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
|
// @file oplog.cpp
/**
* Copyright (C) 2008-2014 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::kReplication
#include "mongo/platform/basic.h"
#include "mongo/db/repl/oplog.h"
#include <deque>
#include <set>
#include <vector>
#include "mongo/db/auth/action_set.h"
#include "mongo/db/auth/action_type.h"
#include "mongo/db/auth/authorization_manager.h"
#include "mongo/db/auth/authorization_manager_global.h"
#include "mongo/db/auth/privilege.h"
#include "mongo/db/background.h"
#include "mongo/db/catalog/apply_ops.h"
#include "mongo/db/catalog/capped_utils.h"
#include "mongo/db/catalog/create_collection.h"
#include "mongo/db/catalog/coll_mod.h"
#include "mongo/db/catalog/collection.h"
#include "mongo/db/catalog/collection_catalog_entry.h"
#include "mongo/db/catalog/drop_collection.h"
#include "mongo/db/catalog/drop_database.h"
#include "mongo/db/catalog/drop_indexes.h"
#include "mongo/db/catalog/rename_collection.h"
#include "mongo/db/commands.h"
#include "mongo/db/commands/dbhash.h"
#include "mongo/db/concurrency/write_conflict_exception.h"
#include "mongo/db/db_raii.h"
#include "mongo/db/dbdirectclient.h"
#include "mongo/db/dbhelpers.h"
#include "mongo/db/service_context.h"
#include "mongo/db/global_timestamp.h"
#include "mongo/db/index_builder.h"
#include "mongo/db/namespace_string.h"
#include "mongo/db/op_observer.h"
#include "mongo/db/operation_context_impl.h"
#include "mongo/db/ops/delete.h"
#include "mongo/db/ops/update.h"
#include "mongo/db/ops/update_lifecycle_impl.h"
#include "mongo/db/repl/bgsync.h"
#include "mongo/db/repl/optime.h"
#include "mongo/db/repl/repl_client_info.h"
#include "mongo/db/repl/replication_coordinator_global.h"
#include "mongo/db/stats/counters.h"
#include "mongo/db/storage/storage_engine.h"
#include "mongo/db/storage_options.h"
#include "mongo/s/d_state.h"
#include "mongo/scripting/engine.h"
#include "mongo/util/elapsed_tracker.h"
#include "mongo/util/file.h"
#include "mongo/util/log.h"
#include "mongo/util/mongoutils/str.h"
#include "mongo/util/stacktrace.h"
#include "mongo/util/startup_test.h"
namespace mongo {
using std::endl;
using std::string;
using std::stringstream;
namespace repl {
std::string rsOplogName = "local.oplog.rs";
std::string masterSlaveOplogName = "local.oplog.$main";
int OPLOG_VERSION = 2;
namespace {
// cached copies of these...so don't rename them, drop them, etc.!!!
Database* _localDB = nullptr;
Collection* _localOplogCollection = nullptr;
// Synchronizes the section where a new Timestamp is generated and when it actually
// appears in the oplog.
mongo::mutex newOpMutex;
stdx::condition_variable newTimestampNotifier;
static std::string _oplogCollectionName;
// so we can fail the same way
void checkOplogInsert( StatusWith<RecordId> result ) {
massert( 17322,
str::stream() << "write to oplog failed: " << result.getStatus().toString(),
result.isOK() );
}
/**
* Allocates an optime for a new entry in the oplog, and updates the replication coordinator to
* reflect that new optime. Returns the new optime and the correct value of the "h" field for
* the new oplog entry.
*
* NOTE: From the time this function returns to the time that the new oplog entry is written
* to the storage system, all errors must be considered fatal. This is because the this
* function registers the new optime with the storage system and the replication coordinator,
* and provides no facility to revert those registrations on rollback.
*/
std::pair<OpTime, long long> getNextOpTime(OperationContext* txn,
Collection* oplog,
const char* ns,
ReplicationCoordinator* replCoord,
const char* opstr) {
stdx::lock_guard<stdx::mutex> lk(newOpMutex);
Timestamp ts = getNextGlobalTimestamp();
newTimestampNotifier.notify_all();
fassert(28560, oplog->getRecordStore()->oplogDiskLocRegister(txn, ts));
long long hashNew = 0;
long long term = 0;
// Set hash and term if we're in replset mode, otherwise they remain 0 in master/slave.
if (replCoord->getReplicationMode() == ReplicationCoordinator::modeReplSet) {
// Current term. If we're not a replset of pv=1, it could be the default value (0) or
// the last valid term before downgrade.
