forked from mongodb/mongo
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathchunk.cpp
1016 lines (806 loc) · 37.5 KB
/
chunk.cpp
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
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
// @file chunk.cpp
/**
* Copyright (C) 2008 10gen 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/>.
*/
#include "pch.h"
#include "../client/connpool.h"
#include "../db/querypattern.h"
#include "../db/queryutil.h"
#include "../util/unittest.h"
#include "../util/timer.h"
#include "chunk.h"
#include "config.h"
#include "cursors.h"
#include "grid.h"
#include "strategy.h"
#include "client.h"
namespace mongo {
inline bool allOfType(BSONType type, const BSONObj& o) {
BSONObjIterator it(o);
while(it.more()) {
if (it.next().type() != type)
return false;
}
return true;
}
// ------- Shard --------
string Chunk::chunkMetadataNS = "config.chunks";
int Chunk::MaxChunkSize = 1024 * 1024 * 64;
int Chunk::MaxObjectPerChunk = 250000;
Chunk::Chunk(const ChunkManager * manager, BSONObj from)
: _manager(manager), _lastmod(0), _dataWritten(mkDataWritten())
{
string ns = from.getStringField( "ns" );
_shard.reset( from.getStringField( "shard" ) );
_lastmod = from["lastmod"];
assert( _lastmod > 0 );
_min = from.getObjectField( "min" ).getOwned();
_max = from.getObjectField( "max" ).getOwned();
uassert( 10170 , "Chunk needs a ns" , ! ns.empty() );
uassert( 13327 , "Chunk ns must match server ns" , ns == _manager->getns() );
uassert( 10171 , "Chunk needs a server" , _shard.ok() );
uassert( 10172 , "Chunk needs a min" , ! _min.isEmpty() );
uassert( 10173 , "Chunk needs a max" , ! _max.isEmpty() );
}
Chunk::Chunk(const ChunkManager * info , const BSONObj& min, const BSONObj& max, const Shard& shard)
: _manager(info), _min(min), _max(max), _shard(shard), _lastmod(0), _dataWritten(mkDataWritten())
{}
long Chunk::mkDataWritten() {
return rand() % ( MaxChunkSize / 5 );
}
string Chunk::getns() const {
assert( _manager );
return _manager->getns();
}
bool Chunk::contains( const BSONObj& obj ) const {
return
_manager->getShardKey().compare( getMin() , obj ) <= 0 &&
_manager->getShardKey().compare( obj , getMax() ) < 0;
}
bool ChunkRange::contains(const BSONObj& obj) const {
// same as Chunk method
return
_manager->getShardKey().compare( getMin() , obj ) <= 0 &&
_manager->getShardKey().compare( obj , getMax() ) < 0;
}
bool Chunk::minIsInf() const {
return _manager->getShardKey().globalMin().woCompare( getMin() ) == 0;
}
bool Chunk::maxIsInf() const {
return _manager->getShardKey().globalMax().woCompare( getMax() ) == 0;
}
BSONObj Chunk::_getExtremeKey( int sort ) const {
ShardConnection conn( getShard().getConnString() , _manager->getns() );
Query q;
if ( sort == 1 ) {
q.sort( _manager->getShardKey().key() );
}
else {
// need to invert shard key pattern to sort backwards
// TODO: make a helper in ShardKeyPattern?
BSONObj k = _manager->getShardKey().key();
BSONObjBuilder r;
BSONObjIterator i(k);
while( i.more() ) {
BSONElement e = i.next();
uassert( 10163 , "can only handle numbers here - which i think is correct" , e.isNumber() );
r.append( e.fieldName() , -1 * e.number() );
}
q.sort( r.obj() );
}
// find the extreme key
BSONObj end = conn->findOne( _manager->getns() , q );
conn.done();
if ( end.isEmpty() )
return BSONObj();
return _manager->getShardKey().extractKey( end );
}
void Chunk::pickMedianKey( BSONObj& medianKey ) const {
// Ask the mongod holding this chunk to figure out the split points.
