summaryrefslogtreecommitdiff
path: root/src/mongo/s/query/cluster_aggregate.cpp
blob: 9fd49e2e0047dde76c8cebe550eff4b63777b15b (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
/**
 *    Copyright (C) 2018-present MongoDB, Inc.
 *
 *    This program is free software: you can redistribute it and/or modify
 *    it under the terms of the Server Side Public License, version 1,
 *    as published by MongoDB, Inc.
 *
 *    This program is distributed in the hope that it will be useful,
 *    but WITHOUT ANY WARRANTY; without even the implied warranty of
 *    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 *    Server Side Public License for more details.
 *
 *    You should have received a copy of the Server Side Public License
 *    along with this program. If not, see
 *    <http://www.mongodb.com/licensing/server-side-public-license>.
 *
 *    As a special exception, the copyright holders give permission to link the
 *    code of portions of this program with the OpenSSL library under certain
 *    conditions as described in each individual source file and distribute
 *    linked combinations including the program with the OpenSSL library. You
 *    must comply with the Server Side Public License in all respects for
 *    all of the code used other than as permitted herein. If you modify file(s)
 *    with this exception, you may extend this exception to your version of the
 *    file(s), but you are not obligated to do so. If you do not wish to do so,
 *    delete this exception statement from your version. If you delete this
 *    exception statement from all source files in the program, then also delete
 *    it in the license file.
 */


#include "mongo/platform/basic.h"

#include "mongo/s/query/cluster_aggregate.h"

#include <boost/intrusive_ptr.hpp>

#include "mongo/db/api_parameters.h"
#include "mongo/db/auth/authorization_session.h"
#include "mongo/db/catalog/collection_uuid_mismatch_info.h"
#include "mongo/db/catalog_shard_feature_flag_gen.h"
#include "mongo/db/client.h"
#include "mongo/db/commands.h"
#include "mongo/db/curop.h"
#include "mongo/db/fle_crud.h"
#include "mongo/db/operation_context.h"
#include "mongo/db/pipeline/document_source_change_stream.h"
#include "mongo/db/pipeline/document_source_internal_unpack_bucket.h"
#include "mongo/db/pipeline/document_source_out.h"
#include "mongo/db/pipeline/expression_context.h"
#include "mongo/db/pipeline/lite_parsed_pipeline.h"
#include "mongo/db/pipeline/pipeline.h"
#include "mongo/db/pipeline/process_interface/mongos_process_interface.h"
#include "mongo/db/pipeline/sharded_agg_helpers.h"
#include "mongo/db/query/collation/collator_factory_interface.h"
#include "mongo/db/query/cursor_response.h"
#include "mongo/db/query/explain_common.h"
#include "mongo/db/query/find_common.h"
#include "mongo/db/query/fle/server_rewrite.h"
#include "mongo/db/query/query_stats.h"
#include "mongo/db/timeseries/timeseries_gen.h"
#include "mongo/db/timeseries/timeseries_options.h"
#include "mongo/db/views/resolved_view.h"
#include "mongo/db/views/view.h"
#include "mongo/db/write_concern_options.h"
#include "mongo/executor/task_executor_pool.h"
#include "mongo/logv2/log.h"
#include "mongo/rpc/get_status_from_command_result.h"
#include "mongo/rpc/op_msg_rpc_impls.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/multi_statement_transaction_requests_sender.h"
#include "mongo/s/query/cluster_aggregation_planner.h"
#include "mongo/s/query/cluster_client_cursor_impl.h"
#include "mongo/s/query/cluster_client_cursor_params.h"
#include "mongo/s/query/cluster_cursor_manager.h"
#include "mongo/s/query/cluster_query_knobs_gen.h"
#include "mongo/s/query/document_source_merge_cursors.h"
#include "mongo/s/query/establish_cursors.h"
#include "mongo/s/query/owned_remote_cursor.h"
#include "mongo/s/query/router_stage_pipeline.h"
#include "mongo/s/query/store_possible_cursor.h"
#include "mongo/s/stale_exception.h"
#include "mongo/s/transaction_router.h"
#include "mongo/util/net/socket_utils.h"