term = ReplClientInfo::forClient(txn->getClient()).getTerm();
hashNew = BackgroundSync::get()->getLastAppliedHash();
// Check to make sure logOp() is legal at this point.
if (*opstr == 'n') {
// 'n' operations are always logged
invariant(*ns == '\0');
// 'n' operations do not advance the hash, since they are not rolled back
}
else {
// Advance the hash
hashNew = (hashNew * 131 + ts.asLL()) * 17 + replCoord->getMyId();
BackgroundSync::get()->setLastAppliedHash(hashNew);
}
}
OpTime opTime(ts, term);
replCoord->setMyLastOptime(opTime);
return std::pair<OpTime,long long>(opTime, hashNew);
}
/**
* This allows us to stream the oplog entry directly into data region
* main goal is to avoid copying the o portion
* which can be very large
* TODO: can have this build the entire doc
*/
class OplogDocWriter : public DocWriter {
public:
OplogDocWriter( const BSONObj& frame, const BSONObj& oField )
: _frame( frame ), _oField( oField ) {
}
~OplogDocWriter(){}
void writeDocument( char* start ) const {
char* buf = start;
memcpy( buf, _frame.objdata(), _frame.objsize() - 1 ); // don't copy final EOO
reinterpret_cast<int*>( buf )[0] = documentSize();
buf += ( _frame.objsize() - 1 );
buf[0] = (char)Object;
buf[1] = 'o';
buf[2] = 0;
memcpy( buf+3, _oField.objdata(), _oField.objsize() );
buf += 3 + _oField.objsize();
buf[0] = EOO;
verify( static_cast<size_t>( ( buf + 1 ) - start ) == documentSize() ); // DEV?
}
size_t documentSize() const {
return _frame.objsize() + _oField.objsize() + 1 /* type */ + 2 /* "o" */;
}
private:
BSONObj _frame;
BSONObj _oField;
};
} // namespace
void setOplogCollectionName() {
if (getGlobalReplicationCoordinator()->getReplicationMode() ==
ReplicationCoordinator::modeReplSet) {
_oplogCollectionName = rsOplogName;
}
else {
_oplogCollectionName = masterSlaveOplogName;
}
}
/* we write to local.oplog.rs:
{ ts : ..., h: ..., v: ..., op: ..., etc }
ts: an OpTime timestamp
h: hash
v: version
op:
"i" insert
"u" update
"d" delete
"c" db cmd
"db" declares presence of a database (ns is set to the db name + '.')
"n" no op
bb param:
if not null, specifies a boolean to pass along to the other side as b: param.
used for "justOne" or "upsert" flags on 'd', 'u'
*/
void _logOp(OperationContext* txn,
const char *opstr,
const char *ns,
const BSONObj& obj,
BSONObj *o2,
bool fromMigrate) {
NamespaceString nss(ns);
if (nss.db() == "local") {
return;
}
if (nss.isSystemDotProfile()) {
return;
}
if (!getGlobalReplicationCoordinator()->isReplEnabled()) {
return;
}
if (!txn->writesAreReplicated()) {
return;
}
fassert(28626, txn->recoveryUnit());
Lock::DBLock lk(txn->lockState(), "local", MODE_IX);
ReplicationCoordinator* replCoord = getGlobalReplicationCoordinator();
if (ns[0] && replCoord->getReplicationMode() == ReplicationCoordinator::modeReplSet &&
!replCoord->canAcceptWritesFor(nss)) {
severe() << "logOp() but can't accept write to collection " << ns;
fassertFailed(17405);
}
Lock::CollectionLock lk2(txn->lockState(), _oplogCollectionName, MODE_IX);
if (_localOplogCollection == nullptr) {
OldClientContext ctx(txn, _oplogCollectionName);
_localDB = ctx.db();
invariant(_localDB);
_localOplogCollection = _localDB->getCollection(_oplogCollectionName);
massert(13347,
"the oplog collection " + _oplogCollectionName +
" missing. did you drop it? if so, restart the server",
_localOplogCollection);
}
std::pair<OpTime, long long> slot = getNextOpTime(txn,
_localOplogCollection,
ns,
replCoord,
opstr);
/* we jump through a bunch of hoops here to avoid copying the obj buffer twice --
instead we do a single copy to the destination position in the memory mapped file.