ScopedDbConnection conn( getShard().getConnString() );
BSONObj result;
BSONObjBuilder cmd;
cmd.append( "splitVector" , _manager->getns() );
cmd.append( "keyPattern" , _manager->getShardKey().key() );
cmd.append( "min" , getMin() );
cmd.append( "max" , getMax() );
cmd.appendBool( "force" , true );
BSONObj cmdObj = cmd.obj();
if ( ! conn->runCommand( "admin" , cmdObj , result )) {
conn.done();
ostringstream os;
os << "splitVector command (median key) failed: " << result;
uassert( 13503 , os.str() , 0 );
}
BSONObjIterator it( result.getObjectField( "splitKeys" ) );
if ( it.more() ) {
medianKey = it.next().Obj().getOwned();
}
conn.done();
}
void Chunk::pickSplitVector( vector<BSONObj>& splitPoints , int chunkSize /* bytes */, int maxPoints, int maxObjs ) const {
// Ask the mongod holding this chunk to figure out the split points.
ScopedDbConnection conn( getShard().getConnString() );
BSONObj result;
BSONObjBuilder cmd;
cmd.append( "splitVector" , _manager->getns() );
cmd.append( "keyPattern" , _manager->getShardKey().key() );
cmd.append( "min" , getMin() );
cmd.append( "max" , getMax() );
cmd.append( "maxChunkSizeBytes" , chunkSize );
cmd.append( "maxSplitPoints" , maxPoints );
cmd.append( "maxChunkObjects" , maxObjs );
BSONObj cmdObj = cmd.obj();
if ( ! conn->runCommand( "admin" , cmdObj , result )) {
conn.done();
ostringstream os;
os << "splitVector command failed: " << result;
uassert( 13345 , os.str() , 0 );
}
BSONObjIterator it( result.getObjectField( "splitKeys" ) );
while ( it.more() ) {
splitPoints.push_back( it.next().Obj().getOwned() );
}
conn.done();
}
BSONObj Chunk::singleSplit( bool force , BSONObj& res ) const {
vector<BSONObj> splitPoint;
// if splitting is not obligatory we may return early if there are not enough data
// we cap the number of objects that would fall in the first half (before the split point)
// the rationale is we'll find a split point without traversing all the data
if ( ! force ) {
vector<BSONObj> candidates;
const int maxPoints = 2;
pickSplitVector( candidates , getManager()->getCurrentDesiredChunkSize() , maxPoints , MaxObjectPerChunk );
if ( candidates.size() <= 1 ) {
// no split points means there isn't enough data to split on
// 1 split point means we have between half the chunk size to full chunk size
// so we shouldn't split
log(1) << "chunk not full enough to trigger auto-split" << endl;
return BSONObj();
}
splitPoint.push_back( candidates.front() );
}
else {
// if forcing a split, use the chunk's median key
BSONObj medianKey;
pickMedianKey( medianKey );
if ( ! medianKey.isEmpty() )
splitPoint.push_back( medianKey );
}
// We assume that if the chunk being split is the first (or last) one on the collection, this chunk is
// likely to see more insertions. Instead of splitting mid-chunk, we use the very first (or last) key
// as a split point.
if ( minIsInf() ) {
splitPoint.clear();
BSONObj key = _getExtremeKey( 1 );
if ( ! key.isEmpty() ) {
splitPoint.push_back( key );
}
}
else if ( maxIsInf() ) {
splitPoint.clear();
BSONObj key = _getExtremeKey( -1 );
if ( ! key.isEmpty() ) {
splitPoint.push_back( key );
}
}
// Normally, we'd have a sound split point here if the chunk is not empty. It's also a good place to
// sanity check.
