summaryrefslogtreecommitdiff
path: root/src/mongo/s/commands/cluster_write_cmd.cpp
blob: c0370d168b02c660533f34fdc55e241633378ce4 (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
/**
 *    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::kSharding

#include "mongo/platform/basic.h"

#include "mongo/base/error_codes.h"
#include "mongo/base/owned_pointer_vector.h"
#include "mongo/client/remote_command_targeter.h"
#include "mongo/db/catalog/document_validation.h"
#include "mongo/db/commands.h"
#include "mongo/db/commands/write_commands/write_commands_common.h"
#include "mongo/db/curop.h"
#include "mongo/db/lasterror.h"
#include "mongo/db/logical_clock.h"
#include "mongo/db/stats/counters.h"
#include "mongo/db/storage/duplicate_key_error_info.h"
#include "mongo/executor/task_executor_pool.h"
#include "mongo/rpc/get_status_from_command_result.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/cluster_last_error_info.h"
#include "mongo/s/commands/cluster_explain.h"
#include "mongo/s/commands/document_shard_key_update_util.h"
#include "mongo/s/grid.h"
#include "mongo/s/multi_statement_transaction_requests_sender.h"
#include "mongo/s/session_catalog_router.h"
#include "mongo/s/transaction_router.h"
#include "mongo/s/would_change_owning_shard_exception.h"
#include "mongo/s/write_ops/batched_command_request.h"
#include "mongo/s/write_ops/batched_command_response.h"
#include "mongo/s/write_ops/chunk_manager_targeter.h"
#include "mongo/s/write_ops/cluster_write.h"
#include "mongo/util/log.h"
#include "mongo/util/timer.h"

namespace mongo {
namespace {

MONGO_FAIL_POINT_DEFINE(hangAfterThrowWouldChangeOwningShardRetryableWrite);

void batchErrorToLastError(const BatchedCommandRequest& request,
                           const BatchedCommandResponse& response,
                           LastError* error) {
    error->reset();

    std::unique_ptr<WriteErrorDetail> commandError;
    WriteErrorDetail* lastBatchError = nullptr;

    if (!response.getOk()) {
        // Command-level error, all writes failed
        commandError = std::make_unique<WriteErrorDetail>();
        commandError->setStatus(response.getTopLevelStatus());
        lastBatchError = commandError.get();
    } else if (response.isErrDetailsSet()) {
        // The last error in the batch is always reported - this matches expected COE semantics for
        // insert batches. For updates and deletes, error is only reported if the error was on the
        // last item.
        const bool lastOpErrored = response.getErrDetails().back()->getIndex() ==
            static_cast<int>(request.sizeWriteOps() - 1);
        if (request.getBatchType() == BatchedCommandRequest::BatchType_Insert || lastOpErrored) {
            lastBatchError = response.getErrDetails().back();
        }
    } else {
        // We don't care about write concern errors, these happen in legacy mode in GLE.
    }

    // Record an error if one exists
    if (lastBatchError) {
        const auto& errMsg = lastBatchError->toStatus().reason();
        error->setLastError(lastBatchError->toStatus().code(),
                            errMsg.empty() ? "see code for details" : errMsg);
        return;
    }

    // Record write stats otherwise
    //
    // NOTE: For multi-write batches, our semantics change a little because we don't have
    // un-aggregated "n" stats
    if (request.getBatchType() == BatchedCommandRequest::BatchType_Update) {
        BSONObj upsertedId;
        if (response.isUpsertDetailsSet()) {
            // Only report the very last item's upserted id if applicable
            if (response.getUpsertDetails().back()->getIndex() + 1 ==
                static_cast<int>(request.sizeWriteOps())) {
                upsertedId = response.getUpsertDetails().back()->getUpsertedID();
            }
        }

        const int numUpserted = response.isUpsertDetailsSet() ? response.sizeUpsertDetails() : 0;
        const auto numMatched = response.getN() - numUpserted;
        invariant(numMatched >= 0);

        // Wrap upserted id in "upserted" field
        BSONObj leUpsertedId;
        if (!upsertedId.isEmpty()) {
            leUpsertedId = upsertedId.firstElement().wrap(kUpsertedFieldName);
        }

        error->recordUpdate(numMatched > 0, response.getN(), leUpsertedId);
    } else if (request.getBatchType() == BatchedCommandRequest::BatchType_Delete) {
        error->recordDelete(response.getN());
    }
}