*/
BSONObjBuilder b(256);
b.append("ts", slot.first.getTimestamp());
b.append("t", slot.first.getTerm());
b.append("h", slot.second);
b.append("v", OPLOG_VERSION);
b.append("op", opstr);
b.append("ns", ns);
if (fromMigrate) {
b.appendBool("fromMigrate", true);
}
if ( o2 ) {
b.append("o2", *o2);
}
BSONObj partial = b.done();
OplogDocWriter writer( partial, obj );
checkOplogInsert( _localOplogCollection->insertDocument( txn, &writer, false ) );
ReplClientInfo::forClient(txn->getClient()).setLastOp( slot.first );
}
OpTime writeOpsToOplog(OperationContext* txn, const std::deque<BSONObj>& ops) {
ReplicationCoordinator* replCoord = getGlobalReplicationCoordinator();
OpTime lastOptime;
MONGO_WRITE_CONFLICT_RETRY_LOOP_BEGIN {
lastOptime = replCoord->getMyLastOptime();
invariant(!ops.empty());
ScopedTransaction transaction(txn, MODE_IX);
Lock::DBLock lk(txn->lockState(), "local", MODE_X);
if ( _localOplogCollection == 0 ) {
OldClientContext ctx(txn, rsOplogName);
_localDB = ctx.db();
verify( _localDB );
_localOplogCollection = _localDB->getCollection(rsOplogName);
massert(13389,
"local.oplog.rs missing. did you drop it? if so restart server",
_localOplogCollection);
}
OldClientContext ctx(txn, rsOplogName, _localDB);
WriteUnitOfWork wunit(txn);
for (std::deque<BSONObj>::const_iterator it = ops.begin();
it != ops.end();
++it) {
const BSONObj& op = *it;
const OpTime optime = extractOpTime(op);
checkOplogInsert(_localOplogCollection->insertDocument(txn, op, false));
if (!(lastOptime < optime)) {
severe() << "replication oplog stream went back in time. "
"previous timestamp: " << lastOptime << " newest timestamp: " << optime
<< ". Op being applied: " << op;
fassertFailedNoTrace(18905);
}
lastOptime = optime;
}
wunit.commit();
} MONGO_WRITE_CONFLICT_RETRY_LOOP_END(txn, "writeOps", _localOplogCollection->ns().ns());
BackgroundSync* bgsync = BackgroundSync::get();
// Keep this up-to-date, in case we step up to primary.
long long hash = ops.back()["h"].numberLong();
bgsync->setLastAppliedHash(hash);
return lastOptime;
}
void createOplog(OperationContext* txn) {
ScopedTransaction transaction(txn, MODE_X);
Lock::GlobalWrite lk(txn->lockState());
const ReplSettings& replSettings = getGlobalReplicationCoordinator()->getSettings();
bool rs = !replSettings.replSet.empty();
OldClientContext ctx(txn, _oplogCollectionName);
Collection* collection = ctx.db()->getCollection( _oplogCollectionName );
if ( collection ) {
if (replSettings.oplogSize != 0) {
const CollectionOptions oplogOpts =
collection->getCatalogEntry()->getCollectionOptions(txn);
int o = (int)(oplogOpts.cappedSize / ( 1024 * 1024 ) );
int n = (int)(replSettings.oplogSize / (1024 * 1024));
if ( n != o ) {
stringstream ss;
ss << "cmdline oplogsize (" << n << ") different than existing (" << o << ") see: http://dochub.mongodb.org/core/increase-oplog";
log() << ss.str() << endl;
throw UserException( 13257 , ss.str() );
}
}
if ( !rs )
initTimestampFromOplog(txn, _oplogCollectionName);
return;
}
/* create an oplog collection, if it doesn't yet exist. */
long long sz = 0;
if ( replSettings.oplogSize != 0 ) {
sz = replSettings.oplogSize;
}
else {
/* not specified. pick a default size */
sz = 50LL * 1024LL * 1024LL;
if ( sizeof(int *) >= 8 ) {
#if defined(__APPLE__)
// typically these are desktops (dev machines), so keep it smallish
sz = (256-64) * 1024 * 1024;
#else
sz = 990LL * 1024 * 1024;
double free =
File::freeSpace(storageGlobalParams.dbpath); //-1 if call not supported.
