->attach(state->conn.get());  // Closes connection for us

                LOG(2) << "pcursor: finished on shard " << shardId
                       << ", current connection state is " << mdata.toBSON();
            }
        } catch (StaleConfigException& e) {
            retry = true;

            std::string staleNS = e->getNss().ns();

            // Will retry all at once
            staleNSExceptions.emplace(staleNS, e);

            // Fully clear this cursor, as it needs to be re-established
            mdata.cleanup(true);
            continue;
        } catch (NetworkException& e) {
            warning() << "socket exception when finishing on " << shardId
                      << ", current connection state is " << mdata.toBSON() << causedBy(redact(e));
            mdata.errored = true;
            if (returnPartial) {
                mdata.cleanup(true);
                continue;
            }
            throw;
        } catch (DBException& e) {
            // NOTE: RECV() WILL NOT THROW A SOCKET EXCEPTION - WE GET THIS AS ERROR 15988 FROM
            // ABOVE
            if (e.code() == 15988) {
                warning() << "exception when receiving data from " << shardId
                          << ", current connection state is " << mdata.toBSON()
                          << causedBy(redact(e));

                mdata.errored = true;
                if (returnPartial) {
                    mdata.cleanup(true);
                    continue;
                }
                throw;
            } else {
                // the InvalidBSON exception indicates that the BSON is malformed ->
                // don't print/call "mdata.toBSON()" to avoid unexpected errors e.g. a segfault
                if (e.code() == ErrorCodes::InvalidBSON)
                    warning() << "bson is malformed :: db exception when finishing on " << shardId
                              << causedBy(redact(e));
                else
                    warning() << "db exception when finishing on " << shardId
                              << ", current connection state is " << mdata.toBSON()
                              << causedBy(redact(e));
                mdata.errored = true;
                throw;
            }
        } catch (std::exception& e) {
            warning() << "exception when finishing on " << shardId
                      << ", current connection state is " << mdata.toBSON() << causedBy(e);
            mdata.errored = true;
            throw;
        } catch (...) {
            warning() << "unknown exception when finishing on " << shardId
                      << ", current connection state is " << mdata.toBSON();
            mdata.errored = true;
            throw;
        }
    }

    // Retry logic for single refresh of namespaces / retry init'ing connections
    if (retry) {
        // Refresh stale namespaces
        if (staleNSExceptions.size()) {
            for (const auto& exEntry : staleNSExceptions) {
                const NamespaceString staleNS(exEntry.first);
                const StaleConfigException& ex = exEntry.second;

                _markStaleNS(staleNS, ex);
                Grid::get(opCtx)->catalogCache()->invalidateShardedCollection(staleNS);

                LOG(1) << "stale config of ns " << staleNS << " on finishing query, will retry"
                       << causedBy(redact(ex));

                // This is somewhat strange
                if (staleNS.ns() != ns) {
                    warning() << "versioned ns " << ns << " doesn't match stale config namespace "
                              << staleNS;
                }
            }
        }

        // Re-establish connections we need to
        startInit(opCtx);
        finishInit(opCtx);
        return;
    }

    // Sanity check and clean final connections
    for (auto i = _cursorMap.begin(); i != _cursorMap.end();) {
        auto& mdata = i->second;

        // Erase empty stuff
        if (!mdata.pcState) {
            log() << "PCursor erasing empty state " << mdata.toBSON();
            _cursorMap.erase(i++);
            continue;
        } else {
            ++i;
        }

        // Make sure all state is in shards
        verify(mdata.initialized == true);
        verify(mdata.finished == true);
        verify(mdata.completed == true);
        verify(!mdata.pcState->conn->ok());
        verify(mdata.pcState->cursor);
        verify(mdata.pcState->primary || mdata.pcState->manager);
    }

    // TODO : More cleanup of metadata?

    // LEGACY STUFF NOW

    _cursors = new DBClientCursorHolder[_cursorMap.size()];

    // Put the cursors in the legacy format
    int index = 0;
    for (auto& cmEntry : _cursorMap) {
        const auto& shardId = cmEntry.first;
        auto& mdata = cmEntry.second;