/**
 * Checks if the response contains a WouldChangeOwningShard error. If it does, asserts that the
 * batch size is 1 and returns the extra info attached to the exception.
 */
boost::optional<WouldChangeOwningShardInfo> getWouldChangeOwningShardErrorInfo(
    OperationContext* opCtx,
    const BatchedCommandRequest& request,
    BatchedCommandResponse* response,
    bool originalCmdInTxn) {

    if (!response->getOk() || !response->isErrDetailsSet()) {
        return boost::none;
    }

    // Updating the shard key when batch size > 1 is disallowed when the document would move
    // shards. If the update is in a transaction uassert. If the write is not in a transaction,
    // change any WouldChangeOwningShard errors in this batch to InvalidOptions to be reported
    // to the user.
    if (request.sizeWriteOps() != 1U) {
        for (auto it = response->getErrDetails().begin(); it != response->getErrDetails().end();
             ++it) {
            if ((*it)->toStatus() != ErrorCodes::WouldChangeOwningShard) {
                continue;
            }

            if (originalCmdInTxn)
                uasserted(ErrorCodes::InvalidOptions,
                          "Document shard key value updates that cause the doc to move shards "
                          "must be sent with write batch of size 1");

            (*it)->setStatus({ErrorCodes::InvalidOptions,
                              "Document shard key value updates that cause the doc to move shards "
                              "must be sent with write batch of size 1"});
        }

        return boost::none;
    } else {
        for (const auto& err : response->getErrDetails()) {
            if (err->toStatus() != ErrorCodes::WouldChangeOwningShard) {
                continue;
            }

            BSONObjBuilder extraInfoBuilder;
            err->toStatus().extraInfo()->serialize(&extraInfoBuilder);
            auto extraInfo = extraInfoBuilder.obj();
            return WouldChangeOwningShardInfo::parseFromCommandError(extraInfo);
        }
    }

    return boost::none;
}

/**
 * Changes the shard key for the document if the response object contains a WouldChangeOwningShard
 * error. If the original command was sent as a retryable write, starts a transaction on the same
 * session and txnNum, deletes the original document, inserts the new one, and commits the
 * transaction. If the original command is part of a transaction, deletes the original document and
 * inserts the new one. Returns whether or not we actually complete the delete and insert.
 */
bool handleWouldChangeOwningShardError(OperationContext* opCtx,
                                       BatchedCommandRequest* request,
                                       BatchedCommandResponse* response,
                                       BatchWriteExecStats stats) {
    auto txnRouter = TransactionRouter::get(opCtx);
    bool isRetryableWrite = opCtx->getTxnNumber() && !txnRouter;

    auto wouldChangeOwningShardErrorInfo =
        getWouldChangeOwningShardErrorInfo(opCtx, *request, response, !isRetryableWrite);
    if (!wouldChangeOwningShardErrorInfo)
        return false;

    bool updatedShardKey = false;
    boost::optional<BSONObj> upsertedId;
    if (isRetryableWrite) {
        if (MONGO_unlikely(hangAfterThrowWouldChangeOwningShardRetryableWrite.shouldFail())) {
            log() << "Hit hangAfterThrowWouldChangeOwningShardRetryableWrite failpoint";
            hangAfterThrowWouldChangeOwningShardRetryableWrite.pauseWhileSet(opCtx);
        }
        RouterOperationContextSession routerSession(opCtx);
        try {
            // Start transaction and re-run the original update command
            auto& readConcernArgs = repl::ReadConcernArgs::get(opCtx);
            readConcernArgs = repl::ReadConcernArgs(repl::ReadConcernLevel::kLocalReadConcern);

            // Ensure the retried operation does not include WC inside the transaction.  The
            // transaction commit will still use the WC, because it uses the WC from the opCtx
            // (which has been set previously in Strategy).
            request->unsetWriteConcern();

            documentShardKeyUpdateUtil::startTransactionForShardKeyUpdate(opCtx);
            // Clear the error details from the response object before sending the write again
            response->unsetErrDetails();
            ClusterWriter::write(opCtx, *request, &stats, response);
            wouldChangeOwningShardErrorInfo =
                getWouldChangeOwningShardErrorInfo(opCtx, *request, response, !isRetryableWrite);
            if (!wouldChangeOwningShardErrorInfo)
                uassertStatusOK(response->toStatus());