long long fivePct = static_cast<long long>( free * 0.05 );
if ( fivePct > sz )
sz = fivePct;
// we use 5% of free space up to 50GB (1TB free)
static long long upperBound = 50LL * 1024 * 1024 * 1024;
if (fivePct > upperBound)
sz = upperBound;
#endif
}
}
log() << "******" << endl;
log() << "creating replication oplog of size: " << (int)( sz / ( 1024 * 1024 ) ) << "MB..." << endl;
CollectionOptions options;
options.capped = true;
options.cappedSize = sz;
options.autoIndexId = CollectionOptions::NO;
MONGO_WRITE_CONFLICT_RETRY_LOOP_BEGIN {
WriteUnitOfWork uow( txn );
invariant(ctx.db()->createCollection(txn, _oplogCollectionName, options));
if( !rs )
getGlobalServiceContext()->getOpObserver()->onOpMessage(txn, BSONObj());
uow.commit();
} MONGO_WRITE_CONFLICT_RETRY_LOOP_END(txn, "createCollection", _oplogCollectionName);
/* sync here so we don't get any surprising lag later when we try to sync */
StorageEngine* storageEngine = getGlobalServiceContext()->getGlobalStorageEngine();
storageEngine->flushAllFiles(true);
log() << "******" << endl;
}
// -------------------------------------
namespace {
NamespaceString parseNs(const string& ns, const BSONObj& cmdObj) {
BSONElement first = cmdObj.firstElement();
uassert(28635,
"no collection name specified",
first.canonicalType() == canonicalizeBSONType(mongo::String)
&& first.valuestrsize() > 0);
std::string coll = first.valuestr();
return NamespaceString(NamespaceString(ns).db().toString(), coll);
}
using OpApplyFn = stdx::function<Status (OperationContext*, const char*, BSONObj&)>;
struct ApplyOpMetadata {
OpApplyFn applyFunc;
std::set<ErrorCodes::Error> acceptableErrors;
ApplyOpMetadata(OpApplyFn fun) {
applyFunc = fun;
}
ApplyOpMetadata(OpApplyFn fun, std::set<ErrorCodes::Error> theAcceptableErrors) {
applyFunc = fun;
acceptableErrors = theAcceptableErrors;
}
};
std::map<std::string, ApplyOpMetadata> opsMap = {
{"create",
{
[](OperationContext* txn, const char* ns, BSONObj& cmd) -> Status {
return createCollection(txn, NamespaceString(ns).db().toString(), cmd);
},
{ErrorCodes::NamespaceExists}
}
},
{"collMod",
{
[](OperationContext* txn, const char* ns, BSONObj& cmd) -> Status {
BSONObjBuilder resultWeDontCareAbout;
return collMod(txn, parseNs(ns, cmd), cmd, &resultWeDontCareAbout);
}
}
},
{"dropDatabase",
{
[](OperationContext* txn, const char* ns, BSONObj& cmd) -> Status {
return dropDatabase(txn, NamespaceString(ns).db().toString());
},
{ErrorCodes::DatabaseNotFound}
}
},
{"drop",
{
[](OperationContext* txn, const char* ns, BSONObj& cmd) -> Status {
BSONObjBuilder resultWeDontCareAbout;
return dropCollection(txn, parseNs(ns, cmd), resultWeDontCareAbout);
},
// IllegalOperation is necessary because in 3.0 we replicate drops of system.profile
// TODO(dannenberg) remove IllegalOperation once we no longer need 3.0 compatibility
{ErrorCodes::NamespaceNotFound, ErrorCodes::IllegalOperation}
}
},
// deleteIndex(es) is deprecated but still works as of April 10, 2015
{"deleteIndex",
{
[](OperationContext* txn, const char* ns, BSONObj& cmd) -> Status {
BSONObjBuilder resultWeDontCareAbout;
return dropIndexes(txn, parseNs(ns, cmd), cmd, &resultWeDontCareAbout);
},
{ErrorCodes::NamespaceNotFound, ErrorCodes::IndexNotFound}
}
},
{"deleteIndexes",
{
[](OperationContext* txn, const char* ns, BSONObj& cmd) -> Status {
BSONObjBuilder resultWeDontCareAbout;
return dropIndexes(txn, parseNs(ns, cmd), cmd, &resultWeDontCareAbout);
},
{ErrorCodes::NamespaceNotFound, ErrorCodes::IndexNotFound}
}
},
{"dropIndex",
{
[](OperationContext* txn, const char* ns, BSONObj& cmd) -> Status {
BSONObjBuilder resultWeDontCareAbout;
return dropIndexes(txn, parseNs(ns, cmd), cmd, &resultWeDontCareAbout);
},
{ErrorCodes::NamespaceNotFound, ErrorCodes::IndexNotFound}
}
},
{"dropIndexes",
{
[](OperationContext* txn, const char* ns, BSONObj& cmd) -> Status {
BSONObjBuilder resultWeDontCareAbout;
