1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
|
/**
* Copyright (C) 2015 MongoDB Inc.
*
* This program is free software: you can redistribute it and/or modify
* it under the terms of the GNU Affero General Public License, version 3,
* as published by the Free Software Foundation.
*
* 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
* GNU Affero General Public License for more details.
*
* You should have received a copy of the GNU Affero General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*
* 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 GNU Affero General 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 <iterator>
#include <map>
#include <set>
#include "mongo/bson/util/bson_extract.h"
#include "mongo/client/read_preference.h"
#include "mongo/client/remote_command_targeter.h"
#include "mongo/db/matcher/extensions_callback_noop.h"
#include "mongo/db/namespace_string.h"
#include "mongo/db/operation_context.h"
#include "mongo/db/query/collation/collation_index_key.h"
#include "mongo/db/query/collation/collator_factory_interface.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/rpc/get_status_from_command_result.h"
#include "mongo/s/balancer/balancer_configuration.h"
#include "mongo/s/catalog/catalog_cache.h"
#include "mongo/s/catalog/sharding_catalog_client.h"
#include "mongo/s/catalog/type_collection.h"
#include "mongo/s/chunk.h"
#include "mongo/s/chunk_diff.h"
#include "mongo/s/client/shard_registry.h"
#include "mongo/s/config.h"
#include "mongo/s/grid.h"
#include "mongo/s/shard_util.h"
#include "mongo/util/log.h"
#include "mongo/util/timer.h"
namespace mongo {
using std::make_pair;
using std::map;
using std::max;
using std::pair;
using std::set;
using std::shared_ptr;
using std::string;
using std::unique_ptr;
using std::vector;
namespace {
/**
* This is an adapter so we can use config diffs - mongos and mongod do them slightly
* differently
*
* The mongos adapter here tracks all shards, and stores ranges by (max, Chunk) in the map.
*/
class CMConfigDiffTracker : public ConfigDiffTracker<shared_ptr<Chunk>> {
public:
CMConfigDiffTracker(ChunkManager* manager) : _manager(manager) {}
bool isTracked(const ChunkType& chunk) const final {
// Mongos tracks all shards
return true;
}
bool isMinKeyIndexed() const final {
return false;
}
pair<BSONObj, shared_ptr<Chunk>> rangeFor(OperationContext* txn,
const ChunkType& chunk) const final {
shared_ptr<Chunk> c(new Chunk(txn, _manager, chunk));
return make_pair(chunk.getMax(), c);
}
ShardId shardFor(OperationContext* txn, const ShardId& shardId) const final {
const auto shard = grid.shardRegistry()->getShard(txn, shardId);
return shard->getId();
}
private:
ChunkManager* const _manager;
};
bool allOfType(BSONType type, const BSONObj& o) {
BSONObjIterator it(o);
while (it.more()) {
if (it.next().type() != type) {
return false;
}
}
return true;
}
bool isChunkMapValid(const ChunkMap& chunkMap) {
#define ENSURE(x) \
do { \
if (!(x)) { \
log() << "ChunkManager::_isValid failed: " #x; \
return false; \
} \
} while (0)
if (chunkMap.empty()) {
return true;
}
// Check endpoints
ENSURE(allOfType(MinKey, chunkMap.begin()->second->getMin()));
ENSURE(allOfType(MaxKey, boost::prior(chunkMap.end())->second->getMax()));
// Make sure there are no gaps or overlaps
for (ChunkMap::const_iterator it = boost::next(chunkMap.begin()), end = chunkMap.end();
it != end;
++it) {
ChunkMap::const_iterator last = boost::prior(it);
if (!(it->second->getMin() == last->second->getMax())) {
log() << last->second->toString();
log() << it->second->toString();
log() << it->second->getMin();
log() << last->second->getMax();
}
ENSURE(it->second->getMin() == last->second->getMax());
}
return true;
#undef ENSURE
}
} // namespace
AtomicUInt32 ChunkManager::NextSequenceNumber(1U);
ChunkManager::ChunkManager(const string& ns,
const ShardKeyPattern& pattern,
std::unique_ptr<CollatorInterface> defaultCollator,
bool unique)
: _ns(ns),
_keyPattern(pattern.getKeyPattern()),
_defaultCollator(std::move(defaultCollator)),
_unique(unique),
_sequenceNumber(NextSequenceNumber.addAndFetch(1)) {}
ChunkManager::ChunkManager(OperationContext* txn, const CollectionType& coll)
: _ns(coll.getNs().ns()),
_keyPattern(coll.getKeyPattern()),
_unique(coll.getUnique()),
_sequenceNumber(NextSequenceNumber.addAndFetch(1)) {
// coll does not have correct version. Use same initial version as _load and createFirstChunks.