            // If we do not get WouldChangeOwningShard when re-running the update, the document has
            // been modified or deleted concurrently and we do not need to delete it and insert a
            // new one.
            updatedShardKey = wouldChangeOwningShardErrorInfo &&
                documentShardKeyUpdateUtil::updateShardKeyForDocument(
                                  opCtx, request->getNS(), wouldChangeOwningShardErrorInfo.get());

            // If the operation was an upsert, record the _id of the new document.
            if (updatedShardKey && wouldChangeOwningShardErrorInfo->getShouldUpsert()) {
                upsertedId = wouldChangeOwningShardErrorInfo->getPostImage()["_id"].wrap();
            }

            // Commit the transaction
            auto commitResponse =
                documentShardKeyUpdateUtil::commitShardKeyUpdateTransaction(opCtx);

            uassertStatusOK(getStatusFromCommandResult(commitResponse));

            auto writeConcernDetail = getWriteConcernErrorDetailFromBSONObj(commitResponse);
            if (writeConcernDetail && !writeConcernDetail->toStatus().isOK())
                response->setWriteConcernError(writeConcernDetail.release());
        } catch (DBException& e) {
            if (e.code() == ErrorCodes::DuplicateKey &&
                e.extraInfo<DuplicateKeyErrorInfo>()->getKeyPattern().hasField("_id")) {
                e.addContext(documentShardKeyUpdateUtil::kDuplicateKeyErrorContext);
            } else {
                e.addContext(documentShardKeyUpdateUtil::kNonDuplicateKeyErrorContext);
            }

            if (!response->isErrDetailsSet() || !response->getErrDetails().back()) {
                auto error = std::make_unique<WriteErrorDetail>();
                error->setIndex(0);
                response->addToErrDetails(error.release());
            }

            // Set the error status to the status of the failed command and abort the transaction.
            auto status = e.toStatus();
            response->getErrDetails().back()->setStatus(status);

            auto txnRouterForAbort = TransactionRouter::get(opCtx);
            if (txnRouterForAbort)
                txnRouterForAbort.implicitlyAbortTransaction(opCtx, status);

            return false;
        }
    } else {
        try {
            // Delete the original document and insert the new one
            updatedShardKey = documentShardKeyUpdateUtil::updateShardKeyForDocument(
                opCtx, request->getNS(), wouldChangeOwningShardErrorInfo.get());

            // If the operation was an upsert, record the _id of the new document.
            if (updatedShardKey && wouldChangeOwningShardErrorInfo->getShouldUpsert()) {
                upsertedId = wouldChangeOwningShardErrorInfo->getPostImage()["_id"].wrap();
            }
        } catch (const ExceptionFor<ErrorCodes::DuplicateKey>& ex) {
            Status status = ex->getKeyPattern().hasField("_id")
                ? ex.toStatus().withContext(documentShardKeyUpdateUtil::kDuplicateKeyErrorContext)
                : ex.toStatus();
            uassertStatusOK(status);
        }
    }

    if (updatedShardKey) {
        // If we get here, the batch size is 1 and we have successfully deleted the old doc
        // and inserted the new one, so it is safe to unset the error details.
        response->unsetErrDetails();
        response->setN(response->getN() + 1);

        if (upsertedId) {
            auto upsertDetail = std::make_unique<BatchedUpsertDetail>();
            upsertDetail->setIndex(0);
            upsertDetail->setUpsertedID(upsertedId.get());
            response->addToUpsertDetails(upsertDetail.release());
        } else {
            response->setNModified(response->getNModified() + 1);
        }
    }

    return updatedShardKey;
}

void updateHostsTargetedMetrics(OperationContext* opCtx,
                                BatchedCommandRequest::BatchType batchType,
                                int nShardsOwningChunks,
                                int nShardsTargeted) {
    NumHostsTargetedMetrics::QueryType writeType;
    switch (batchType) {
        case BatchedCommandRequest::BatchType_Insert:
            writeType = NumHostsTargetedMetrics::QueryType::kInsertCmd;
            break;
        case BatchedCommandRequest::BatchType_Update:
            writeType = NumHostsTargetedMetrics::QueryType::kUpdateCmd;
            break;
        case BatchedCommandRequest::BatchType_Delete:
            writeType = NumHostsTargetedMetrics::QueryType::kDeleteCmd;
            break;