        _cursors[index].reset(mdata.pcState->cursor.get(), &mdata);

        const auto shard =
            uassertStatusOK(Grid::get(opCtx)->shardRegistry()->getShard(opCtx, shardId));
        _servers.insert(shard->getConnString().toString());

        index++;
    }

    _numServers = _cursorMap.size();
}

void ParallelSortClusteredCursor::getQueryShardIds(set<ShardId>& shardIds) const {
    std::transform(_cursorMap.begin(),
                   _cursorMap.end(),
                   std::inserter(shardIds, shardIds.end()),
                   [](const ShardCursorsMap::value_type& pair) { return pair.first; });
}

std::shared_ptr<DBClientCursor> ParallelSortClusteredCursor::getShardCursor(
    const ShardId& shardId) const {
    auto it = _cursorMap.find(shardId);
    if (it == _cursorMap.end()) {
        return nullptr;
    }

    return it->second.pcState->cursor;
}

// DEPRECATED (but still used by map/reduce)
void ParallelSortClusteredCursor::_oldInit(OperationContext* opCtx) {
    // make sure we're not already initialized
    verify(!_cursors);
    _cursors = new DBClientCursorHolder[_numServers];

    bool returnPartial = (_options & QueryOption_PartialResults);

    vector<string> serverHosts(_servers.begin(), _servers.end());
    set<int> retryQueries;
    int finishedQueries = 0;

    vector<shared_ptr<ShardConnection>> conns;
    vector<string> servers;

    // Since we may get all sorts of errors, record them all as they come and throw them later if
    // necessary
    vector<string> staleConfigExs;
    vector<string> socketExs;
    vector<string> otherExs;
    bool allConfigStale = false;

    int retries = -1;

    // Loop through all the queries until we've finished or gotten a socket exception on all of them
    // We break early for non-socket exceptions, and socket exceptions if we aren't returning
    // partial results
    do {
        retries++;

        bool firstPass = retryQueries.size() == 0;

        if (!firstPass) {
            log() << "retrying " << (returnPartial ? "(partial) " : "")
                  << "parallel connection to ";
            for (set<int>::const_iterator it = retryQueries.begin(); it != retryQueries.end();
                 ++it) {
                log() << serverHosts[*it] << ", ";
            }
            log() << finishedQueries << " finished queries.";
        }

        size_t num = 0;
        for (vector<string>::const_iterator it = serverHosts.begin(); it != serverHosts.end();
             ++it) {
            size_t i = num++;

            const string& serverHost = *it;

            // If we're not retrying this cursor on later passes, continue
            if (!firstPass && retryQueries.find(i) == retryQueries.end())
                continue;

            const string errLoc = " @ " + serverHost;

            if (firstPass) {
                // This may be the first time connecting to this shard, if so we can get an error
                // here
                try {
                    conns.push_back(shared_ptr<ShardConnection>(new ShardConnection(
                        opCtx, uassertStatusOK(ConnectionString::parse(serverHost)), _ns)));
                } catch (std::exception& e) {
                    socketExs.push_back(e.what() + errLoc);
                    if (!returnPartial) {
                        num--;
                        break;
                    }

                    conns.push_back(shared_ptr<ShardConnection>());
                    continue;
                }

                servers.push_back(serverHost);
            }

            if (conns[i]->setVersion()) {
                conns[i]->done();

                // Version is zero b/c this is deprecated codepath
                staleConfigExs.push_back(str::stream() << "stale config detected for " << _ns
                                                       << " in ParallelCursor::_init "
                                                       << errLoc);
                break;
            }

            LOG(5) << "ParallelSortClusteredCursor::init server:" << serverHost << " ns:" << _ns
                   << " query:" << redact(_query) << " fields:" << redact(_fields)
                   << " options: " << _options;

            if (!_cursors[i].get())
                _cursors[i].reset(
                    new DBClientCursor(conns[i]->get(),
                                       NamespaceString(_ns),
                                       _query,
                                       0,                                 // nToReturn
                                       0,                                 // nToSkip
                                       _fields.isEmpty() ? 0 : &_fields,  // fieldsToReturn
                                       _options,
                                       _batchSize == 0 ? 0 : _batchSize + _needToSkip  // batchSize
                                       ),
                    NULL);

            try {
                _cursors[i].get()->initLazy(!firstPass);
            } catch (NetworkException& e) {
                socketExs.push_back(e.what() + errLoc);
                _cursors[i].reset(NULL, NULL);
                conns[i]->done();
                if (!returnPartial)
                    break;
            } catch (std::exception& e) {
                otherExs.push_back(e.what() + errLoc);
                _cursors[i].reset(NULL, NULL);
                conns[i]->done();
                break;
            }
        }