_version = ChunkVersion(0, 0, coll.getEpoch());
if (!coll.getDefaultCollation().isEmpty()) {
auto statusWithCollator = CollatorFactoryInterface::get(txn->getServiceContext())
->makeFromBSON(coll.getDefaultCollation());
// The collation was validated upon collection creation.
invariantOK(statusWithCollator.getStatus());
_defaultCollator = std::move(statusWithCollator.getValue());
}
}
void ChunkManager::loadExistingRanges(OperationContext* txn, const ChunkManager* oldManager) {
int tries = 3;
while (tries--) {
ChunkMap chunkMap;
set<ShardId> shardIds;
ShardVersionMap shardVersions;
Timer t;
bool success = _load(txn, chunkMap, shardIds, &shardVersions, oldManager);
if (success) {
log() << "ChunkManager: time to load chunks for " << _ns << ": " << t.millis() << "ms"
<< " sequenceNumber: " << _sequenceNumber << " version: " << _version.toString()
<< " based on: "
<< (oldManager ? oldManager->getVersion().toString() : "(empty)");
// TODO: Merge into diff code above, so we validate in one place
if (isChunkMapValid(chunkMap)) {
_chunkMap.swap(chunkMap);
_shardIds.swap(shardIds);
_shardVersions.swap(shardVersions);
_chunkRangeMap = _constructRanges(_chunkMap);
return;
}
}
warning() << "ChunkManager loaded an invalid config for " << _ns << ", trying again";
sleepmillis(10 * (3 - tries));
}
// This will abort construction so we should never have a reference to an invalid config
msgasserted(13282,
str::stream() << "Couldn't load a valid config for " << _ns
<< " after 3 attempts. Please try again.");
}
bool ChunkManager::_load(OperationContext* txn,
ChunkMap& chunkMap,
set<ShardId>& shardIds,
ShardVersionMap* shardVersions,
const ChunkManager* oldManager) {
// Reset the max version, but not the epoch, when we aren't loading from the oldManager
_version = ChunkVersion(0, 0, _version.epoch());
// If we have a previous version of the ChunkManager to work from, use that info to reduce
// our config query
if (oldManager && oldManager->getVersion().isSet()) {
// Get the old max version
_version = oldManager->getVersion();
// Load a copy of the old versions
*shardVersions = oldManager->_shardVersions;
// Load a copy of the chunk map, replacing the chunk manager with our own
const ChunkMap& oldChunkMap = oldManager->getChunkMap();
// Could be v.expensive
// TODO: If chunks were immutable and didn't reference the manager, we could do more
// interesting things here
for (const auto& oldChunkMapEntry : oldChunkMap) {
shared_ptr<Chunk> oldC = oldChunkMapEntry.second;
shared_ptr<Chunk> newC(new Chunk(this,
oldC->getMin(),
oldC->getMax(),
oldC->getShardId(),
oldC->getLastmod(),
oldC->getBytesWritten()));
chunkMap.insert(make_pair(oldC->getMax(), newC));
}
LOG(2) << "loading chunk manager for collection " << _ns
<< " using old chunk manager w/ version " << _version.toString() << " and "
<< oldChunkMap.size() << " chunks";
}
// Attach a diff tracker for the versioned chunk data
CMConfigDiffTracker differ(this);
differ.attach(_ns, chunkMap, _version, *shardVersions);
// Diff tracker should *always* find at least one chunk if collection exists
// Get the diff query required
auto diffQuery = differ.configDiffQuery();
repl::OpTime opTime;
std::vector<ChunkType> chunks;
uassertStatusOK(grid.catalogClient(txn)->getChunks(
txn, diffQuery.query, diffQuery.sort, boost::none, &chunks, &opTime));
invariant(opTime >= _configOpTime);
_configOpTime = opTime;
int diffsApplied = differ.calculateConfigDiff(txn, chunks);
if (diffsApplied > 0) {
LOG(2) << "loaded " << diffsApplied << " chunks into new chunk manager for " << _ns