if ( splitPoint.empty() || _min == splitPoint.front() || _max == splitPoint.front() ) {
log() << "want to split chunk, but can't find split point chunk " << toString()
<< " got: " << ( splitPoint.empty() ? "<empty>" : splitPoint.front().toString() ) << endl;
return BSONObj();
}
if (multiSplit( splitPoint , res ))
return splitPoint.front();
else
return BSONObj();
}
bool Chunk::multiSplit( const vector<BSONObj>& m , BSONObj& res ) const {
const size_t maxSplitPoints = 8192;
uassert( 10165 , "can't split as shard doesn't have a manager" , _manager );
uassert( 13332 , "need a split key to split chunk" , !m.empty() );
uassert( 13333 , "can't split a chunk in that many parts", m.size() < maxSplitPoints );
uassert( 13003 , "can't split a chunk with only one distinct value" , _min.woCompare(_max) );
ScopedDbConnection conn( getShard().getConnString() );
BSONObjBuilder cmd;
cmd.append( "splitChunk" , _manager->getns() );
cmd.append( "keyPattern" , _manager->getShardKey().key() );
cmd.append( "min" , getMin() );
cmd.append( "max" , getMax() );
cmd.append( "from" , getShard().getConnString() );
cmd.append( "splitKeys" , m );
cmd.append( "shardId" , genID() );
cmd.append( "configdb" , configServer.modelServer() );
BSONObj cmdObj = cmd.obj();
if ( ! conn->runCommand( "admin" , cmdObj , res )) {
warning() << "splitChunk failed - cmd: " << cmdObj << " result: " << res << endl;
conn.done();
// reloading won't strictly solve all problems, e.g. the collection's metadata lock can be taken
// but we issue here so that mongos may refresh without needing to be written/read against
_manager->reload();
return false;
}
conn.done();
// force reload of config
_manager->reload();
return true;
}
bool Chunk::moveAndCommit( const Shard& to , long long chunkSize /* bytes */, BSONObj& res ) const {
uassert( 10167 , "can't move shard to its current location!" , getShard() != to );
log() << "moving chunk ns: " << _manager->getns() << " moving ( " << toString() << ") " << _shard.toString() << " -> " << to.toString() << endl;
Shard from = _shard;
ScopedDbConnection fromconn( from);
bool worked = fromconn->runCommand( "admin" ,
BSON( "moveChunk" << _manager->getns() <<
"from" << from.getConnString() <<
"to" << to.getConnString() <<
"min" << _min <<
"max" << _max <<
"maxChunkSizeBytes" << chunkSize <<
"shardId" << genID() <<
"configdb" << configServer.modelServer()
) ,
res
);
fromconn.done();
log( worked ) << "moveChunk result: " << res << endl;
// if succeeded, needs to reload to pick up the new location
// if failed, mongos may be stale
// reload is excessive here as the failure could be simply because collection metadata is taken
_manager->reload();
return worked;
}
bool Chunk::splitIfShould( long dataWritten ) const {
LastError::Disabled d( lastError.get() );
try {
_dataWritten += dataWritten;
int splitThreshold = getManager()->getCurrentDesiredChunkSize();
if ( minIsInf() || maxIsInf() ) {
splitThreshold = (int) ((double)splitThreshold * .9);
}
if ( _dataWritten < splitThreshold / 5 )
return false;
// this is a bit ugly
// we need it so that mongos blocks for the writes to actually be committed
// this does mean mongos has more back pressure than mongod alone
// since it nots 100% tcp queue bound
// this was implicit before since we did a splitVector on the same socket
ShardConnection::sync();
log(1) << "about to initiate autosplit: " << *this << " dataWritten: " << _dataWritten << " splitThreshold: " << splitThreshold << endl;
_dataWritten = 0; // reset so we check often enough
BSONObj res;
BSONObj splitPoint = singleSplit( false /* does not force a split if not enough data */ , res );
if ( splitPoint.isEmpty() ) {
// singleSplit would have issued a message if we got here
_dataWritten = 0; // this means there wasn't enough data to split, so don't want to try again until considerable more data
return false;
}
log() << "autosplitted " << _manager->getns() << " shard: " << toString()
<< " on: " << splitPoint << "(splitThreshold " << splitThreshold << ")"
#ifdef _DEBUG
<< " size: " << getPhysicalSize() // slow - but can be useful when debugging
#endif
<< endl;
BSONElement shouldMigrate = res["shouldMigrate"]; // not in mongod < 1.9.1 but that is ok
if (!shouldMigrate.eoo() && grid.shouldBalance()){
BSONObj range = shouldMigrate.embeddedObject();
BSONObj min = range["min"].embeddedObject();
BSONObj max = range["max"].embeddedObject();
Shard newLocation = Shard::pick( getShard() );
if ( getShard() == newLocation ) {
// if this is the best shard, then we shouldn't do anything (Shard::pick already logged our shard).
