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
|
/**
* Copyright (C) 2013 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/s/chunk_manager_targeter.h"
#include "mongo/s/config.h"
#include "mongo/s/grid.h"
#include "mongo/util/log.h"
#include "mongo/util/mongoutils/str.h"
namespace mongo {
using std::endl;
using std::map;
using std::set;
using std::string;
using std::vector;
using mongoutils::str::stream;
/**
* Helper to get the DBConfigPtr object in an exception-safe way.
*/
static bool getDBConfigSafe( StringData db, DBConfigPtr& config, string* errMsg ) {
try {
config = grid.getDBConfig( db, true );
if ( !config ) *errMsg = stream() << "could not load or create database " << db;
}
catch ( const DBException& ex ) {
*errMsg = ex.toString();
}
return config.get();
}
ChunkManagerTargeter::ChunkManagerTargeter() :
_needsTargetingRefresh( false ), _stats( new TargeterStats ) {
}
Status ChunkManagerTargeter::init( const NamespaceString& nss ) {
_nss = nss;
//
// Get the latest metadata information from the cache
//
DBConfigPtr config;
string errMsg;
if ( !getDBConfigSafe( _nss.db(), config, &errMsg ) ) {
return Status( ErrorCodes::DatabaseNotFound, errMsg );
}
// Get either the chunk manager or primary shard
config->getChunkManagerOrPrimary( _nss.ns(), _manager, _primary );
return Status::OK();
}
const NamespaceString& ChunkManagerTargeter::getNS() const {
return _nss;
}
Status ChunkManagerTargeter::targetInsert( const BSONObj& doc,
ShardEndpoint** endpoint ) const {
BSONObj shardKey;
if ( _manager ) {
//
// Sharded collections have the following requirements for targeting:
//
// Inserts must contain the exact shard key.
//
shardKey = _manager->getShardKeyPattern().extractShardKeyFromDoc(doc);
// Check shard key exists
if (shardKey.isEmpty()) {
return Status(ErrorCodes::ShardKeyNotFound,
stream() << "document " << doc
<< " does not contain shard key for pattern "
<< _manager->getShardKeyPattern().toString());
}
// Check shard key size on insert
Status status = ShardKeyPattern::checkShardKeySize(shardKey);
if (!status.isOK())
return status;
}
// Target the shard key or database primary
if (!shardKey.isEmpty()) {
return targetShardKey(shardKey, doc.objsize(), endpoint);
}
else {
if (!_primary) {
return Status(ErrorCodes::NamespaceNotFound,
str::stream() << "could not target insert in collection "
<< getNS().ns() << "; no metadata found");
}
*endpoint = new ShardEndpoint(_primary->getName(), ChunkVersion::UNSHARDED());
return Status::OK();
}
}
namespace {
// TODO: Expose these for unit testing via dbtests
enum UpdateType {
UpdateType_Replacement, UpdateType_OpStyle, UpdateType_Unknown
};
/**
* There are two styles of update expressions:
* coll.update({ x : 1 }, { y : 2 }) // Replacement style
* coll.update({ x : 1 }, { $set : { y : 2 } }) // OpStyle
*/
UpdateType getUpdateExprType( const BSONObj& updateExpr ) {
UpdateType updateType = UpdateType_Unknown;
// Empty update is replacement-style, by default
if ( updateExpr.isEmpty() ) return UpdateType_Replacement;
BSONObjIterator it( updateExpr );
while ( it.more() ) {
BSONElement next = it.next();
if ( next.fieldName()[0] == '$' ) {
if ( updateType == UpdateType_Unknown ) {
updateType = UpdateType_OpStyle;
}
else if ( updateType == UpdateType_Replacement ) {
return UpdateType_Unknown;
}
}
else {
if ( updateType == UpdateType_Unknown ) {
updateType = UpdateType_Replacement;
}
else if ( updateType == UpdateType_OpStyle ) {
return UpdateType_Unknown;
}
}
}
return updateType;
}
/**
* This returns "does the query have an _id field" and "is the _id field
* querying for a direct value like _id : 3 and not _id : { $gt : 3 }"
*
* Ex: { _id : 1 } => true
* { foo : <anything>, _id : 1 } => true
* { _id : { $lt : 30 } } => false
* { foo : <anything> } => false
*/
bool isExactIdQuery(const BSONObj& query) {
static const ShardKeyPattern virtualIdShardKey(BSON("_id" << 1));
StatusWith<BSONObj> status = virtualIdShardKey.extractShardKeyFromQuery(query);
if (!status.isOK())
return false;
return !status.getValue()["_id"].eoo();
}
}
Status ChunkManagerTargeter::targetUpdate( const BatchedUpdateDocument& updateDoc,
vector<ShardEndpoint*>* endpoints ) const {
//
// Update targeting may use either the query or the update. This is to support save-style
// updates, of the form:
//
// coll.update({ _id : xxx }, { _id : xxx, shardKey : 1, foo : bar }, { upsert : true })
//
// Because drivers do not know the shard key, they can't pull the shard key automatically
// into the query doc, and to correctly support upsert we must target a single shard.