<< " with version " << _version;
// Add all existing shards we find to the shards set
for (ShardVersionMap::iterator it = shardVersions->begin(); it != shardVersions->end();) {
shared_ptr<Shard> shard = grid.shardRegistry()->getShard(txn, it->first);
if (shard) {
shardIds.insert(it->first);
++it;
} else {
shardVersions->erase(it++);
}
}
_configOpTime = opTime;
return true;
} else if (diffsApplied == 0) {
// No chunks were found for the ns
warning() << "no chunks found when reloading " << _ns << ", previous version was "
<< _version;
// Set all our data to empty
chunkMap.clear();
shardVersions->clear();
_version = ChunkVersion(0, 0, OID());
_configOpTime = opTime;
return true;
} else { // diffsApplied < 0
bool allInconsistent = (differ.numValidDiffs() == 0);
if (allInconsistent) {
// All versions are different, this can be normal
warning() << "major change in chunk information found when reloading " << _ns
<< ", previous version was " << _version;
} else {
// Inconsistent load halfway through (due to yielding cursor during load)
// should be rare
warning() << "inconsistent chunks found when reloading " << _ns
<< ", previous version was " << _version << ", this should be rare";
}
// Set all our data to empty to be extra safe
chunkMap.clear();
shardVersions->clear();
_version = ChunkVersion(0, 0, OID());
return allInconsistent;
}
}
shared_ptr<ChunkManager> ChunkManager::reload(OperationContext* txn, bool force) const {
const NamespaceString nss(_ns);
auto config = uassertStatusOK(grid.catalogCache()->getDatabase(txn, nss.db().toString()));
return config->getChunkManagerIfExists(txn, getns(), force);
}
void ChunkManager::_printChunks() const {
for (ChunkMap::const_iterator it = _chunkMap.begin(), end = _chunkMap.end(); it != end; ++it) {
log() << redact((*it->second).toString());
}
}
void ChunkManager::calcInitSplitsAndShards(OperationContext* txn,
const ShardId& primaryShardId,
const vector<BSONObj>* initPoints,
const set<ShardId>* initShardIds,
vector<BSONObj>* splitPoints,
vector<ShardId>* shardIds) const {
invariant(_chunkMap.empty());
if (!initPoints || initPoints->empty()) {
// discover split points
const auto primaryShard = grid.shardRegistry()->getShard(txn, primaryShardId);
const NamespaceString nss{getns()};
auto result = uassertStatusOK(
primaryShard->runCommand(txn,
ReadPreferenceSetting{ReadPreference::PrimaryPreferred},
nss.db().toString(),
BSON("count" << nss.coll()),
Shard::RetryPolicy::kIdempotent));
long long numObjects = 0;
uassertStatusOK(result.commandStatus);
uassertStatusOK(bsonExtractIntegerField(result.response, "n", &numObjects));
if (numObjects > 0) {
*splitPoints = uassertStatusOK(shardutil::selectChunkSplitPoints(
txn,
primaryShardId,
NamespaceString(_ns),
_keyPattern,
_keyPattern.getKeyPattern().globalMin(),
_keyPattern.getKeyPattern().globalMax(),
Grid::get(txn)->getBalancerConfiguration()->getMaxChunkSizeBytes(),
0,
0));
}
// since docs already exists, must use primary shard
shardIds->push_back(primaryShardId);
} else {
// make sure points are unique and ordered
set<BSONObj> orderedPts;
for (unsigned i = 0; i < initPoints->size(); ++i) {
BSONObj pt = (*initPoints)[i];
orderedPts.insert(pt);
}
for (set<BSONObj>::iterator it = orderedPts.begin(); it != orderedPts.end(); ++it) {
splitPoints->push_back(*it);
}
if (!initShardIds || !initShardIds->size()) {
// If not specified, only use the primary shard (note that it's not safe for mongos
// to put initial chunks on other shards without the primary mongod knowing).