log(1) << "recently split chunk: " << range << " already in the best shard: " << getShard() << endl;
return true; // we did split even if we didn't migrate
}
ChunkManagerPtr cm = _manager->reload(false/*just reloaded in mulitsplit*/);
ChunkPtr toMove = cm->findChunk(min);
if ( ! (toMove->getMin() == min && toMove->getMax() == max) ){
log(1) << "recently split chunk: " << range << " modified before we could migrate " << toMove << endl;
return true;
}
log() << "moving chunk (auto): " << toMove << " to: " << newLocation.toString() << endl;
BSONObj res;
massert( 10412 ,
str::stream() << "moveAndCommit failed: " << res ,
toMove->moveAndCommit( newLocation , MaxChunkSize , res ) );
// update our config
_manager->reload();
}
return true;
}
catch ( std::exception& e ) {
// if the collection lock is taken (e.g. we're migrating), it is fine for the split to fail.
warning() << "could have autosplit on collection: " << _manager->getns() << " but: " << e.what() << endl;
return false;
}
}
long Chunk::getPhysicalSize() const {
ScopedDbConnection conn( getShard().getConnString() );
BSONObj result;
uassert( 10169 , "datasize failed!" , conn->runCommand( "admin" ,
BSON( "datasize" << _manager->getns()
<< "keyPattern" << _manager->getShardKey().key()
<< "min" << getMin()
<< "max" << getMax()
<< "maxSize" << ( MaxChunkSize + 1 )
<< "estimate" << true
) , result ) );
conn.done();
return (long)result["size"].number();
}
void Chunk::appendShortVersion( const char * name , BSONObjBuilder& b ) const {
BSONObjBuilder bb( b.subobjStart( name ) );
bb.append( "min" , _min );
bb.append( "max" , _max );
bb.done();
}
bool Chunk::operator==( const Chunk& s ) const {
return
_manager->getShardKey().compare( _min , s._min ) == 0 &&
_manager->getShardKey().compare( _max , s._max ) == 0
;
}
void Chunk::serialize(BSONObjBuilder& to,ShardChunkVersion myLastMod) {
to.append( "_id" , genID( _manager->getns() , _min ) );
if ( myLastMod.isSet() ) {
to.appendTimestamp( "lastmod" , myLastMod );
}
else if ( _lastmod.isSet() ) {
assert( _lastmod > 0 && _lastmod < 1000 );
to.appendTimestamp( "lastmod" , _lastmod );
}
else {
assert(0);
}
to << "ns" << _manager->getns();
to << "min" << _min;
to << "max" << _max;
to << "shard" << _shard.getName();
}
string Chunk::genID( const string& ns , const BSONObj& o ) {
StringBuilder buf( ns.size() + o.objsize() + 16 );
buf << ns << "-";
BSONObjIterator i(o);
while ( i.more() ) {
BSONElement e = i.next();
buf << e.fieldName() << "_" << e.toString(false, true);
}
return buf.str();
}
string Chunk::toString() const {
stringstream ss;
ss << "ns:" << _manager->getns() << " at: " << _shard.toString() << " lastmod: " << _lastmod.toString() << " min: " << _min << " max: " << _max;
return ss.str();
}
ShardKeyPattern Chunk::skey() const {
return _manager->getShardKey();
}
// ------- ChunkManager --------
AtomicUInt ChunkManager::NextSequenceNumber = 1;
ChunkManager::ChunkManager( string ns , ShardKeyPattern pattern , bool unique ) :
_ns( ns ) , _key( pattern ) , _unique( unique ) , _chunkRanges(), _mutex("ChunkManager"),
_nsLock( ConnectionString( configServer.modelServer() , ConnectionString::SYNC ) , ns ),
// The shard versioning mechanism hinges on keeping track of the number of times we reloaded ChunkManager's.
// Increasing this number here will prompt checkShardVersion() to refresh the connection-level versions to
// the most up to date value.