#define MONGO_LOGV2_DEFAULT_COMPONENT ::mongo::logv2::LogComponent::kCommand


namespace mongo {

constexpr unsigned ClusterAggregate::kMaxViewRetries;

namespace {

// "Resolve" involved namespaces into a map. We won't try to execute anything on a mongos, but we
// still have to populate this map so that any $lookups, etc. will be able to have a resolved view
// definition. It's okay that this is incorrect, we will repopulate the real namespace map on the
// mongod. Note that this function must be called before forwarding an aggregation command on an
// unsharded collection, in order to verify that the involved namespaces are allowed to be sharded.
auto resolveInvolvedNamespaces(stdx::unordered_set<NamespaceString> involvedNamespaces) {
    StringMap<ExpressionContext::ResolvedNamespace> resolvedNamespaces;
    for (auto&& nss : involvedNamespaces) {
        resolvedNamespaces.try_emplace(nss.coll(), nss, std::vector<BSONObj>{});
    }
    return resolvedNamespaces;
}

// Build an appropriate ExpressionContext for the pipeline. This helper instantiates an appropriate
// collator, creates a MongoProcessInterface for use by the pipeline's stages, and sets the
// collection UUID if provided.
boost::intrusive_ptr<ExpressionContext> makeExpressionContext(
    OperationContext* opCtx,
    const AggregateCommandRequest& request,
    BSONObj collationObj,
    boost::optional<UUID> uuid,
    StringMap<ExpressionContext::ResolvedNamespace> resolvedNamespaces,
    bool hasChangeStream) {

    std::unique_ptr<CollatorInterface> collation;
    if (!collationObj.isEmpty()) {
        // This will be null if attempting to build an interface for the simple collator.
        collation = uassertStatusOK(
            CollatorFactoryInterface::get(opCtx->getServiceContext())->makeFromBSON(collationObj));
    }

    // Create the expression context, and set 'inMongos' to true. We explicitly do *not* set
    // mergeCtx->tempDir.
    auto mergeCtx = make_intrusive<ExpressionContext>(
        opCtx,
        request,
        std::move(collation),
        std::make_shared<MongosProcessInterface>(
            Grid::get(opCtx)->getExecutorPool()->getArbitraryExecutor()),
        std::move(resolvedNamespaces),
        uuid);

    mergeCtx->inMongos = true;

    // Serialize the 'AggregateCommandRequest' and save it so that the original command can be
    // reconstructed for dispatch to a new shard, which is sometimes necessary for change streams
    // pipelines.
    if (hasChangeStream) {
        mergeCtx->originalAggregateCommand =
            aggregation_request_helper::serializeToCommandObj(request);
    }

    return mergeCtx;
}

void appendEmptyResultSetWithStatus(OperationContext* opCtx,
                                    const NamespaceString& nss,
                                    Status status,
                                    BSONObjBuilder* result) {
    // Rewrite ShardNotFound as NamespaceNotFound so that appendEmptyResultSet swallows it.
    if (status == ErrorCodes::ShardNotFound) {
        status = {ErrorCodes::NamespaceNotFound, status.reason()};
    }
    appendEmptyResultSet(opCtx, *result, status, nss);
}

void updateHostsTargetedMetrics(OperationContext* opCtx,
                                const NamespaceString& executionNss,
                                const boost::optional<ChunkManager>& cm,
                                stdx::unordered_set<NamespaceString> involvedNamespaces) {
    if (!cm)
        return;