//
// The rule is simple - If the update is replacement style (no '$set'), we target using the
// update. If the update is replacement style, we target using the query.
//
// If we have the exact shard key in either the query or replacement doc, we target using
// that extracted key.
//
BSONObj query = updateDoc.getQuery();
BSONObj updateExpr = updateDoc.getUpdateExpr();
UpdateType updateType = getUpdateExprType( updateDoc.getUpdateExpr() );
if ( updateType == UpdateType_Unknown ) {
return Status( ErrorCodes::UnsupportedFormat,
stream() << "update document " << updateExpr
<< " has mixed $operator and non-$operator style fields" );
}
BSONObj shardKey;
if ( _manager ) {
//
// Sharded collections have the following futher requirements for targeting:
//
// Upserts must be targeted exactly by shard key.
// Non-multi updates must be targeted exactly by shard key *or* exact _id.
//
// Get the shard key
if (updateType == UpdateType_OpStyle) {
// Target using the query
StatusWith<BSONObj> status =
_manager->getShardKeyPattern().extractShardKeyFromQuery(query);
// Bad query
if (!status.isOK())
return status.getStatus();
shardKey = status.getValue();
}
else {
// Target using the replacement document
shardKey = _manager->getShardKeyPattern().extractShardKeyFromDoc(updateExpr);
}
//
// Extra sharded update validation
//
if (updateDoc.getUpsert()) {
// Sharded upserts *always* need to be exactly targeted by shard key
if (shardKey.isEmpty()) {
return Status(ErrorCodes::ShardKeyNotFound,
stream() << "upsert " << updateDoc.toBSON()
<< " does not contain shard key for pattern "
<< _manager->getShardKeyPattern().toString());
}
// Also check shard key size on upsert
Status status = ShardKeyPattern::checkShardKeySize(shardKey);
if (!status.isOK())
return status;
}
// Validate that single (non-multi) sharded updates are targeted by shard key or _id
if (!updateDoc.getMulti() && shardKey.isEmpty()
&& !isExactIdQuery(updateDoc.getQuery())) {
return Status(ErrorCodes::ShardKeyNotFound,
stream() << "update " << updateDoc.toBSON()
<< " does not contain _id or shard key for pattern "
<< _manager->getShardKeyPattern().toString());
}
}
// Target the shard key, query, or replacement doc
if (!shardKey.isEmpty()) {
// We can't rely on our query targeting to be exact
ShardEndpoint* endpoint = NULL;
Status result = targetShardKey(shardKey,
(query.objsize() + updateExpr.objsize()),
&endpoint);
endpoints->push_back(endpoint);
return result;
}
else if (updateType == UpdateType_OpStyle) {
return targetQuery(query, endpoints);
}
else {
return targetDoc(updateExpr, endpoints);
}
}
Status ChunkManagerTargeter::targetDelete( const BatchedDeleteDocument& deleteDoc,
vector<ShardEndpoint*>* endpoints ) const {
BSONObj shardKey;
if ( _manager ) {
//
// Sharded collections have the following further requirements for targeting:
//
// Limit-1 deletes must be targeted exactly by shard key *or* exact _id
//
// Get the shard key
StatusWith<BSONObj> status =
_manager->getShardKeyPattern().extractShardKeyFromQuery(deleteDoc.getQuery());
// Bad query
if (!status.isOK())
return status.getStatus();
shardKey = status.getValue();
// Validate that single (limit-1) sharded deletes are targeted by shard key or _id
if (deleteDoc.getLimit() == 1 && shardKey.isEmpty()
&& !isExactIdQuery(deleteDoc.getQuery())) {
return Status(ErrorCodes::ShardKeyNotFound,
stream() << "delete " << deleteDoc.toBSON()
<< " does not contain _id or shard key for pattern "
<< _manager->getShardKeyPattern().toString());
}
}
// Target the shard key or delete query
if (!shardKey.isEmpty()) {
// We can't rely on our query targeting to be exact
ShardEndpoint* endpoint = NULL;
Status result = targetShardKey(shardKey, 0, &endpoint);
endpoints->push_back(endpoint);
return result;
}
else {
return targetQuery(deleteDoc.getQuery(), endpoints);
}
}
Status ChunkManagerTargeter::targetDoc(const BSONObj& doc,
vector<ShardEndpoint*>* endpoints) const {
// NOTE: This is weird and fragile, but it's the way our language works right now -
// documents are either A) invalid or B) valid equality queries over themselves.