_sequenceNumber(++NextSequenceNumber)
{
int tries = 3;
while (tries--) {
ChunkMap chunkMap;
set<Shard> shards;
ShardVersionMap shardVersions;
Timer t;
_load(chunkMap, shards, shardVersions);
{
int ms = t.millis();
log() << "ChunkManager: time to load chunks for " << ns << ": " << ms << "ms"
<< " sequenceNumber: " << _sequenceNumber
<< " version: " << _version.toString()
<< endl;
}
if (_isValid(chunkMap)) {
// These variables are const for thread-safety. Since the
// constructor can only be called from one thread, we don't have
// to worry about that here.
const_cast<ChunkMap&>(_chunkMap).swap(chunkMap);
const_cast<set<Shard>&>(_shards).swap(shards);
const_cast<ShardVersionMap&>(_shardVersions).swap(shardVersions);
const_cast<ChunkRangeManager&>(_chunkRanges).reloadAll(_chunkMap);
return;
}
if (_chunkMap.size() < 10) {
_printChunks();
}
warning() << "ChunkManager loaded an invalid config, trying again" << endl;
sleepmillis(10 * (3-tries));
}
// this will abort construction so we should never have a reference to an invalid config
msgasserted(13282, "Couldn't load a valid config for " + _ns + " after 3 attempts. Please try again.");
}
ChunkManagerPtr ChunkManager::reload(bool force) const {
return grid.getDBConfig(getns())->getChunkManager(getns(), force);
}
void ChunkManager::_load(ChunkMap& chunkMap, set<Shard>& shards, ShardVersionMap& shardVersions) {
ScopedDbConnection conn( configServer.modelServer() );
// TODO really need the sort?
auto_ptr<DBClientCursor> cursor = conn->query( Chunk::chunkMetadataNS, QUERY("ns" << _ns).sort("lastmod",-1), 0, 0, 0, 0,
(DEBUG_BUILD ? 2 : 1000000)); // batch size. Try to induce potential race conditions in debug builds
assert( cursor.get() );
while ( cursor->more() ) {
BSONObj d = cursor->next();
if ( d["isMaxMarker"].trueValue() ) {
continue;
}
ChunkPtr c( new Chunk( this, d ) );
chunkMap[c->getMax()] = c;
shards.insert(c->getShard());
// set global max
if ( c->getLastmod() > _version )
_version = c->getLastmod();
// set shard max
ShardChunkVersion& shardMax = shardVersions[c->getShard()];
if ( c->getLastmod() > shardMax )
shardMax = c->getLastmod();
}
conn.done();
}
bool ChunkManager::_isValid(const ChunkMap& chunkMap) {
#define ENSURE(x) do { if(!(x)) { log() << "ChunkManager::_isValid failed: " #x << endl; return false; } } while(0)
if (chunkMap.empty())
return true;
// Check endpoints
ENSURE(allOfType(MinKey, chunkMap.begin()->second->getMin()));
ENSURE(allOfType(MaxKey, 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 = prior(it);
if (!(it->second->getMin() == last->second->getMax())) {
PRINT(it->second->toString());
PRINT(it->second->getMin());
PRINT(last->second->getMax());
}
ENSURE(it->second->getMin() == last->second->getMax());
}
return true;
#undef ENSURE
}
void ChunkManager::_printChunks() const {
for (ChunkMap::const_iterator it=_chunkMap.begin(), end=_chunkMap.end(); it != end; ++it) {
log() << *it->second << endl;
}
}
bool ChunkManager::hasShardKey( const BSONObj& obj ) const {
return _key.hasShardKey( obj );
}
void ChunkManager::createFirstChunk( const Shard& shard ) const {
// TODO distlock?
assert( _chunkMap.size() == 0 );
Chunk c (this, _key.globalMin(), _key.globalMax(), shard);
// this is the first chunk; start the versioning from scratch
ShardChunkVersion version;
version.incMajor();
// build update for the chunk collection
BSONObjBuilder chunkBuilder;
c.serialize( chunkBuilder , version );
BSONObj chunkCmd = chunkBuilder.obj();
log() << "about to create first chunk for: " << _ns << endl;
ScopedDbConnection conn( configServer.modelServer() );
BSONObj res;
conn->update( Chunk::chunkMetadataNS, QUERY( "_id" << c.genID() ), chunkCmd, true, false );
string errmsg = conn->getLastError();
if ( errmsg.size() ) {
stringstream ss;
ss << "saving first chunk failed. cmd: " << chunkCmd << " result: " << errmsg;
log( LL_ERROR ) << ss.str() << endl;
msgasserted( 13592 , ss.str() );
}
conn.done();
// the ensure index will have the (desired) indirect effect of creating the collection on the
// assigned shard, as it sets up the index over the sharding keys.
