summaryrefslogtreecommitdiff
path: root/src/mongo/s/chunk_manager.cpp
blob: a8af892ff28f9b743cb436a57654b2238dd2d378 (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
/**
 *    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/s/chunk_manager.h"

#include "mongo/base/owned_pointer_vector.h"
#include "mongo/bson/simple_bsonobj_comparator.h"
#include "mongo/db/matcher/extensions_callback_noop.h"
#include "mongo/db/query/collation/collation_index_key.h"
#include "mongo/db/query/index_bounds_builder.h"
#include "mongo/db/query/query_planner.h"
#include "mongo/db/query/query_planner_common.h"
#include "mongo/db/storage/key_string.h"
#include "mongo/logv2/log.h"
#include "mongo/s/chunk_writes_tracker.h"
#include "mongo/s/shard_invalidated_for_targeting_exception.h"
#include "mongo/util/log.h"

namespace mongo {
namespace {

// Used to generate sequence numbers to assign to each newly created RoutingTableHistory
AtomicWord<unsigned> nextCMSequenceNumber(0);

void checkAllElementsAreOfType(BSONType type, const BSONObj& o) {
    for (auto&& element : o) {
        uassert(ErrorCodes::ConflictingOperationInProgress,
                str::stream() << "Not all elements of " << o << " are of type " << typeName(type),
                element.type() == type);
    }
}

std::string extractKeyStringInternal(const BSONObj& shardKeyValue, Ordering ordering) {
    BSONObjBuilder strippedKeyValue;
    for (const auto& elem : shardKeyValue) {
        strippedKeyValue.appendAs(elem, ""_sd);
    }

    KeyString::Builder ks(KeyString::Version::V1, strippedKeyValue.done(), ordering);
    return {ks.getBuffer(), ks.getSize()};
}

}  // namespace

ShardVersionTargetingInfo::ShardVersionTargetingInfo(const OID& epoch)
    : shardVersion(0, 0, epoch) {}

RoutingTableHistory::RoutingTableHistory(NamespaceString nss,
                                         boost::optional<UUID> uuid,
                                         KeyPattern shardKeyPattern,
                                         std::unique_ptr<CollatorInterface> defaultCollator,
                                         bool unique,
                                         ChunkInfoMap chunkMap,
                                         ChunkVersion collectionVersion)
    : _sequenceNumber(nextCMSequenceNumber.addAndFetch(1)),
      _nss(std::move(nss)),
      _uuid(uuid),
      _shardKeyPattern(shardKeyPattern),
      _shardKeyOrdering(Ordering::make(_shardKeyPattern.toBSON())),
      _defaultCollator(std::move(defaultCollator)),
      _unique(unique),
      _chunkMap(std::move(chunkMap)),
      _collectionVersion(collectionVersion),
      _shardVersions(_constructShardVersionMap()) {}

void RoutingTableHistory::setShardStale(const ShardId& shardId) {
    auto it = _shardVersions.find(shardId);
    if (it != _shardVersions.end()) {
        it->second.isStale.store(true);
    }
}

void RoutingTableHistory::setAllShardsRefreshed() {
    for (auto& [shard, targetingInfo] : _shardVersions) {
        targetingInfo.isStale.store(false);
    }
}

Chunk ChunkManager::findIntersectingChunk(const BSONObj& shardKey, const BSONObj& collation) const {
    const bool hasSimpleCollation = (collation.isEmpty() && !_rt->getDefaultCollator()) ||
        SimpleBSONObjComparator::kInstance.evaluate(collation == CollationSpec::kSimpleSpec);
    if (!hasSimpleCollation) {
        for (BSONElement elt : shardKey) {
            uassert(ErrorCodes::ShardKeyNotFound,
                    str::stream() << "Cannot target single shard due to collation of key "
                                  << elt.fieldNameStringData() << " for namespace " << getns(),
                    !CollationIndexKey::isCollatableType(elt.type()));
        }
    }