            MONGO_UNREACHABLE;
    }

    auto targetType = NumHostsTargetedMetrics::get(opCtx).parseTargetType(
        opCtx, nShardsTargeted, nShardsOwningChunks);
    NumHostsTargetedMetrics::get(opCtx).addNumHostsTargeted(writeType, targetType);
}

/**
 * Base class for mongos write commands.
 */
class ClusterWriteCmd : public Command {
public:
    virtual ~ClusterWriteCmd() {}

    AllowedOnSecondary secondaryAllowed(ServiceContext*) const final {
        return AllowedOnSecondary::kNever;
    }

protected:
    class InvocationBase;

    ClusterWriteCmd(StringData name) : Command(name) {}

private:
    /**
     * Executes a write command against a particular database, and targets the command based on
     * a write operation.
     *
     * Does *not* retry or retarget if the metadata is stale.
     */
    static Status _commandOpWrite(OperationContext* opCtx,
                                  const std::string& dbName,
                                  const BSONObj& command,
                                  BatchItemRef targetingBatchItem,
                                  std::vector<Strategy::CommandResult>* results) {
        // Note that this implementation will not handle targeting retries and does not completely
        // emulate write behavior
        ChunkManagerTargeter targeter(targetingBatchItem.getRequest()->getNS());
        Status status = targeter.init(opCtx);
        if (!status.isOK())
            return status;

        auto swEndpoints = [&]() -> StatusWith<std::vector<ShardEndpoint>> {
            if (targetingBatchItem.getOpType() == BatchedCommandRequest::BatchType_Insert) {
                auto swEndpoint = targeter.targetInsert(opCtx, targetingBatchItem.getDocument());
                if (!swEndpoint.isOK())
                    return swEndpoint.getStatus();
                return std::vector<ShardEndpoint>{std::move(swEndpoint.getValue())};
            } else if (targetingBatchItem.getOpType() == BatchedCommandRequest::BatchType_Update) {
                return targeter.targetUpdate(opCtx, targetingBatchItem.getUpdate());
            } else if (targetingBatchItem.getOpType() == BatchedCommandRequest::BatchType_Delete) {
                return targeter.targetDelete(opCtx, targetingBatchItem.getDelete());
            } else {
                MONGO_UNREACHABLE;
            }
        }();

        if (!swEndpoints.isOK())
            return swEndpoints.getStatus();

        // Assemble requests
        std::vector<AsyncRequestsSender::Request> requests;
        for (const auto& endpoint : swEndpoints.getValue()) {
            BSONObj cmdObjWithVersions = BSONObj(command);
            if (endpoint.databaseVersion) {
                cmdObjWithVersions =
                    appendDbVersionIfPresent(cmdObjWithVersions, *endpoint.databaseVersion);
            }
            cmdObjWithVersions = appendShardVersion(cmdObjWithVersions, endpoint.shardVersion);
            requests.emplace_back(endpoint.shardName, cmdObjWithVersions);
        }

        // Send the requests.

        const ReadPreferenceSetting readPref(ReadPreference::PrimaryOnly, TagSet());
        MultiStatementTransactionRequestsSender ars(
            opCtx,
            Grid::get(opCtx)->getExecutorPool()->getArbitraryExecutor(),
            dbName,
            requests,
            readPref,
            Shard::RetryPolicy::kNoRetry);

        // Receive the responses.

        Status dispatchStatus = Status::OK();
        while (!ars.done()) {
            // Block until a response is available.
            auto response = ars.next();

            if (!response.swResponse.isOK()) {
                dispatchStatus = std::move(response.swResponse.getStatus());
                break;
            }

            Strategy::CommandResult result;

            // If the response status was OK, the response must contain which host was targeted.
            invariant(response.shardHostAndPort);
            result.target = ConnectionString(std::move(*response.shardHostAndPort));

            result.shardTargetId = std::move(response.shardId);
            result.result = std::move(response.swResponse.getValue().data);

            results->push_back(result);
        }

        return dispatchStatus;
    }
};

class ClusterWriteCmd::InvocationBase : public CommandInvocation {
public:
    InvocationBase(const ClusterWriteCmd* command,
                   const OpMsgRequest& request,
                   BatchedCommandRequest batchedRequest)
        : CommandInvocation(command),
          _bypass{shouldBypassDocumentValidationForCommand(request.body)},
          _request{&request},
          _batchedRequest{std::move(batchedRequest)} {}

    const BatchedCommandRequest& getBatchedRequest() const {
        return _batchedRequest;
    }