ScopedDbConnection shardConn( c.getShard().getConnString() );
shardConn->ensureIndex( getns() , getShardKey().key() , _unique , "" , false /* do not cache ensureIndex SERVER-1691 */ );
shardConn.done();
log() << "successfully created first chunk for " << c.toString() << endl;
}
ChunkPtr ChunkManager::findChunk( const BSONObj & obj ) const {
BSONObj key = _key.extractKey(obj);
{
BSONObj foo;
ChunkPtr c;
{
ChunkMap::const_iterator it = _chunkMap.upper_bound(key);
if (it != _chunkMap.end()) {
foo = it->first;
c = it->second;
}
}
if ( c ) {
if ( c->contains( obj ) )
return c;
PRINT(foo);
PRINT(*c);
PRINT(key);
reload();
massert(13141, "Chunk map pointed to incorrect chunk", false);
}
}
throw UserException( 8070 , str::stream() << "couldn't find a chunk which should be impossible: " << key );
}
ChunkPtr ChunkManager::findChunkOnServer( const Shard& shard ) const {
for ( ChunkMap::const_iterator i=_chunkMap.begin(); i!=_chunkMap.end(); ++i ) {
ChunkPtr c = i->second;
if ( c->getShard() == shard )
return c;
}
return ChunkPtr();
}
void ChunkManager::getShardsForQuery( set<Shard>& shards , const BSONObj& query ) const {
//TODO look into FieldRangeSetOr
OrRangeGenerator org(_ns.c_str(), query, false);
const string special = org.getSpecial();
if (special == "2d") {
BSONForEach(field, query) {
if (getGtLtOp(field) == BSONObj::opNEAR) {
uassert(13501, "use geoNear command rather than $near query", false);
// TODO: convert to geoNear rather than erroring out
}
// $within queries are fine
}
}
else if (!special.empty()) {
uassert(13502, "unrecognized special query type: " + special, false);
}
do {
boost::scoped_ptr<FieldRangeSetPair> frsp (org.topFrsp());
{
// special case if most-significant field isn't in query
FieldRange range = frsp->singleKeyRange(_key.key().firstElementFieldName());
if ( !range.nontrivial() ) {
DEV PRINT(range.nontrivial());
getAllShards(shards);
return;
}
}
BoundList ranges = frsp->singleKeyIndexBounds(_key.key(), 1);
for (BoundList::const_iterator it=ranges.begin(), end=ranges.end(); it != end; ++it) {
BSONObj minObj = it->first.replaceFieldNames(_key.key());
BSONObj maxObj = it->second.replaceFieldNames(_key.key());
ChunkRangeMap::const_iterator min, max;
min = _chunkRanges.upper_bound(minObj);
max = _chunkRanges.upper_bound(maxObj);
massert( 13507 , str::stream() << "invalid chunk config minObj: " << minObj , min != _chunkRanges.ranges().end());
// make max non-inclusive like end iterators
if(max != _chunkRanges.ranges().end())
++max;
for (ChunkRangeMap::const_iterator it=min; it != max; ++it) {
shards.insert(it->second->getShard());
}
// once we know we need to visit all shards no need to keep looping
//if (shards.size() == _shards.size())
//return;
}
if (org.moreOrClauses())
org.popOrClauseSingleKey();
}
while (org.moreOrClauses());
}
void ChunkManager::getShardsForRange(set<Shard>& shards, const BSONObj& min, const BSONObj& max) const {
uassert(13405, "min must have shard key", hasShardKey(min));
uassert(13406, "max must have shard key", hasShardKey(max));
ChunkRangeMap::const_iterator it = _chunkRanges.upper_bound(min);
ChunkRangeMap::const_iterator end = _chunkRanges.lower_bound(max);
for (; it!=end; ++ it) {
shards.insert(it->second->getShard());
// once we know we need to visit all shards no need to keep looping
if (shards.size() == _shards.size())
break;
}
}