    const auto it = _rt->getChunkMap().upper_bound(_rt->_extractKeyString(shardKey));
    uassert(ErrorCodes::ShardKeyNotFound,
            str::stream() << "Cannot target single shard using key " << shardKey
                          << " for namespace " << getns(),
            it != _rt->getChunkMap().end() && it->second->containsKey(shardKey));

    return Chunk(*(it->second), _clusterTime);
}

bool ChunkManager::keyBelongsToShard(const BSONObj& shardKey, const ShardId& shardId) const {
    if (shardKey.isEmpty())
        return false;

    const auto it = _rt->getChunkMap().upper_bound(_rt->_extractKeyString(shardKey));
    if (it == _rt->getChunkMap().end())
        return false;

    invariant(it->second->containsKey(shardKey));

    return it->second->getShardIdAt(_clusterTime) == shardId;
}

void ChunkManager::getShardIdsForQuery(OperationContext* opCtx,
                                       const BSONObj& query,
                                       const BSONObj& collation,
                                       std::set<ShardId>* shardIds) const {
    auto qr = std::make_unique<QueryRequest>(_rt->getns());
    qr->setFilter(query);

    if (!collation.isEmpty()) {
        qr->setCollation(collation);
    } else if (_rt->getDefaultCollator()) {
        qr->setCollation(_rt->getDefaultCollator()->getSpec().toBSON());
    }

    const boost::intrusive_ptr<ExpressionContext> expCtx;
    auto cq = uassertStatusOK(
        CanonicalQuery::canonicalize(opCtx,
                                     std::move(qr),
                                     expCtx,
                                     ExtensionsCallbackNoop(),
                                     MatchExpressionParser::kAllowAllSpecialFeatures));

    // Fast path for targeting equalities on the shard key.
    auto shardKeyToFind = _rt->getShardKeyPattern().extractShardKeyFromQuery(*cq);
    if (!shardKeyToFind.isEmpty()) {
        try {
            auto chunk = findIntersectingChunk(shardKeyToFind, collation);
            shardIds->insert(chunk.getShardId());
            return;
        } catch (const DBException&) {
            // The query uses multiple shards
        }
    }

    // Transforms query into bounds for each field in the shard key
    // for example :
    //   Key { a: 1, b: 1 },
    //   Query { a : { $gte : 1, $lt : 2 },
    //            b : { $gte : 3, $lt : 4 } }
    //   => Bounds { a : [1, 2), b : [3, 4) }
    IndexBounds bounds = getIndexBoundsForQuery(_rt->getShardKeyPattern().toBSON(), *cq);

    // Transforms bounds for each shard key field into full shard key ranges
    // for example :
    //   Key { a : 1, b : 1 }
    //   Bounds { a : [1, 2), b : [3, 4) }
    //   => Ranges { a : 1, b : 3 } => { a : 2, b : 4 }
    BoundList ranges = _rt->getShardKeyPattern().flattenBounds(bounds);

    for (BoundList::const_iterator it = ranges.begin(); it != ranges.end(); ++it) {
        getShardIdsForRange(it->first /*min*/, it->second /*max*/, shardIds);

        // once we know we need to visit all shards no need to keep looping
        if (shardIds->size() == _rt->_shardVersions.size()) {
            break;
        }
    }

    // SERVER-4914 Some clients of getShardIdsForQuery() assume at least one shard will be returned.
    // For now, we satisfy that assumption by adding a shard with no matches rather than returning
    // an empty set of shards.
    if (shardIds->empty()) {
        shardIds->insert(_rt->getChunkMap().begin()->second->getShardIdAt(_clusterTime));
    }
}

void ChunkManager::getShardIdsForRange(const BSONObj& min,
                                       const BSONObj& max,
                                       std::set<ShardId>* shardIds) const {
    const auto bounds = _rt->overlappingRanges(min, max, true);
    for (auto it = bounds.first; it != bounds.second; ++it) {
        shardIds->insert(it->second->getShardIdAt(_clusterTime));