shardIds->push_back(primaryShardId);
} else {
std::copy(initShardIds->begin(), initShardIds->end(), std::back_inserter(*shardIds));
}
}
}
Status ChunkManager::createFirstChunks(OperationContext* txn,
const ShardId& primaryShardId,
const vector<BSONObj>* initPoints,
const set<ShardId>* initShardIds) {
// TODO distlock?
// TODO: Race condition if we shard the collection and insert data while we split across
// the non-primary shard.
vector<BSONObj> splitPoints;
vector<ShardId> shardIds;
calcInitSplitsAndShards(txn, primaryShardId, initPoints, initShardIds, &splitPoints, &shardIds);
// this is the first chunk; start the versioning from scratch
ChunkVersion version(1, 0, OID::gen());
log() << "going to create " << splitPoints.size() + 1 << " chunk(s) for: " << _ns
<< " using new epoch " << version.epoch();
for (unsigned i = 0; i <= splitPoints.size(); i++) {
BSONObj min = i == 0 ? _keyPattern.getKeyPattern().globalMin() : splitPoints[i - 1];
BSONObj max =
i < splitPoints.size() ? splitPoints[i] : _keyPattern.getKeyPattern().globalMax();
ChunkType chunk;
chunk.setNS(_ns);
chunk.setMin(min);
chunk.setMax(max);
chunk.setShard(shardIds[i % shardIds.size()]);
chunk.setVersion(version);
Status status = grid.catalogClient(txn)->insertConfigDocument(
txn, ChunkType::ConfigNS, chunk.toBSON(), ShardingCatalogClient::kMajorityWriteConcern);
if (!status.isOK()) {
const string errMsg = str::stream() << "Creating first chunks failed: "
<< redact(status.reason());
error() << errMsg;
return Status(status.code(), errMsg);
}
version.incMinor();
}
_version = ChunkVersion(0, 0, version.epoch());
return Status::OK();
}
shared_ptr<Chunk> ChunkManager::findIntersectingChunk(OperationContext* txn,
const BSONObj& shardKey,
const CollatorInterface* collator) const {
{
if (collator) {
for (BSONElement elt : shardKey) {
if (CollationIndexKey::isCollatableType(elt.type())) {
msgasserted(ErrorCodes::ShardKeyNotFound,
"cannot target single shard due to collation");
}
}
}
BSONObj chunkMin;
shared_ptr<Chunk> chunk;
{
ChunkMap::const_iterator it = _chunkMap.upper_bound(shardKey);
if (it != _chunkMap.end()) {
chunkMin = it->first;
chunk = it->second;
}
}
if (chunk) {
if (chunk->containsKey(shardKey)) {
return chunk;
}
log() << redact(chunkMin.toString());
log() << redact((*chunk).toString());
log() << redact(shardKey);
reload(txn);
msgasserted(13141, "Chunk map pointed to incorrect chunk");
}
}
msgasserted(8070,
str::stream() << "couldn't find a chunk intersecting: " << shardKey << " for ns: "
<< _ns
<< " at version: "
<< _version.toString()
<< ", number of chunks: "
<< _chunkMap.size());
}
shared_ptr<Chunk> ChunkManager::findIntersectingChunkWithSimpleCollation(
OperationContext* txn, const BSONObj& shardKey) const {
const CollatorInterface* collator = nullptr;
return findIntersectingChunk(txn, shardKey, collator);
}
void ChunkManager::getShardIdsForQuery(OperationContext* txn,
const BSONObj& query,
const BSONObj& collation,
set<ShardId>* shardIds) const {
auto qr = stdx::make_unique<QueryRequest>(NamespaceString(_ns));
qr->setFilter(query);
if (!collation.isEmpty()) {
qr->setCollation(collation);
} else if (_defaultCollator) {
qr->setCollation(_defaultCollator->getSpec().toBSON());
}
auto statusWithCQ = CanonicalQuery::canonicalize(txn, std::move(qr), ExtensionsCallbackNoop());