void ChunkManager::getAllShards( set<Shard>& all ) const {
all.insert(_shards.begin(), _shards.end());
}
void ChunkManager::drop( ChunkManagerPtr me ) const {
scoped_lock lk( _mutex );
configServer.logChange( "dropCollection.start" , _ns , BSONObj() );
dist_lock_try dlk;
try{
dlk = dist_lock_try( &_nsLock , "drop" );
}
catch( LockException& e ){
uassert( 14022, str::stream() << "Error locking distributed lock for chunk drop." << causedBy( e ), false);
}
uassert( 13331 , "collection's metadata is undergoing changes. Please try again." , dlk.got() );
uassert( 10174 , "config servers not all up" , configServer.allUp() );
set<Shard> seen;
log(1) << "ChunkManager::drop : " << _ns << endl;
// lock all shards so no one can do a split/migrate
for ( ChunkMap::const_iterator i=_chunkMap.begin(); i!=_chunkMap.end(); ++i ) {
ChunkPtr c = i->second;
seen.insert( c->getShard() );
}
log(1) << "ChunkManager::drop : " << _ns << "\t all locked" << endl;
// delete data from mongod
for ( set<Shard>::iterator i=seen.begin(); i!=seen.end(); i++ ) {
ScopedDbConnection conn( *i );
conn->dropCollection( _ns );
conn.done();
}
log(1) << "ChunkManager::drop : " << _ns << "\t removed shard data" << endl;
// remove chunk data
ScopedDbConnection conn( configServer.modelServer() );
conn->remove( Chunk::chunkMetadataNS , BSON( "ns" << _ns ) );
conn.done();
log(1) << "ChunkManager::drop : " << _ns << "\t removed chunk data" << endl;
for ( set<Shard>::iterator i=seen.begin(); i!=seen.end(); i++ ) {
ScopedDbConnection conn( *i );
BSONObj res;
// this is horrible
// we need a special command for dropping on the d side
// this hack works for the moment
if ( ! setShardVersion( conn.conn() , _ns , 0 , true , res ) )
throw UserException( 8071 , str::stream() << "cleaning up after drop failed: " << res );
conn->simpleCommand( "admin", 0, "unsetSharding" );
conn.done();
}
log(1) << "ChunkManager::drop : " << _ns << "\t DONE" << endl;
configServer.logChange( "dropCollection" , _ns , BSONObj() );
}
void ChunkManager::maybeChunkCollection() const {
uassert( 13346 , "can't pre-split already splitted collection" , (_chunkMap.size() == 1) );
ChunkPtr soleChunk = _chunkMap.begin()->second;
vector<BSONObj> splitPoints;
soleChunk->pickSplitVector( splitPoints , Chunk::MaxChunkSize );
if ( splitPoints.empty() ) {
log(1) << "not enough data to warrant chunking " << getns() << endl;
return;
}
BSONObj res;
ChunkPtr p;
bool worked = soleChunk->multiSplit( splitPoints , res );
if (!worked) {
log( LL_WARNING ) << "could not split '" << getns() << "': " << res << endl;
return;
}
}
ShardChunkVersion ChunkManager::getVersion( const Shard& shard ) const {
ShardVersionMap::const_iterator i = _shardVersions.find( shard );
if ( i == _shardVersions.end() )
return 0;
return i->second;
}
ShardChunkVersion ChunkManager::getVersion() const {
return _version;
}
string ChunkManager::toString() const {
stringstream ss;
ss << "ChunkManager: " << _ns << " key:" << _key.toString() << '\n';
for ( ChunkMap::const_iterator i=_chunkMap.begin(); i!=_chunkMap.end(); ++i ) {
const ChunkPtr c = i->second;
ss << "\t" << c->toString() << '\n';
}
return ss.str();
}
void ChunkRangeManager::assertValid() const {
if (_ranges.empty())
return;
try {
// No Nulls
for (ChunkRangeMap::const_iterator it=_ranges.begin(), end=_ranges.end(); it != end; ++it) {
assert(it->second);
}
// Check endpoints