        // No need to iterate through the rest of the ranges, because we already know we need to use
        // all shards.
        if (shardIds->size() == _rt->_shardVersions.size()) {
            break;
        }
    }
}

bool ChunkManager::rangeOverlapsShard(const ChunkRange& range, const ShardId& shardId) const {
    const auto bounds = _rt->overlappingRanges(range.getMin(), range.getMax(), false);
    const auto it = std::find_if(bounds.first, bounds.second, [this, &shardId](const auto& scr) {
        return scr.second->getShardIdAt(_clusterTime) == shardId;
    });

    return it != bounds.second;
}

ChunkManager::ConstRangeOfChunks ChunkManager::getNextChunkOnShard(const BSONObj& shardKey,
                                                                   const ShardId& shardId) const {
    for (auto it = _rt->getChunkMap().upper_bound(_rt->_extractKeyString(shardKey));
         it != _rt->getChunkMap().end();
         ++it) {
        const auto& chunk = it->second;
        if (chunk->getShardIdAt(_clusterTime) == shardId) {
            const auto begin = it;
            const auto end = ++it;
            return {ConstChunkIterator(begin, _clusterTime), ConstChunkIterator(end, _clusterTime)};
        }
    }

    return {ConstChunkIterator(), ConstChunkIterator()};
}

void RoutingTableHistory::getAllShardIds(std::set<ShardId>* all) const {
    std::transform(_shardVersions.begin(),
                   _shardVersions.end(),
                   std::inserter(*all, all->begin()),
                   [](const ShardVersionMap::value_type& pair) { return pair.first; });
}

int RoutingTableHistory::getNShardsOwningChunks() const {
    return _shardVersions.size();
}

std::pair<ChunkInfoMap::const_iterator, ChunkInfoMap::const_iterator>
RoutingTableHistory::overlappingRanges(const BSONObj& min,
                                       const BSONObj& max,
                                       bool isMaxInclusive) const {

    const auto itMin = _chunkMap.upper_bound(_extractKeyString(min));
    const auto itMax = [this, &max, isMaxInclusive]() {
        auto it = isMaxInclusive ? _chunkMap.upper_bound(_extractKeyString(max))
                                 : _chunkMap.lower_bound(_extractKeyString(max));
        return it == _chunkMap.end() ? it : ++it;
    }();

    return {itMin, itMax};
}

IndexBounds ChunkManager::getIndexBoundsForQuery(const BSONObj& key,
                                                 const CanonicalQuery& canonicalQuery) {
    // $text is not allowed in planning since we don't have text index on mongos.
    // TODO: Treat $text query as a no-op in planning on mongos. So with shard key {a: 1},
    //       the query { a: 2, $text: { ... } } will only target to {a: 2}.
    if (QueryPlannerCommon::hasNode(canonicalQuery.root(), MatchExpression::TEXT)) {
        IndexBounds bounds;
        IndexBoundsBuilder::allValuesBounds(key, &bounds);  // [minKey, maxKey]
        return bounds;
    }

    // Similarly, ignore GEO_NEAR queries in planning, since we do not have geo indexes on mongos.
    if (QueryPlannerCommon::hasNode(canonicalQuery.root(), MatchExpression::GEO_NEAR)) {
        IndexBounds bounds;
        IndexBoundsBuilder::allValuesBounds(key, &bounds);
        return bounds;
    }

    // Consider shard key as an index
    std::string accessMethod = IndexNames::findPluginName(key);
    dassert(accessMethod == IndexNames::BTREE || accessMethod == IndexNames::HASHED);
    const auto indexType = IndexNames::nameToType(accessMethod);