uassertStatusOK(statusWithCQ.getStatus());
unique_ptr<CanonicalQuery> cq = std::move(statusWithCQ.getValue());
// Query validation
if (QueryPlannerCommon::hasNode(cq->root(), MatchExpression::GEO_NEAR)) {
uasserted(13501, "use geoNear command rather than $near query");
}
// Fast path for targeting equalities on the shard key.
auto shardKeyToFind = _keyPattern.extractShardKeyFromQuery(*cq);
if (shardKeyToFind.isOK() && !shardKeyToFind.getValue().isEmpty()) {
try {
auto chunk = findIntersectingChunk(txn, shardKeyToFind.getValue(), cq->getCollator());
shardIds->insert(chunk->getShardId());
return;
} catch (const MsgAssertionException& msg) {
// If we failed to find the intersecting chunk for a reason other than the inability to
// target a single shard, throw.
if (msg.getCode() != ErrorCodes::ShardKeyNotFound) {
throw msg;
}
}
}
// 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(_keyPattern.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 = _keyPattern.flattenBounds(bounds);
for (BoundList::const_iterator it = ranges.begin(); it != ranges.end(); ++it) {
getShardIdsForRange(*shardIds, it->first /*min*/, it->second /*max*/);
// once we know we need to visit all shards no need to keep looping
if (shardIds->size() == _shardIds.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(_chunkRangeMap.begin()->second.getShardId());
}
}
void ChunkManager::getShardIdsForRange(set<ShardId>& shardIds,
const BSONObj& min,
const BSONObj& max) const {
auto it = _chunkRangeMap.upper_bound(min);
auto end = _chunkRangeMap.upper_bound(max);
// The chunk range map must always cover the entire key space
invariant(it != _chunkRangeMap.end());
// We need to include the last chunk
if (end != _chunkRangeMap.cend()) {
++end;
}
for (; it != end; ++it) {
shardIds.insert(it->second.getShardId());
// No need to iterate through the rest of the ranges, because we already know we need to use
// all shards.
if (shardIds.size() == _shardIds.size()) {
break;
}
}
}
void ChunkManager::getAllShardIds(set<ShardId>* all) const {
dassert(all);
all->insert(_shardIds.begin(), _shardIds.end());
}
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;
}
// Consider shard key as an index
string accessMethod = IndexNames::findPluginName(key);
dassert(accessMethod == IndexNames::BTREE || accessMethod == IndexNames::HASHED);
// Use query framework to generate index bounds
QueryPlannerParams plannerParams;
// Must use "shard key" index
plannerParams.options = QueryPlannerParams::NO_TABLE_SCAN;
IndexEntry indexEntry(key,
accessMethod,
false /* multiKey */,
MultikeyPaths{},
false /* sparse */,
false /* unique */,
"shardkey",
NULL /* filterExpr */,
BSONObj(),
NULL /* collator */);
plannerParams.indices.push_back(indexEntry);
OwnedPointerVector<QuerySolution> solutions;
Status status = QueryPlanner::plan(canonicalQuery, plannerParams, &solutions.mutableVector());
uassert(status.code(), status.reason(), status.isOK());
IndexBounds bounds;
for (vector<QuerySolution*>::const_iterator it = solutions.begin();
bounds.size() == 0 && it != solutions.end();
it++) {
// Try next solution if we failed to generate index bounds, i.e. bounds.size() == 0
bounds = collapseQuerySolution((*it)->root.get());
}
if (bounds.size() == 0) {
// We cannot plan the query without collection scan, so target to all shards.