        // Go through all the potentially started cursors and finish initializing them or log any
        // errors and potentially retry
        // TODO:  Better error classification would make this easier, errors are indicated in all
        // sorts of ways here that we need to trap.
        for (size_t i = 0; i < num; i++) {
            const string errLoc = " @ " + serverHosts[i];

            if (!_cursors[i].get() || (!firstPass && retryQueries.find(i) == retryQueries.end())) {
                if (conns[i])
                    conns[i].get()->done();
                continue;
            }

            verify(conns[i]);
            retryQueries.erase(i);

            bool retry = false;

            try {
                if (!_cursors[i].get()->initLazyFinish(retry)) {
                    warning() << "invalid result from " << conns[i]->getHost()
                              << (retry ? ", retrying" : "");
                    _cursors[i].reset(NULL, NULL);

                    if (!retry) {
                        socketExs.push_back(str::stream() << "error querying server: "
                                                          << servers[i]);
                        conns[i]->done();
                    } else {
                        retryQueries.insert(i);
                    }

                    continue;
                }
            } catch (StaleConfigException& e) {
                // Our stored configuration data is actually stale, we need to reload it
                // when we throw our exception
                allConfigStale = true;

                staleConfigExs.push_back(
                    (string) "stale config detected when receiving response for " + e.toString() +
                    errLoc);
                _cursors[i].reset(NULL, NULL);
                conns[i]->done();
                continue;
            } catch (NetworkException& e) {
                socketExs.push_back(e.what() + errLoc);
                _cursors[i].reset(NULL, NULL);
                conns[i]->done();
                continue;
            } catch (std::exception& e) {
                otherExs.push_back(e.what() + errLoc);
                _cursors[i].reset(NULL, NULL);
                conns[i]->done();
                continue;
            }

            try {
                _cursors[i].get()->attach(conns[i].get());  // this calls done on conn
                // Rethrow stale config or other errors
                throwCursorStale(_cursors[i].get());
                throwCursorError(_cursors[i].get());

                finishedQueries++;
            } catch (StaleConfigException& e) {
                // Our stored configuration data is actually stale, we need to reload it
                // when we throw our exception
                allConfigStale = true;

                staleConfigExs.push_back((string) "stale config detected for " + e.toString() +
                                         errLoc);
                _cursors[i].reset(NULL, NULL);
                conns[i]->done();
                continue;
            } catch (std::exception& e) {
                otherExs.push_back(e.what() + errLoc);
                _cursors[i].reset(NULL, NULL);
                conns[i]->done();
                continue;
            }
        }

        // Don't exceed our max retries, should not happen
        verify(retries < 5);
    } while (retryQueries.size() > 0 /* something to retry */ &&
             (socketExs.size() == 0 || returnPartial) /* no conn issues */ &&
             staleConfigExs.size() == 0 /* no config issues */ &&
             otherExs.size() == 0 /* no other issues */);

    // Assert that our conns are all closed!
    for (vector<shared_ptr<ShardConnection>>::iterator i = conns.begin(); i < conns.end(); ++i) {
        verify(!(*i) || !(*i)->ok());
    }

    // Handle errors we got during initialization.
    // If we're returning partial results, we can ignore socketExs, but nothing else
    // Log a warning in any case, so we don't lose these messages
    bool throwException = (socketExs.size() > 0 && !returnPartial) || staleConfigExs.size() > 0 ||
        otherExs.size() > 0;

    if (socketExs.size() > 0 || staleConfigExs.size() > 0 || otherExs.size() > 0) {
        vector<string> errMsgs;

        errMsgs.insert(errMsgs.end(), staleConfigExs.begin(), staleConfigExs.end());
        errMsgs.insert(errMsgs.end(), otherExs.begin(), otherExs.end());
        errMsgs.insert(errMsgs.end(), socketExs.begin(), socketExs.end());

        std::stringstream errMsg;
        errMsg << "could not initialize cursor across all shards because : ";
        for (vector<string>::iterator i = errMsgs.begin(); i != errMsgs.end(); i++) {
            if (i != errMsgs.begin())
                errMsg << " :: and :: ";
            errMsg << *i;
        }

        if (throwException && staleConfigExs.size() > 0) {
            // Version is zero b/c this is deprecated codepath
            uasserted(StaleConfigInfo(NamespaceString(_ns),
                                      ChunkVersion(0, 0, OID()),
                                      ChunkVersion(0, 0, OID())),
                      errMsg.str());
        } else if (throwException) {
            uasserted(14827, errMsg.str());
        } else {
            warning() << redact(errMsg.str());
        }
    }

    if (retries > 0)
        log() << "successfully finished parallel query after " << retries << " retries";
}

bool ParallelSortClusteredCursor::more() {
    if (_needToSkip > 0) {
        int n = _needToSkip;
        _needToSkip = 0;

        while (n > 0 && more()) {
            next();
            n--;
        }

        _needToSkip = n;
    }

    for (int i = 0; i < _numServers; i++) {
        if (_cursors[i].get() && _cursors[i].get()->more())
            return true;
    }

    return false;
}

BSONObj ParallelSortClusteredCursor::next() {
    BSONObj best = BSONObj();
    int bestFrom = -1;

    for (int j = 0; j < _numServers; j++) {
        // Iterate _numServers times, starting one past the last server we used.
        // This means we actually start at server #1, not #0, but shouldn't matter

        int i = (j + _lastFrom + 1) % _numServers;