    // Use query framework to generate index bounds
    QueryPlannerParams plannerParams;
    // Must use "shard key" index
    plannerParams.options = QueryPlannerParams::NO_TABLE_SCAN;
    IndexEntry indexEntry(key,
                          indexType,
                          // The shard key index cannot be multikey.
                          false,
                          // Empty multikey paths, since the shard key index cannot be multikey.
                          MultikeyPaths{},
                          // Empty multikey path set, since the shard key index cannot be multikey.
                          {},
                          false /* sparse */,
                          false /* unique */,
                          IndexEntry::Identifier{"shardkey"},
                          nullptr /* filterExpr */,
                          BSONObj(),
                          nullptr, /* collator */
                          nullptr /* projExec */);
    plannerParams.indices.push_back(std::move(indexEntry));

    auto plannerResult = QueryPlanner::plan(canonicalQuery, plannerParams);
    if (plannerResult.getStatus().code() != ErrorCodes::NoQueryExecutionPlans) {
        auto solutions = uassertStatusOK(std::move(plannerResult));

        // Pick any solution that has non-trivial IndexBounds. bounds.size() == 0 represents a
        // trivial IndexBounds where none of the fields' values are bounded.
        for (auto&& soln : solutions) {
            IndexBounds bounds = collapseQuerySolution(soln->root.get());
            if (bounds.size() > 0) {
                return bounds;
            }
        }
    }

    // We cannot plan the query without collection scan, so target to all shards.
    IndexBounds bounds;
    IndexBoundsBuilder::allValuesBounds(key, &bounds);  // [minKey, maxKey]
    return bounds;
}

IndexBounds ChunkManager::collapseQuerySolution(const QuerySolutionNode* node) {
    if (node->children.empty()) {
        invariant(node->getType() == STAGE_IXSCAN);

        const IndexScanNode* ixNode = static_cast<const IndexScanNode*>(node);
        return ixNode->bounds;
    }

    if (node->children.size() == 1) {
        // e.g. FETCH -> IXSCAN
        return collapseQuerySolution(node->children.front());
    }

    // children.size() > 1, assert it's OR / SORT_MERGE.
    if (node->getType() != STAGE_OR && node->getType() != STAGE_SORT_MERGE) {
        // Unexpected node. We should never reach here.
        LOGV2_ERROR(23833,
                    "could not generate index bounds on query solution tree: {node}",
                    "node"_attr = redact(node->toString()));
        dassert(false);  // We'd like to know this error in testing.

        // Bail out with all shards in production, since this isn't a fatal error.
        return IndexBounds();
    }

    IndexBounds bounds;

    for (std::vector<QuerySolutionNode*>::const_iterator it = node->children.begin();
         it != node->children.end();
         it++) {
        // The first branch under OR
        if (it == node->children.begin()) {
            invariant(bounds.size() == 0);
            bounds = collapseQuerySolution(*it);
            if (bounds.size() == 0) {  // Got unexpected node in query solution tree
                return IndexBounds();
            }
            continue;
        }

        IndexBounds childBounds = collapseQuerySolution(*it);
        if (childBounds.size() == 0) {
            // Got unexpected node in query solution tree
            return IndexBounds();
        }

        invariant(childBounds.size() == bounds.size());

        for (size_t i = 0; i < bounds.size(); i++) {
            bounds.fields[i].intervals.insert(bounds.fields[i].intervals.end(),
                                              childBounds.fields[i].intervals.begin(),
                                              childBounds.fields[i].intervals.end());
        }
    }

    for (size_t i = 0; i < bounds.size(); i++) {
        IndexBoundsBuilder::unionize(&bounds.fields[i]);
    }

    return bounds;
}

bool RoutingTableHistory::compatibleWith(const RoutingTableHistory& other,
                                         const ShardId& shardName) const {
    // Return true if the shard version is the same in the two chunk managers
    // TODO: This doesn't need to be so strong, just major vs
    return other.getVersion(shardName) == getVersion(shardName);
}