    // Create a set of ShardIds that own a chunk belonging to any of the collections involved in
    // this pipeline. This will be used to determine whether the pipeline targeted all of the shards
    // that own chunks for any collection involved or not.
    std::set<ShardId> shardsOwningChunks = [&]() {
        std::set<ShardId> shardsIds;

        if (cm->isSharded()) {
            std::set<ShardId> shardIdsForNs;
            cm->getAllShardIds(&shardIdsForNs);
            for (const auto& shardId : shardIdsForNs) {
                shardsIds.insert(shardId);
            }
        }

        for (const auto& nss : involvedNamespaces) {
            if (nss == executionNss)
                continue;

            const auto [resolvedNsCM, _] =
                uassertStatusOK(getCollectionRoutingInfoForTxnCmd(opCtx, nss));
            if (resolvedNsCM.isSharded()) {
                std::set<ShardId> shardIdsForNs;
                resolvedNsCM.getAllShardIds(&shardIdsForNs);
                for (const auto& shardId : shardIdsForNs) {
                    shardsIds.insert(shardId);
                }
            }
        }

        return shardsIds;
    }();

    auto nShardsTargeted = CurOp::get(opCtx)->debug().nShards;
    if (nShardsTargeted > 0) {
        auto targetType = NumHostsTargetedMetrics::get(opCtx).parseTargetType(
            opCtx, nShardsTargeted, shardsOwningChunks.size());
        NumHostsTargetedMetrics::get(opCtx).addNumHostsTargeted(
            NumHostsTargetedMetrics::QueryType::kAggregateCmd, targetType);
    }
}

/**
 * Performs validations related to API versioning, time-series stages, and general command
 * validation.
 * Throws UserAssertion if any of the validations fails
 *     - validation of API versioning on each stage on the pipeline
 *     - validation of API versioning on 'AggregateCommandRequest' request
 *     - validation of time-series related stages
 *     - validation of command parameters
 */
void performValidationChecks(const OperationContext* opCtx,
                             const AggregateCommandRequest& request,
                             const LiteParsedPipeline& liteParsedPipeline) {
    liteParsedPipeline.validate(opCtx);
    aggregation_request_helper::validateRequestForAPIVersion(opCtx, request);
    aggregation_request_helper::validateRequestFromClusterQueryWithoutShardKey(request);
}

/**
 * Rebuilds the pipeline and uses a different granularity value for the 'bucketMaxSpanSeconds' field
 * in the $_internalUnpackBucket stage.
 */
std::vector<BSONObj> rebuildPipelineWithTimeSeriesGranularity(
    const std::vector<BSONObj>& pipeline,
    boost::optional<BucketGranularityEnum> granularity,
    boost::optional<int32_t> maxSpanSeconds) {
    int32_t bucketSpan = 0;

    if (maxSpanSeconds) {
        bucketSpan = *maxSpanSeconds;
    } else {
        bucketSpan = timeseries::getMaxSpanSecondsFromGranularity(
            granularity.get_value_or(BucketGranularityEnum::Seconds));
    }

    std::vector<BSONObj> newPipeline;
    for (auto& stage : pipeline) {
        if (stage.firstElementFieldNameStringData() ==
                DocumentSourceInternalUnpackBucket::kStageNameInternal ||
            stage.firstElementFieldNameStringData() ==
                DocumentSourceInternalUnpackBucket::kStageNameExternal) {
            BSONObjBuilder newOptions;
            for (auto& elem : stage.firstElement().Obj()) {
                if (elem.fieldNameStringData() ==
                    DocumentSourceInternalUnpackBucket::kBucketMaxSpanSeconds) {
                    newOptions.append(DocumentSourceInternalUnpackBucket::kBucketMaxSpanSeconds,
                                      bucketSpan);
                } else {
                    newOptions.append(elem);
                }
            }
            newPipeline.push_back(
                BSON(stage.firstElementFieldNameStringData() << newOptions.obj()));
            continue;
        }
        newPipeline.push_back(stage);
    }
    return newPipeline;
}

}  // namespace

Status ClusterAggregate::runAggregate(OperationContext* opCtx,
                                      const Namespaces& namespaces,
                                      AggregateCommandRequest& request,
                                      const PrivilegeVector& privileges,
                                      BSONObjBuilder* result) {
    return runAggregate(opCtx, namespaces, request, {request}, privileges, result);
}