assert(allOfType(MinKey, _ranges.begin()->second->getMin()));
assert(allOfType(MaxKey, prior(_ranges.end())->second->getMax()));
// Make sure there are no gaps or overlaps
for (ChunkRangeMap::const_iterator it=boost::next(_ranges.begin()), end=_ranges.end(); it != end; ++it) {
ChunkRangeMap::const_iterator last = prior(it);
assert(it->second->getMin() == last->second->getMax());
}
// Check Map keys
for (ChunkRangeMap::const_iterator it=_ranges.begin(), end=_ranges.end(); it != end; ++it) {
assert(it->first == it->second->getMax());
}
// Make sure we match the original chunks
const ChunkMap chunks = _ranges.begin()->second->getManager()->_chunkMap;
for ( ChunkMap::const_iterator i=chunks.begin(); i!=chunks.end(); ++i ) {
const ChunkPtr chunk = i->second;
ChunkRangeMap::const_iterator min = _ranges.upper_bound(chunk->getMin());
ChunkRangeMap::const_iterator max = _ranges.lower_bound(chunk->getMax());
assert(min != _ranges.end());
assert(max != _ranges.end());
assert(min == max);
assert(min->second->getShard() == chunk->getShard());
assert(min->second->contains( chunk->getMin() ));
assert(min->second->contains( chunk->getMax() ) || (min->second->getMax() == chunk->getMax()));
}
}
catch (...) {
log( LL_ERROR ) << "\t invalid ChunkRangeMap! printing ranges:" << endl;
for (ChunkRangeMap::const_iterator it=_ranges.begin(), end=_ranges.end(); it != end; ++it)
cout << it->first << ": " << *it->second << endl;
throw;
}
}
void ChunkRangeManager::reloadAll(const ChunkMap& chunks) {
_ranges.clear();
_insertRange(chunks.begin(), chunks.end());
DEV assertValid();
}
void ChunkRangeManager::_insertRange(ChunkMap::const_iterator begin, const ChunkMap::const_iterator end) {
while (begin != end) {
ChunkMap::const_iterator first = begin;
Shard shard = first->second->getShard();
while (begin != end && (begin->second->getShard() == shard))
++begin;
shared_ptr<ChunkRange> cr (new ChunkRange(first, begin));
_ranges[cr->getMax()] = cr;
}
}
int ChunkManager::getCurrentDesiredChunkSize() const {
// split faster in early chunks helps spread out an initial load better
const int minChunkSize = 1 << 20; // 1 MBytes
int splitThreshold = Chunk::MaxChunkSize;
int nc = numChunks();
if ( nc < 10 ) {
splitThreshold = max( splitThreshold / 4 , minChunkSize );
}
else if ( nc < 20 ) {
splitThreshold = max( splitThreshold / 2 , minChunkSize );
}
return splitThreshold;
}
class ChunkObjUnitTest : public UnitTest {
public:
void runShardChunkVersion() {
vector<ShardChunkVersion> all;
all.push_back( ShardChunkVersion(1,1) );
all.push_back( ShardChunkVersion(1,2) );
all.push_back( ShardChunkVersion(2,1) );
all.push_back( ShardChunkVersion(2,2) );
for ( unsigned i=0; i<all.size(); i++ ) {
for ( unsigned j=i+1; j<all.size(); j++ ) {
assert( all[i] < all[j] );
}
}
}
void run() {
runShardChunkVersion();
log(1) << "shardObjTest passed" << endl;
}
} shardObjTest;
// ----- to be removed ---
extern OID serverID;
// NOTE (careful when deprecating)
// currently the sharding is enabled because of a write or read (as opposed to a split or migrate), the shard learns
// its name and through the 'setShardVersion' command call
bool setShardVersion( DBClientBase & conn , const string& ns , ShardChunkVersion version , bool authoritative , BSONObj& result ) {
BSONObjBuilder cmdBuilder;
cmdBuilder.append( "setShardVersion" , ns.c_str() );
cmdBuilder.append( "configdb" , configServer.modelServer() );