    bool getBypass() const {
        return _bypass;
    }

private:
    virtual void doCheckAuthorizationHook(AuthorizationSession* authzSession) const = 0;

    bool runImpl(OperationContext* opCtx,
                 const OpMsgRequest& request,
                 BatchedCommandRequest& batchedRequest,
                 BSONObjBuilder& result) const {
        BatchWriteExecStats stats;
        BatchedCommandResponse response;
        ClusterWriter::write(opCtx, batchedRequest, &stats, &response);

        bool updatedShardKey = false;
        if (_batchedRequest.getBatchType() == BatchedCommandRequest::BatchType_Update) {
            updatedShardKey =
                handleWouldChangeOwningShardError(opCtx, &batchedRequest, &response, stats);
        }

        // Populate the lastError object based on the write response
        batchErrorToLastError(batchedRequest, response, &LastError::get(opCtx->getClient()));

        size_t numAttempts;

        if (!response.getOk()) {
            numAttempts = 0;
        } else if (batchedRequest.getWriteCommandBase().getOrdered() &&
                   response.isErrDetailsSet()) {
            // Add one failed attempt
            numAttempts = response.getErrDetailsAt(0)->getIndex() + 1;
        } else {
            numAttempts = batchedRequest.sizeWriteOps();
        }

        // TODO: increase opcounters by more than one
        auto& debug = CurOp::get(opCtx)->debug();
        switch (_batchedRequest.getBatchType()) {
            case BatchedCommandRequest::BatchType_Insert:
                for (size_t i = 0; i < numAttempts; ++i) {
                    globalOpCounters.gotInsert();
                }
                debug.additiveMetrics.ninserted = response.getN();
                break;
            case BatchedCommandRequest::BatchType_Update:
                for (size_t i = 0; i < numAttempts; ++i) {
                    globalOpCounters.gotUpdate();
                }
                debug.upsert = response.isUpsertDetailsSet();
                debug.additiveMetrics.nMatched =
                    response.getN() - (debug.upsert ? response.sizeUpsertDetails() : 0);
                debug.additiveMetrics.nModified = response.getNModified();
                break;
            case BatchedCommandRequest::BatchType_Delete:
                for (size_t i = 0; i < numAttempts; ++i) {
                    globalOpCounters.gotDelete();
                }
                debug.additiveMetrics.ndeleted = response.getN();
                break;
        }

        // Save the last opTimes written on each shard for this client, to allow GLE to work
        ClusterLastErrorInfo::get(opCtx->getClient())->addHostOpTimes(stats.getWriteOpTimes());

        // Record the number of shards targeted by this write.
        CurOp::get(opCtx)->debug().nShards =
            stats.getTargetedShards().size() + (updatedShardKey ? 1 : 0);

        if (stats.getNumShardsOwningChunks().is_initialized())
            updateHostsTargetedMetrics(opCtx,
                                       _batchedRequest.getBatchType(),
                                       stats.getNumShardsOwningChunks().get(),
                                       stats.getTargetedShards().size() +
                                           (updatedShardKey ? 1 : 0));

        if (auto txnRouter = TransactionRouter::get(opCtx)) {
            auto writeCmdStatus = response.toStatus();
            if (!writeCmdStatus.isOK()) {
                txnRouter.implicitlyAbortTransaction(opCtx, writeCmdStatus);
            }
        }

        result.appendElements(response.toBSON());
        return response.getOk();
    }

    void run(OperationContext* opCtx, rpc::ReplyBuilderInterface* result) override {
        BSONObjBuilder bob = result->getBodyBuilder();
        bool ok = runImpl(opCtx, *_request, _batchedRequest, bob);
        if (!ok)
            CommandHelpers::appendSimpleCommandStatus(bob, ok);
    }

    void explain(OperationContext* opCtx,
                 ExplainOptions::Verbosity verbosity,
                 rpc::ReplyBuilderInterface* result) override {
        uassert(ErrorCodes::InvalidLength,
                "explained write batches must be of size 1",
                _batchedRequest.sizeWriteOps() == 1U);

        const auto explainCmd = ClusterExplain::wrapAsExplain(_request->body, verbosity);

        // We will time how long it takes to run the commands on the shards.
        Timer timer;