return targetQuery(doc, endpoints);
}
Status ChunkManagerTargeter::targetQuery( const BSONObj& query,
vector<ShardEndpoint*>* endpoints ) const {
if ( !_primary && !_manager ) {
return Status( ErrorCodes::NamespaceNotFound,
str::stream() << "could not target query in "
<< getNS().ns()
<< "; no metadata found" );
}
set<Shard> shards;
if ( _manager ) {
try {
_manager->getShardsForQuery( shards, query );
} catch ( const DBException& ex ) {
return ex.toStatus();
}
}
else {
shards.insert( *_primary );
}
for ( set<Shard>::iterator it = shards.begin(); it != shards.end(); ++it ) {
endpoints->push_back(new ShardEndpoint(it->getName(),
_manager ?
_manager->getVersion(it->getName()) :
ChunkVersion::UNSHARDED()));
}
return Status::OK();
}
Status ChunkManagerTargeter::targetShardKey(const BSONObj& shardKey,
long long estDataSize,
ShardEndpoint** endpoint) const {
invariant(NULL != _manager);
ChunkPtr chunk = _manager->findIntersectingChunk(shardKey);
// Track autosplit stats for sharded collections
// Note: this is only best effort accounting and is not accurate.
if (estDataSize > 0)
_stats->chunkSizeDelta[chunk->getMin()] += estDataSize;
Shard shard = chunk->getShard();
*endpoint = new ShardEndpoint(shard.getName(),
_manager->getVersion(shard.getName()));
return Status::OK();
}
Status ChunkManagerTargeter::targetCollection( vector<ShardEndpoint*>* endpoints ) const {
if ( !_primary && !_manager ) {
return Status( ErrorCodes::NamespaceNotFound,
str::stream() << "could not target full range of "
<< getNS().ns()
<< "; metadata not found" );
}
set<Shard> shards;
if ( _manager ) {
_manager->getAllShards( shards );
}
else {
shards.insert( *_primary );
}
for ( set<Shard>::iterator it = shards.begin(); it != shards.end(); ++it ) {
endpoints->push_back(new ShardEndpoint(it->getName(),
_manager ?
_manager->getVersion(it->getName()) :
ChunkVersion::UNSHARDED()));
}
return Status::OK();
}
Status ChunkManagerTargeter::targetAllShards( vector<ShardEndpoint*>* endpoints ) const {
if ( !_primary && !_manager ) {
return Status( ErrorCodes::NamespaceNotFound,
str::stream() << "could not target every shard with versions for "
<< getNS().ns()
<< "; metadata not found" );
}
vector<Shard> shards;
Shard::getAllShards( shards );
for ( vector<Shard>::iterator it = shards.begin(); it != shards.end(); ++it ) {
endpoints->push_back(new ShardEndpoint(it->getName(),
_manager ?
_manager->getVersion(it->getName()) :
ChunkVersion::UNSHARDED()));
}
return Status::OK();
}
namespace {
//
// Utilities to compare shard versions
//
enum CompareResult {
CompareResult_Unknown, CompareResult_GTE, CompareResult_LT
};
/**
* Returns the relationship of two shard versions. Shard versions of a collection that has
* not been dropped and recreated and where there is at least one chunk on a shard are
* comparable, otherwise the result is ambiguous.
*/
CompareResult compareShardVersions( const ChunkVersion& shardVersionA,
const ChunkVersion& shardVersionB ) {
// Collection may have been dropped
if ( !shardVersionA.hasEqualEpoch( shardVersionB ) ) return CompareResult_Unknown;
// Zero shard versions are only comparable to themselves
if ( !shardVersionA.isSet() || !shardVersionB.isSet() ) {
// If both are zero...
if ( !shardVersionA.isSet() && !shardVersionB.isSet() ) return CompareResult_GTE;
// Otherwise...
return CompareResult_Unknown;
}
if ( shardVersionA < shardVersionB ) return CompareResult_LT;
else return CompareResult_GTE;
}
ChunkVersion getShardVersion( StringData shardName,
const ChunkManagerPtr& manager,
const ShardPtr& primary ) {
dassert( !( manager && primary ) );
dassert( manager || primary );
if ( primary ) return ChunkVersion::UNSHARDED();
return manager->getVersion(shardName.toString());
}
/**
* Returns the relationship between two maps of shard versions. As above, these maps are
* often comparable when the collection has not been dropped and there is at least one
* chunk on the shards.