Status ClusterAggregate::runAggregate(OperationContext* opCtx,
                                      const Namespaces& namespaces,
                                      AggregateCommandRequest& request,
                                      const LiteParsedPipeline& liteParsedPipeline,
                                      const PrivilegeVector& privileges,
                                      BSONObjBuilder* result) {
    return runAggregate(
        opCtx, namespaces, request, liteParsedPipeline, privileges, boost::none, result);
}

Status ClusterAggregate::runAggregate(OperationContext* opCtx,
                                      const Namespaces& namespaces,
                                      AggregateCommandRequest& request,
                                      const LiteParsedPipeline& liteParsedPipeline,
                                      const PrivilegeVector& privileges,
                                      boost::optional<CollectionRoutingInfo> cri,
                                      BSONObjBuilder* result) {
    // Perform some validations on the LiteParsedPipeline and request before continuing with the
    // aggregation command.
    performValidationChecks(opCtx, request, liteParsedPipeline);

    uassert(51028, "Cannot specify exchange option to a mongos", !request.getExchange());
    uassert(51143,
            "Cannot specify runtime constants option to a mongos",
            !request.getLegacyRuntimeConstants());
    uassert(51089,
            str::stream() << "Internal parameter(s) ["
                          << AggregateCommandRequest::kNeedsMergeFieldName << ", "
                          << AggregateCommandRequest::kFromMongosFieldName
                          << "] cannot be set to 'true' when sent to mongos",
            !request.getNeedsMerge() && !request.getFromMongos());

    const auto isSharded = [](OperationContext* opCtx, const NamespaceString& nss) {
        const auto [resolvedNsCM, _] =
            uassertStatusOK(getCollectionRoutingInfoForTxnCmd(opCtx, nss));
        return resolvedNsCM.isSharded();
    };

    liteParsedPipeline.verifyIsSupported(
        opCtx, isSharded, request.getExplain(), serverGlobalParams.enableMajorityReadConcern);
    auto hasChangeStream = liteParsedPipeline.hasChangeStream();
    auto involvedNamespaces = liteParsedPipeline.getInvolvedNamespaces();
    auto shouldDoFLERewrite = ::mongo::shouldDoFLERewrite(request);
    auto startsWithDocuments = liteParsedPipeline.startsWithDocuments();

    if (!shouldDoFLERewrite) {
        query_stats::registerAggRequest(request, opCtx);
    }

    // If the routing table is not already taken by the higher level, fill it now.
    if (!cri) {
        // If the routing table is valid, we obtain a reference to it. If the table is not valid,
        // then either the database does not exist, or there are no shards in the cluster. In the
        // latter case, we always return an empty cursor. In the former case, if the requested
        // aggregation is a $changeStream, we allow the operation to continue so that stream cursors
        // can be established on the given namespace before the database or collection is actually
        // created. If the database does not exist and this is not a $changeStream, then we return
        // an empty cursor.
        auto executionNsRoutingInfoStatus =
            sharded_agg_helpers::getExecutionNsRoutingInfo(opCtx, namespaces.executionNss);

        if (!executionNsRoutingInfoStatus.isOK()) {
            uassert(CollectionUUIDMismatchInfo(request.getDbName(),
                                               *request.getCollectionUUID(),
                                               request.getNamespace().coll().toString(),
                                               boost::none),
                    "Database does not exist",
                    executionNsRoutingInfoStatus != ErrorCodes::NamespaceNotFound ||
                        !request.getCollectionUUID());

            if (liteParsedPipeline.startsWithCollStats()) {
                uassertStatusOKWithContext(executionNsRoutingInfoStatus,
                                           "Unable to retrieve information for $collStats stage");
            }
        }