        // Target the command to the shards based on the singleton batch item.
        BatchItemRef targetingBatchItem(&_batchedRequest, 0);
        std::vector<Strategy::CommandResult> shardResults;
        uassertStatusOK(_commandOpWrite(opCtx,
                                        _request->getDatabase().toString(),
                                        explainCmd,
                                        targetingBatchItem,
                                        &shardResults));
        auto bodyBuilder = result->getBodyBuilder();
        uassertStatusOK(ClusterExplain::buildExplainResult(
            opCtx, shardResults, ClusterExplain::kWriteOnShards, timer.millis(), &bodyBuilder));
    }

    NamespaceString ns() const override {
        return _batchedRequest.getNS();
    }

    bool supportsWriteConcern() const override {
        return true;
    }

    ReadConcernSupportResult supportsReadConcern(repl::ReadConcernLevel level) const final {
        return ReadConcernSupportResult::allSupportedAndDefaultPermitted();
    }

    void doCheckAuthorization(OperationContext* opCtx) const final {
        try {
            doCheckAuthorizationHook(AuthorizationSession::get(opCtx->getClient()));
        } catch (const DBException& e) {
            LastError::get(opCtx->getClient()).setLastError(e.code(), e.reason());
            throw;
        }
    }

    const ClusterWriteCmd* command() const {
        return static_cast<const ClusterWriteCmd*>(definition());
    }

    bool _bypass;
    const OpMsgRequest* _request;
    BatchedCommandRequest _batchedRequest;
};

class ClusterInsertCmd final : public ClusterWriteCmd {
public:
    ClusterInsertCmd() : ClusterWriteCmd("insert") {}

private:
    class Invocation final : public InvocationBase {
    public:
        using InvocationBase::InvocationBase;

    private:
        void doCheckAuthorizationHook(AuthorizationSession* authzSession) const final {
            auth::checkAuthForInsertCommand(
                authzSession, getBypass(), getBatchedRequest().getInsertRequest());
        }
    };

    std::unique_ptr<CommandInvocation> parse(OperationContext* opCtx,
                                             const OpMsgRequest& request) final {
        return std::make_unique<Invocation>(
            this,
            request,
            BatchedCommandRequest::cloneInsertWithIds(BatchedCommandRequest::parseInsert(request)));
    }

    std::string help() const override {
        return "insert documents";
    }

    LogicalOp getLogicalOp() const override {
        return LogicalOp::opInsert;
    }
} clusterInsertCmd;

class ClusterUpdateCmd final : public ClusterWriteCmd {
public:
    ClusterUpdateCmd() : ClusterWriteCmd("update") {}

private:
    class Invocation final : public InvocationBase {
    public:
        using InvocationBase::InvocationBase;

    private:
        void doCheckAuthorizationHook(AuthorizationSession* authzSession) const final {
            auth::checkAuthForUpdateCommand(
                authzSession, getBypass(), getBatchedRequest().getUpdateRequest());
        }
    };

    std::unique_ptr<CommandInvocation> parse(OperationContext* opCtx,
                                             const OpMsgRequest& request) final {
        auto parsedRequest = BatchedCommandRequest::parseUpdate(request);
        uassert(51195,
                "Cannot specify runtime constants option to a mongos",
                !parsedRequest.hasRuntimeConstants());
        parsedRequest.setRuntimeConstants(Variables::generateRuntimeConstants(opCtx));
        return std::make_unique<Invocation>(this, request, std::move(parsedRequest));
    }

    std::string help() const override {
        return "update documents";
    }

    LogicalOp getLogicalOp() const override {
        return LogicalOp::opUpdate;
    }
} clusterUpdateCmd;

class ClusterDeleteCmd final : public ClusterWriteCmd {
public:
    ClusterDeleteCmd() : ClusterWriteCmd("delete") {}

private:
    class Invocation final : public InvocationBase {
    public:
        using InvocationBase::InvocationBase;

    private:
        void doCheckAuthorizationHook(AuthorizationSession* authzSession) const final {
            auth::checkAuthForDeleteCommand(
                authzSession, getBypass(), getBatchedRequest().getDeleteRequest());
        }
    };

    std::unique_ptr<CommandInvocation> parse(OperationContext* opCtx,
                                             const OpMsgRequest& request) final {
        return std::make_unique<Invocation>(
            this, request, BatchedCommandRequest::parseDelete(request));
    }

    std::string help() const override {
        return "delete documents";
    }

    LogicalOp getLogicalOp() const override {
        return LogicalOp::opDelete;
    }
} clusterDeleteCmd;

}  // namespace
}  // namespace mongo