*
* If any versions in the maps are not comparable, the result is _Unknown.
*
* If any versions in the first map (cached) are _LT the versions in the second map
* (remote), the first (cached) versions are _LT the second (remote) versions.
*
* Note that the signature here is weird since our cached map of chunk versions is
* stored in a ChunkManager or is implicit in the primary shard of the collection.
*/
CompareResult //
compareAllShardVersions( const ChunkManagerPtr& cachedShardVersions,
const ShardPtr& cachedPrimary,
const map<string, ChunkVersion>& remoteShardVersions ) {
CompareResult finalResult = CompareResult_GTE;
for ( map<string, ChunkVersion>::const_iterator it = remoteShardVersions.begin();
it != remoteShardVersions.end(); ++it ) {
//
// Get the remote and cached version for the next shard
//
const string& shardName = it->first;
const ChunkVersion& remoteShardVersion = it->second;
ChunkVersion cachedShardVersion;
try {
// Throws b/c shard constructor throws
cachedShardVersion = getShardVersion( shardName,
cachedShardVersions,
cachedPrimary );
}
catch ( const DBException& ex ) {
warning() << "could not lookup shard " << shardName
<< " in local cache, shard metadata may have changed"
<< " or be unavailable" << causedBy( ex ) << endl;
return CompareResult_Unknown;
}
//
// Compare the remote and cached versions
//
CompareResult result = compareShardVersions( cachedShardVersion,
remoteShardVersion );
if ( result == CompareResult_Unknown ) return result;
if ( result == CompareResult_LT ) finalResult = CompareResult_LT;
// Note that we keep going after _LT b/c there could be more _Unknowns.
}
return finalResult;
}
/**
* Whether or not the manager/primary pair is different from the other manager/primary pair
*/
bool isMetadataDifferent( const ChunkManagerPtr& managerA,
const ShardPtr& primaryA,
const ChunkManagerPtr& managerB,
const ShardPtr& primaryB ) {
if ( ( managerA && !managerB ) || ( !managerA && managerB ) || ( primaryA && !primaryB )
|| ( !primaryA && primaryB ) ) return true;
if ( managerA ) {
return !managerA->getVersion().isStrictlyEqualTo( managerB->getVersion() );
}
dassert( NULL != primaryA.get() );
return primaryA->getName() != primaryB->getName();
}
/**
* Whether or not the manager/primary pair was changed or refreshed from a previous version
* of the metadata.
*/
bool wasMetadataRefreshed( const ChunkManagerPtr& managerA,
const ShardPtr& primaryA,
const ChunkManagerPtr& managerB,
const ShardPtr& primaryB ) {
if ( isMetadataDifferent( managerA, primaryA, managerB, primaryB ) )
return true;
if ( managerA ) {
dassert( managerB.get() ); // otherwise metadata would be different
return managerA->getSequenceNumber() != managerB->getSequenceNumber();
}
return false;
}
}
void ChunkManagerTargeter::noteStaleResponse( const ShardEndpoint& endpoint,
const BSONObj& staleInfo ) {
dassert( !_needsTargetingRefresh );
ChunkVersion remoteShardVersion;
if ( staleInfo["vWanted"].eoo() ) {
// If we don't have a vWanted sent, assume the version is higher than our current
// version.
remoteShardVersion = getShardVersion( endpoint.shardName, _manager, _primary );
remoteShardVersion.incMajor();
}
else {
remoteShardVersion = ChunkVersion::fromBSON( staleInfo, "vWanted" );
}
ShardVersionMap::iterator it = _remoteShardVersions.find( endpoint.shardName );
if ( it == _remoteShardVersions.end() ) {
_remoteShardVersions.insert( make_pair( endpoint.shardName, remoteShardVersion ) );
}
else {
ChunkVersion& previouslyNotedVersion = it->second;
if ( previouslyNotedVersion.hasEqualEpoch( remoteShardVersion )) {
if ( previouslyNotedVersion.isOlderThan( remoteShardVersion )) {
remoteShardVersion.cloneTo( &previouslyNotedVersion );
}
}
else {
// Epoch changed midway while applying the batch so set the version to
// something unique and non-existent to force a reload when
// refreshIsNeeded is called.