        if (executionNsRoutingInfoStatus.isOK()) {
            cri = executionNsRoutingInfoStatus.getValue();
        } else if (!((hasChangeStream || startsWithDocuments) &&
                     executionNsRoutingInfoStatus == ErrorCodes::NamespaceNotFound)) {
            appendEmptyResultSetWithStatus(
                opCtx, namespaces.requestedNss, executionNsRoutingInfoStatus.getStatus(), result);
            return Status::OK();
        }
    }

    boost::intrusive_ptr<ExpressionContext> expCtx;
    const auto pipelineBuilder = [&]() {
        // Populate the collection UUID and the appropriate collation to use.
        auto [collationObj, uuid] = [&]() -> std::pair<BSONObj, boost::optional<UUID>> {
            // If this is a change stream, take the user-defined collation if one exists, or an
            // empty BSONObj otherwise. Change streams never inherit the collection's default
            // collation, and since collectionless aggregations generally run on the 'admin'
            // database, the standard logic would attempt to resolve its non-existent UUID and
            // collation by sending a specious 'listCollections' command to the config servers.
            if (hasChangeStream) {
                return {request.getCollation().value_or(BSONObj()), boost::none};
            }

            return cluster_aggregation_planner::getCollationAndUUID(
                opCtx,
                cri ? boost::make_optional(cri->cm) : boost::none,
                namespaces.executionNss,
                request.getCollation().value_or(BSONObj()));
        }();

        // Build an ExpressionContext for the pipeline. This instantiates an appropriate collator,
        // resolves all involved namespaces, and creates a shared MongoProcessInterface for use by
        // the pipeline's stages.
        expCtx = makeExpressionContext(opCtx,
                                       request,
                                       collationObj,
                                       uuid,
                                       resolveInvolvedNamespaces(involvedNamespaces),
                                       hasChangeStream);

        // Parse and optimize the full pipeline.
        auto pipeline = Pipeline::parse(request.getPipeline(), expCtx);

        // If the aggregate command supports encrypted collections, do rewrites of the pipeline to
        // support querying against encrypted fields.
        if (shouldDoFLERewrite) {
            if (!request.getEncryptionInformation()->getCrudProcessed().value_or(false)) {
                pipeline = processFLEPipelineS(opCtx,
                                               namespaces.executionNss,
                                               request.getEncryptionInformation().value(),
                                               std::move(pipeline));
                request.getEncryptionInformation()->setCrudProcessed(true);
            }

            CurOp::get(opCtx)->debug().shouldOmitDiagnosticInformation = true;
        }

        pipeline->optimizePipeline();

        // Validate the pipeline post-optimization.
        const bool alreadyOptimized = true;
        pipeline->validateCommon(alreadyOptimized);

        return pipeline;
    };

    // The pipeline is not allowed to passthrough if any stage is not allowed to passthrough or if
    // the pipeline needs to undergo FLE rewriting first.
    auto allowedToPassthrough =
        liteParsedPipeline.allowedToPassthroughFromMongos() && !shouldDoFLERewrite;
    auto targeter = cluster_aggregation_planner::AggregationTargeter::make(
        opCtx,
        namespaces.executionNss,
        pipelineBuilder,
        cri,
        involvedNamespaces,
        hasChangeStream,
        startsWithDocuments,
        allowedToPassthrough,
        request.getPassthroughToShard().has_value());

    uassert(
        6487500,
        fmt::format("Cannot use {} with an aggregation that executes entirely on mongos",
                    AggregateCommandRequest::kCollectionUUIDFieldName),
        !request.getCollectionUUID() ||
            targeter.policy !=
                cluster_aggregation_planner::AggregationTargeter::TargetingPolicy::kMongosRequired);

    if (!expCtx) {
        // When the AggregationTargeter chooses a "passthrough" policy, it does not call the
        // 'pipelineBuilder' function, so we never get an expression context. Because this is a
        // passthrough, we only need a bare minimum expression context anyway.
        invariant(targeter.policy ==
                      cluster_aggregation_planner::AggregationTargeter::kPassthrough ||
                  targeter.policy ==
                      cluster_aggregation_planner::AggregationTargeter::kSpecificShardOnly);
        expCtx = make_intrusive<ExpressionContext>(
            opCtx, nullptr, namespaces.executionNss, boost::none, request.getLet());
    }