ChunkVersion RoutingTableHistory::getVersion(const ShardId& shardName) const {
    auto it = _shardVersions.find(shardName);
    if (it == _shardVersions.end()) {
        // Shards without explicitly tracked shard versions (meaning they have no chunks) always
        // have a version of (0, 0, epoch)
        return ChunkVersion(0, 0, _collectionVersion.epoch());
    }

    uassert(ShardInvalidatedForTargetingInfo(_nss),
            "shard has been marked stale",
            !it->second.isStale.load());

    return it->second.shardVersion;
}

std::string RoutingTableHistory::toString() const {
    StringBuilder sb;
    sb << "RoutingTableHistory: " << _nss.ns() << " key: " << _shardKeyPattern.toString() << '\n';

    sb << "Chunks:\n";
    for (const auto& chunk : _chunkMap) {
        sb << "\t" << chunk.second->toString() << '\n';
    }

    sb << "Shard versions:\n";
    for (const auto& entry : _shardVersions) {
        sb << "\t" << entry.first << ": " << entry.second.shardVersion.toString() << '\n';
    }

    return sb.str();
}

ShardVersionMap RoutingTableHistory::_constructShardVersionMap() const {
    const OID& epoch = _collectionVersion.epoch();

    ShardVersionMap shardVersions;
    ChunkInfoMap::const_iterator current = _chunkMap.cbegin();

    boost::optional<BSONObj> firstMin = boost::none;
    boost::optional<BSONObj> lastMax = boost::none;

    while (current != _chunkMap.cend()) {
        const auto& firstChunkInRange = current->second;
        const auto& currentRangeShardId = firstChunkInRange->getShardIdAt(boost::none);

        // Tracks the max shard version for the shard on which the current range will reside
        auto shardVersionIt = shardVersions.find(currentRangeShardId);
        if (shardVersionIt == shardVersions.end()) {
            shardVersionIt = shardVersions.emplace(currentRangeShardId, epoch).first;
        }

        auto& maxShardVersion = shardVersionIt->second.shardVersion;

        current =
            std::find_if(current,
                         _chunkMap.cend(),
                         [&currentRangeShardId,
                          &maxShardVersion](const ChunkInfoMap::value_type& chunkMapEntry) {
                             const auto& currentChunk = chunkMapEntry.second;

                             if (currentChunk->getShardIdAt(boost::none) != currentRangeShardId)
                                 return true;

                             if (currentChunk->getLastmod() > maxShardVersion)
                                 maxShardVersion = currentChunk->getLastmod();

                             return false;
                         });

        const auto rangeLast = std::prev(current);

        const auto& rangeMin = firstChunkInRange->getMin();
        const auto& rangeMax = rangeLast->second->getMax();

        // Check the continuity of the chunks map
        if (lastMax && !SimpleBSONObjComparator::kInstance.evaluate(*lastMax == rangeMin)) {
            if (SimpleBSONObjComparator::kInstance.evaluate(*lastMax < rangeMin))
                uasserted(ErrorCodes::ConflictingOperationInProgress,
                          str::stream()
                              << "Gap exists in the routing table between chunks "
                              << _chunkMap.at(_extractKeyString(*lastMax))->getRange().toString()
                              << " and " << rangeLast->second->getRange().toString());
            else
                uasserted(ErrorCodes::ConflictingOperationInProgress,
                          str::stream()
                              << "Overlap exists in the routing table between chunks "
                              << _chunkMap.at(_extractKeyString(*lastMax))->getRange().toString()
                              << " and " << rangeLast->second->getRange().toString());
        }

        if (!firstMin)
            firstMin = rangeMin;

        lastMax = rangeMax;

        // If a shard has chunks it must have a shard version, otherwise we have an invalid chunk
        // somewhere, which should have been caught at chunk load time
        invariant(maxShardVersion.isSet());
    }

    if (!_chunkMap.empty()) {
        invariant(!shardVersions.empty());
        invariant(firstMin.is_initialized());
        invariant(lastMax.is_initialized());

        checkAllElementsAreOfType(MinKey, firstMin.get());
        checkAllElementsAreOfType(MaxKey, lastMax.get());
    }

    return shardVersions;
}

std::string RoutingTableHistory::_extractKeyString(const BSONObj& shardKeyValue) const {
    return extractKeyStringInternal(shardKeyValue, _shardKeyOrdering);
}