ChunkVersion::IGNORED().cloneTo( &previouslyNotedVersion );
}
}
}
void ChunkManagerTargeter::noteCouldNotTarget() {
dassert( _remoteShardVersions.empty() );
_needsTargetingRefresh = true;
}
const TargeterStats* ChunkManagerTargeter::getStats() const {
return _stats.get();
}
Status ChunkManagerTargeter::refreshIfNeeded( bool *wasChanged ) {
bool dummy;
if ( !wasChanged )
wasChanged = &dummy;
*wasChanged = false;
//
// Did we have any stale config or targeting errors at all?
//
if ( !_needsTargetingRefresh && _remoteShardVersions.empty() ) return Status::OK();
//
// Get the latest metadata information from the cache if there were issues
//
ChunkManagerPtr lastManager = _manager;
ShardPtr lastPrimary = _primary;
DBConfigPtr config;
string errMsg;
if ( !getDBConfigSafe( _nss.db(), config, &errMsg ) ) {
return Status( ErrorCodes::DatabaseNotFound, errMsg );
}
// Get either the chunk manager or primary shard
config->getChunkManagerOrPrimary( _nss.ns(), _manager, _primary );
// We now have the latest metadata from the cache.
//
// See if and how we need to do a remote refresh.
// Either we couldn't target at all, or we have stale versions, but not both.
//
dassert( !( _needsTargetingRefresh && !_remoteShardVersions.empty() ) );
if ( _needsTargetingRefresh ) {
// Reset the field
_needsTargetingRefresh = false;
// If we couldn't target, we might need to refresh if we haven't remotely refreshed the
// metadata since we last got it from the cache.
bool alreadyRefreshed = wasMetadataRefreshed( lastManager,
lastPrimary,
_manager,
_primary );
// If didn't already refresh the targeting information, refresh it
if ( !alreadyRefreshed ) {
// To match previous behavior, we just need an incremental refresh here
return refreshNow( RefreshType_RefreshChunkManager );
}
*wasChanged = isMetadataDifferent( lastManager, lastPrimary, _manager, _primary );
return Status::OK();
}
else if ( !_remoteShardVersions.empty() ) {
// If we got stale shard versions from remote shards, we may need to refresh
// NOTE: Not sure yet if this can happen simultaneously with targeting issues
CompareResult result = compareAllShardVersions( _manager,
_primary,
_remoteShardVersions );
// Reset the versions
_remoteShardVersions.clear();
if ( result == CompareResult_Unknown ) {
// Our current shard versions aren't all comparable to the old versions, maybe drop
return refreshNow( RefreshType_ReloadDatabase );
}
else if ( result == CompareResult_LT ) {
// Our current shard versions are less than the remote versions, but no drop
return refreshNow( RefreshType_RefreshChunkManager );
}
*wasChanged = isMetadataDifferent( lastManager, lastPrimary, _manager, _primary );
return Status::OK();
}
// unreachable
dassert( false );
return Status::OK();
}
// To match legacy reload behavior, we have to backoff on config reload per-thread
// TODO: Centralize this behavior better by refactoring config reload in mongos
static const int maxWaitMillis = 500;
static boost::thread_specific_ptr<Backoff> perThreadBackoff;
static void refreshBackoff() {
if ( !perThreadBackoff.get() )
perThreadBackoff.reset( new Backoff( maxWaitMillis, maxWaitMillis * 2 ) );
perThreadBackoff.get()->nextSleepMillis();
}
Status ChunkManagerTargeter::refreshNow( RefreshType refreshType ) {
DBConfigPtr config;
string errMsg;
if ( !getDBConfigSafe( _nss.db(), config, &errMsg ) ) {
return Status( ErrorCodes::DatabaseNotFound, errMsg );
}
// Try not to spam the configs
refreshBackoff();
// TODO: Improve synchronization and make more explicit
if ( refreshType == RefreshType_RefreshChunkManager ) {
try {
// Forces a remote check of the collection info, synchronization between threads
// happens internally.
config->getChunkManagerIfExists( _nss.ns(), true );
}
catch ( const DBException& ex ) {
return Status( ErrorCodes::UnknownError, ex.toString() );
}
config->getChunkManagerOrPrimary( _nss.ns(), _manager, _primary );
}
else if ( refreshType == RefreshType_ReloadDatabase ) {
try {
// Dumps the db info, reloads it all, synchronization between threads happens
// internally.
config->reload();
config->getChunkManagerIfExists( _nss.ns(), true, true );
}
catch ( const DBException& ex ) {
return Status( ErrorCodes::UnknownError, ex.toString() );
}
config->getChunkManagerOrPrimary( _nss.ns(), _manager, _primary );
}
return Status::OK();
}
} // namespace mongo
|