    if (request.getExplain()) {
        explain_common::generateServerInfo(result);
        explain_common::generateServerParameters(result);
    }

    auto status = [&]() {
        switch (targeter.policy) {
            case cluster_aggregation_planner::AggregationTargeter::TargetingPolicy::kPassthrough: {
                // A pipeline with $changeStream should never be allowed to passthrough.
                invariant(!hasChangeStream);
                const bool eligibleForSampling = !request.getExplain();
                return cluster_aggregation_planner::runPipelineOnPrimaryShard(
                    expCtx,
                    namespaces,
                    targeter.cri->cm,
                    request.getExplain(),
                    aggregation_request_helper::serializeToCommandDoc(request),
                    privileges,
                    eligibleForSampling,
                    result);
            }

            case cluster_aggregation_planner::AggregationTargeter::TargetingPolicy::
                kMongosRequired: {
                // If this is an explain write the explain output and return.
                auto expCtx = targeter.pipeline->getContext();
                if (expCtx->explain) {
                    *result << "splitPipeline" << BSONNULL << "mongos"
                            << Document{{"host", getHostNameCachedAndPort()},
                                        {"stages",
                                         targeter.pipeline->writeExplainOps(*expCtx->explain)}};
                    return Status::OK();
                }

                return cluster_aggregation_planner::runPipelineOnMongoS(
                    namespaces,
                    request.getCursor().getBatchSize().value_or(
                        aggregation_request_helper::kDefaultBatchSize),
                    std::move(targeter.pipeline),
                    result,
                    privileges);
            }

            case cluster_aggregation_planner::AggregationTargeter::TargetingPolicy::kAnyShard: {
                const bool eligibleForSampling = !request.getExplain();
                return cluster_aggregation_planner::dispatchPipelineAndMerge(
                    opCtx,
                    std::move(targeter),
                    aggregation_request_helper::serializeToCommandDoc(request),
                    request.getCursor().getBatchSize().value_or(
                        aggregation_request_helper::kDefaultBatchSize),
                    namespaces,
                    privileges,
                    result,
                    hasChangeStream,
                    startsWithDocuments,
                    eligibleForSampling);
            }
            case cluster_aggregation_planner::AggregationTargeter::TargetingPolicy::
                kSpecificShardOnly: {
                // Mark expCtx as tailable and await data so CCC behaves accordingly.
                expCtx->tailableMode = TailableModeEnum::kTailableAndAwaitData;

                uassert(6273801,
                        "per shard cursor pipeline must contain $changeStream",
                        hasChangeStream);

                // Make sure the rest of the pipeline can be pushed down.
                auto pipeline = request.getPipeline();
                std::vector<BSONObj> nonChangeStreamPart(pipeline.begin() + 1, pipeline.end());
                LiteParsedPipeline nonChangeStreamLite(request.getNamespace(), nonChangeStreamPart);
                uassert(6273802,
                        "$_passthroughToShard specified with a stage that is not allowed to "
                        "passthrough from mongos",
                        nonChangeStreamLite.allowedToPassthroughFromMongos());
                ShardId shardId(std::string(request.getPassthroughToShard()->getShard()));
                // (Ignore FCV check): This is in mongos so we expect to ignore FCV.
                uassert(6273803,
                        "$_passthroughToShard not supported for queries against config replica set",
                        shardId != ShardId::kConfigServerId ||
                            gFeatureFlagCatalogShard.isEnabledAndIgnoreFCVUnsafe());
                // This is an aggregation pipeline started internally, so it is not eligible for
                // sampling.
                const bool eligibleForSampling = false;

                return cluster_aggregation_planner::runPipelineOnSpecificShardOnly(
                    expCtx,
                    namespaces,
                    boost::none,
                    request.getExplain(),
                    aggregation_request_helper::serializeToCommandDoc(request),
                    privileges,
                    shardId,
                    true /* forPerShardCursor */,
                    eligibleForSampling,
                    result);
            }