        // Check to see if the cursor is finished
        if (!_cursors[i].get() || !_cursors[i].get()->more()) {
            if (_cursors[i].getMData())
                _cursors[i].getMData()->pcState->done = true;
            continue;
        }

        // We know we have at least one result in this cursor
        BSONObj me = _cursors[i].get()->peekFirst();

        // If this is the first non-empty cursor, save the result as best
        if (bestFrom < 0) {
            best = me;
            bestFrom = i;
            if (_sortKey.isEmpty())
                break;
            continue;
        }

        // Otherwise compare the result to the current best result
        int comp = dps::compareObjectsAccordingToSort(best, me, _sortKey, true);
        if (comp < 0)
            continue;

        best = me;
        bestFrom = i;
    }

    _lastFrom = bestFrom;

    uassert(10019, "no more elements", bestFrom >= 0);
    _cursors[bestFrom].get()->next();

    // Make sure the result data won't go away after the next call to more()
    if (!_cursors[bestFrom].get()->moreInCurrentBatch()) {
        best = best.getOwned();
    }

    if (_cursors[bestFrom].getMData())
        _cursors[bestFrom].getMData()->pcState->count++;

    return best;
}

void ParallelConnectionMetadata::cleanup(bool full) {
    if (full || errored)
        retryNext = false;

    if (!retryNext && pcState) {
        if (initialized && !errored) {
            verify(pcState->cursor);
            verify(pcState->conn);

            if (!finished && pcState->conn->ok()) {
                try {
                    // Complete the call if only halfway done
                    bool retry = false;
                    pcState->cursor->initLazyFinish(retry);
                } catch (std::exception&) {
                    warning() << "exception closing cursor";
                } catch (...) {
                    warning() << "unknown exception closing cursor";
                }
            }
        }

        // Double-check conn is closed
        if (pcState->conn) {
            pcState->conn->done();
        }

        pcState.reset();
    } else
        verify(finished || !initialized);

    initialized = false;
    finished = false;
    completed = false;
    errored = false;
}

BSONObj ParallelConnectionMetadata::toBSON() const {
    return BSON("state" << (pcState ? pcState->toBSON() : BSONObj()) << "retryNext" << retryNext
                        << "init"
                        << initialized
                        << "finish"
                        << finished
                        << "errored"
                        << errored);
}

std::string ParallelConnectionState::toString() const {
    return str::stream() << "PCState : " << toBSON();
}

BSONObj ParallelConnectionState::toBSON() const {
    BSONObj cursorPeek = BSON("no cursor"
                              << "");
    if (cursor) {
        vector<BSONObj> v;
        cursor->peek(v, 1);
        if (v.size() == 0)
            cursorPeek = BSON("no data"
                              << "");
        else
            cursorPeek = BSON("" << v[0]);
    }

    BSONObj stateObj =
        BSON("conn" << (conn ? (conn->ok() ? conn->conn().toString() : "(done)") : "") << "vinfo"
                    << (manager ? (str::stream() << manager->getns().ns() << " @ "
                                                 << manager->getVersion().toString())
                                : primary->toString()));

    // Append cursor data if exists
    BSONObjBuilder stateB;
    stateB.appendElements(stateObj);
    if (!cursor)
        stateB.append("cursor", "(none)");
    else {
        vector<BSONObj> v;
        cursor->peek(v, 1);
        if (v.size() == 0)
            stateB.append("cursor", "(empty)");
        else
            stateB.append("cursor", v[0]);
    }

    stateB.append("count", count);
    stateB.append("done", done);

    return stateB.obj().getOwned();
}

void throwCursorStale(DBClientCursor* cursor) {
    invariant(cursor);

    if (cursor->hasResultFlag(ResultFlag_ShardConfigStale)) {
        BSONObj error;
        cursor->peekError(&error);
        uasserted(StaleConfigInfo::parseFromCommandError(error),
                  "query returned a stale config error");
    }

    if (NamespaceString(cursor->getns()).isCommand()) {
        // Commands that care about versioning (like the count or geoNear command) sometimes return
        // with the stale config error code, but don't set the ShardConfigStale result flag on the
        // cursor.
        //
        // TODO: Standardize stale config reporting.
        BSONObj res = cursor->peekFirst();
        auto status = getStatusFromCommandResult(res);
        if (status == ErrorCodes::StaleConfig) {
            uassertStatusOK(status.withContext("command returned a stale config error"));
        }
    }
}

}  // namespace mongo