std::shared_ptr<RoutingTableHistory> RoutingTableHistory::makeNew(
    NamespaceString nss,
    boost::optional<UUID> uuid,
    KeyPattern shardKeyPattern,
    std::unique_ptr<CollatorInterface> defaultCollator,
    bool unique,
    OID epoch,
    const std::vector<ChunkType>& chunks) {
    return RoutingTableHistory(std::move(nss),
                               std::move(uuid),
                               std::move(shardKeyPattern),
                               std::move(defaultCollator),
                               std::move(unique),
                               {},
                               {0, 0, epoch})
        .makeUpdated(chunks);
}

std::shared_ptr<RoutingTableHistory> RoutingTableHistory::makeUpdated(
    const std::vector<ChunkType>& changedChunks) {

    const auto startingCollectionVersion = getVersion();
    auto chunkMap = _chunkMap;

    ChunkVersion collectionVersion = startingCollectionVersion;
    for (const auto& chunk : changedChunks) {
        const auto& chunkVersion = chunk.getVersion();

        uassert(ErrorCodes::ConflictingOperationInProgress,
                str::stream() << "Chunk with namespace " << chunk.getNS().ns() << " and min key "
                              << chunk.getMin()
                              << " has epoch different from that of the collection "
                              << chunkVersion.epoch(),
                collectionVersion.epoch() == chunkVersion.epoch());

        // Chunks must always come in incrementally sorted order
        invariant(chunkVersion >= collectionVersion);
        collectionVersion = chunkVersion;

        const auto chunkMinKeyString = _extractKeyString(chunk.getMin());
        const auto chunkMaxKeyString = _extractKeyString(chunk.getMax());

        // Returns the first chunk with a max key that is > min - implies that the chunk overlaps
        // min
        const auto low = chunkMap.upper_bound(chunkMinKeyString);

        // Returns the first chunk with a max key that is > max - implies that the next chunk cannot
        // not overlap max
        const auto high = chunkMap.upper_bound(chunkMaxKeyString);

        // If we are in the middle of splitting a chunk, for the first few
        // chunks inserted, low == high, because both lookups will point to the
        // same chunk (the one being split). If we're inserting the last chunk
        // for the current chunk being split, low will point to the chunk that
        // we're splitting, and high will point to the next chunk past the one
        // we're splitting (which could be chunkMap.end()). In this case,
        // std::distance(low, high) == 1. Lastly, this does not apply during
        // the creation of the original routing table, in which case the map is
        // empty and the first chunk that is inserted will find that low ==
        // high, but low == chunkMap.end(), and we aren't doing a split in that
        // case.
        auto foundSingleChunk =
            ((low == high || std::distance(low, high) == 1) && low != chunkMap.end());

        auto newChunk = std::make_shared<ChunkInfo>(chunk);
        if (foundSingleChunk) {
            auto chunkBeingReplacedBySplit = low->second;
            auto bytesInReplacedChunk =
                chunkBeingReplacedBySplit->getWritesTracker()->getBytesWritten();
            newChunk->getWritesTracker()->addBytesWritten(bytesInReplacedChunk);
        }

        // Erase all chunks from the map, which overlap the chunk we got from the persistent store
        chunkMap.erase(low, high);

        // Insert only the chunk itself
        chunkMap.insert(std::make_pair(chunkMaxKeyString, newChunk));
    }

    // If at least one diff was applied, the metadata is correct, but it might not have changed so
    // in this case there is no need to recreate the chunk manager.
    //
    // NOTE: In addition to the above statement, it is also important that we return the same chunk
    // manager object, because the write commands' code relies on changes of the chunk manager's
    // sequence number to detect batch writes not making progress because of chunks moving across
    // shards too frequently.
    if (collectionVersion == startingCollectionVersion) {
        setAllShardsRefreshed();
        return shared_from_this();
    }

    return std::shared_ptr<RoutingTableHistory>(
        new RoutingTableHistory(_nss,
                                _uuid,
                                KeyPattern(getShardKeyPattern().getKeyPattern()),
                                CollatorInterface::cloneCollator(getDefaultCollator()),
                                isUnique(),
                                std::move(chunkMap),
                                collectionVersion));
}

}  // namespace mongo