                MONGO_UNREACHABLE;
        }
        MONGO_UNREACHABLE;
    }();

    if (status.isOK()) {
        updateHostsTargetedMetrics(opCtx,
                                   namespaces.executionNss,
                                   cri ? boost::make_optional(cri->cm) : boost::none,
                                   involvedNamespaces);
        // Report usage statistics for each stage in the pipeline.
        liteParsedPipeline.tickGlobalStageCounters();
        // Add 'command' object to explain output.
        if (expCtx->explain) {
            explain_common::appendIfRoom(
                aggregation_request_helper::serializeToCommandObj(request), "command", result);
        }
    }
    return status;
}

Status ClusterAggregate::retryOnViewError(OperationContext* opCtx,
                                          const AggregateCommandRequest& request,
                                          const ResolvedView& resolvedView,
                                          const NamespaceString& requestedNss,
                                          const PrivilegeVector& privileges,
                                          BSONObjBuilder* result,
                                          unsigned numberRetries) {
    if (numberRetries >= kMaxViewRetries) {
        return Status(ErrorCodes::InternalError,
                      "Failed to resolve view after max number of retries.");
    }

    auto resolvedAggRequest = resolvedView.asExpandedViewAggregation(request);

    result->resetToEmpty();

    if (auto txnRouter = TransactionRouter::get(opCtx)) {
        txnRouter.onViewResolutionError(opCtx, requestedNss);
    }

    // We pass both the underlying collection namespace and the view namespace here. The
    // underlying collection namespace is used to execute the aggregation on mongoD. Any cursor
    // returned will be registered under the view namespace so that subsequent getMore and
    // killCursors calls against the view have access.
    Namespaces nsStruct;
    nsStruct.requestedNss = requestedNss;
    nsStruct.executionNss = resolvedView.getNamespace();

    // For a sharded time-series collection, the routing is based on both routing table and the
    // bucketMaxSpanSeconds value. We need to make sure we use the bucketMaxSpanSeconds of the same
    // version as the routing table, instead of the one attached in the view error. This way the
    // shard versioning check can correctly catch stale routing information.
    boost::optional<CollectionRoutingInfo> snapshotCri;
    if (nsStruct.executionNss.isTimeseriesBucketsCollection()) {
        auto executionNsRoutingInfoStatus =
            sharded_agg_helpers::getExecutionNsRoutingInfo(opCtx, nsStruct.executionNss);
        if (executionNsRoutingInfoStatus.isOK()) {
            const auto& cri = executionNsRoutingInfoStatus.getValue();
            if (cri.cm.isSharded() && cri.cm.getTimeseriesFields()) {
                const auto patchedPipeline = rebuildPipelineWithTimeSeriesGranularity(
                    resolvedAggRequest.getPipeline(),
                    cri.cm.getTimeseriesFields()->getGranularity(),
                    cri.cm.getTimeseriesFields()->getBucketMaxSpanSeconds());
                resolvedAggRequest.setPipeline(patchedPipeline);
                snapshotCri = cri;
            }
        }
    }

    auto status = ClusterAggregate::runAggregate(
        opCtx, nsStruct, resolvedAggRequest, {resolvedAggRequest}, privileges, snapshotCri, result);

    // If the underlying namespace was changed to a view during retry, then re-run the aggregation
    // on the new resolved namespace.
    if (status.extraInfo<ResolvedView>()) {
        return ClusterAggregate::retryOnViewError(opCtx,
                                                  resolvedAggRequest,
                                                  *status.extraInfo<ResolvedView>(),
                                                  requestedNss,
                                                  privileges,
                                                  result,
                                                  numberRetries + 1);
    }

    return status;
}

}  // namespace mongo