IndexBoundsBuilder::allValuesBounds(key, &bounds); // [minKey, maxKey]
}
return bounds;
}
IndexBounds ChunkManager::collapseQuerySolution(const QuerySolutionNode* node) {
if (node->children.size() == 0) {
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.
error() << "could not generate index bounds on query solution tree: "
<< 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 (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 ChunkManager::compatibleWith(const ChunkManager& 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).equals(getVersion(shardName));
}
ChunkVersion ChunkManager::getVersion(const ShardId& shardName) const {
ShardVersionMap::const_iterator i = _shardVersions.find(shardName);
if (i == _shardVersions.end()) {
// Shards without explicitly tracked shard versions (meaning they have
// no chunks) always have a version of (0, 0, epoch). Note this is
// *different* from the dropped chunk version of (0, 0, OID(000...)).
// See s/chunk_version.h.
return ChunkVersion(0, 0, _version.epoch());
}
return i->second;
}
ChunkVersion ChunkManager::getVersion() const {
return _version;
}
string ChunkManager::toString() const {
StringBuilder sb;
sb << "ChunkManager: " << _ns << " key:" << _keyPattern.toString() << '\n';
for (ChunkMap::const_iterator i = _chunkMap.begin(); i != _chunkMap.end(); ++i) {
sb << "\t" << i->second->toString() << '\n';
}
return sb.str();
}
ChunkManager::ChunkRangeMap ChunkManager::_constructRanges(const ChunkMap& chunkMap) {
ChunkRangeMap chunkRangeMap;
if (chunkMap.empty()) {
return chunkRangeMap;
}
ChunkMap::const_iterator current = chunkMap.cbegin();
while (current != chunkMap.cend()) {
const auto rangeFirst = current;
current = std::find_if(
current, chunkMap.cend(), [&rangeFirst](const ChunkMap::value_type& chunkMapEntry) {
return chunkMapEntry.second->getShardId() != rangeFirst->second->getShardId();
});
const auto rangeLast = std::prev(current);
const BSONObj rangeMin = rangeFirst->second->getMin();
const BSONObj rangeMax = rangeLast->second->getMax();
auto insertResult = chunkRangeMap.insert(std::make_pair(
rangeMax, ShardAndChunkRange(rangeMin, rangeMax, rangeFirst->second->getShardId())));
invariant(insertResult.second);
if (insertResult.first != chunkRangeMap.begin()) {
// Make sure there are no gaps in the ranges
insertResult.first--;
invariant(insertResult.first->first == rangeMin);
}
}
invariant(!chunkRangeMap.empty());
invariant(allOfType(MinKey, chunkRangeMap.begin()->second.getMin()));
invariant(allOfType(MaxKey, chunkRangeMap.rbegin()->first));
return chunkRangeMap;
}
uint64_t ChunkManager::getCurrentDesiredChunkSize() const {
// split faster in early chunks helps spread out an initial load better
const uint64_t minChunkSize = 1 << 20; // 1 MBytes
uint64_t splitThreshold = grid.getBalancerConfiguration()->getMaxChunkSizeBytes();
int nc = numChunks();
if (nc <= 1) {
return 1024;
} else if (nc < 3) {
return minChunkSize / 2;
} else if (nc < 10) {
splitThreshold = max(splitThreshold / 4, minChunkSize);
} else if (nc < 20) {
splitThreshold = max(splitThreshold / 2, minChunkSize);
}
return splitThreshold;
}
repl::OpTime ChunkManager::getConfigOpTime() const {
return _configOpTime;
}
} // namespace mongo
|