return dropIndexes(txn, parseNs(ns, cmd), cmd, &resultWeDontCareAbout);
},
{ErrorCodes::NamespaceNotFound, ErrorCodes::IndexNotFound}
}
},
{"renameCollection",
{
[](OperationContext* txn, const char* ns, BSONObj& cmd) -> Status {
return renameCollection(txn,
NamespaceString(cmd.firstElement().valuestrsafe()),
NamespaceString(cmd["to"].valuestrsafe()),
cmd["stayTemp"].trueValue(),
cmd["dropTarget"].trueValue());
},
{ErrorCodes::NamespaceNotFound, ErrorCodes::NamespaceExists}
}
},
{"applyOps",
{
[](OperationContext* txn, const char* ns, BSONObj& cmd) -> Status {
BSONObjBuilder resultWeDontCareAbout;
return applyOps(txn, nsToDatabase(ns), cmd, &resultWeDontCareAbout);
},
{ErrorCodes::UnknownError}
}
},
{"convertToCapped",
{
[](OperationContext* txn, const char* ns, BSONObj& cmd) -> Status {
return convertToCapped(txn,
parseNs(ns, cmd),
cmd["size"].number());
}
}
},
{"emptycapped",
{
[](OperationContext* txn, const char* ns, BSONObj& cmd) -> Status {
return emptyCapped(txn, parseNs(ns, cmd));
}
}
},
};
} // namespace
// @return failure status if an update should have happened and the document DNE.
// See replset initial sync code.
Status applyOperation_inlock(OperationContext* txn,
Database* db,
const BSONObj& op,
bool convertUpdateToUpsert) {
LOG(3) << "applying op: " << op << endl;
OpCounters * opCounters = txn->writesAreReplicated() ? &globalOpCounters : &replOpCounters;
const char *names[] = { "o", "ns", "op", "b", "o2" };
BSONElement fields[5];
op.getFields(5, names, fields);
BSONElement& fieldO = fields[0];
BSONElement& fieldNs = fields[1];
BSONElement& fieldOp = fields[2];
BSONElement& fieldB = fields[3];
BSONElement& fieldO2 = fields[4];
BSONObj o;
if( fieldO.isABSONObj() )
o = fieldO.embeddedObject();
const char *ns = fieldNs.valuestrsafe();
BSONObj o2;
if (fieldO2.isABSONObj())
o2 = fieldO2.Obj();
bool valueB = fieldB.booleanSafe();
if (nsIsFull(ns)) {
if (supportsDocLocking()) {
// WiredTiger, and others requires MODE_IX since the applier threads driving
// this allow writes to the same collection on any thread.
invariant(txn->lockState()->isCollectionLockedForMode(ns, MODE_IX));
} else {
// mmapV1 ensures that all operations to the same collection are executed from
// the same worker thread, so it takes an exclusive lock (MODE_X)
invariant(txn->lockState()->isCollectionLockedForMode(ns, MODE_X));
}
}
Collection* collection = db->getCollection( ns );
IndexCatalog* indexCatalog = collection == nullptr ? nullptr : collection->getIndexCatalog();
// operation type -- see logOp() comments for types
const char *opType = fieldOp.valuestrsafe();
invariant(*opType != 'c'); // commands are processed in applyCommand_inlock()
if ( *opType == 'i' ) {
opCounters->gotInsert();
const char *p = strchr(ns, '.');
if ( p && nsToCollectionSubstring( p ) == "system.indexes" ) {
if (o["background"].trueValue()) {
IndexBuilder* builder = new IndexBuilder(o);
// This spawns a new thread and returns immediately.
builder->go();
// Wait for thread to start and register itself
Lock::TempRelease release(txn->lockState());
IndexBuilder::waitForBgIndexStarting();
}
else {
IndexBuilder builder(o);
Status status = builder.buildInForeground(txn, db);
uassertStatusOK(status);
}
}
else {
// do upserts for inserts as we might get replayed more than once
OpDebug debug;
uassert(ErrorCodes::NamespaceNotFound, str::stream() <<
"Failed to apply insert due to missing collection: " << op.toString(),
collection);
// No _id.
// This indicates an issue with the upstream server:
// The oplog entry is corrupted; or
// The version of the upstream server is obsolete.
uassert(ErrorCodes::NoSuchKey, str::stream() <<
"Failed to apply insert due to missing _id: " << op.toString(),
o.hasField("_id"));
// TODO: It may be better to do an insert here, and then catch the duplicate
// key exception and do update then. Very few upserts will not be inserts...
BSONObjBuilder b;
b.append(o.getField("_id"));
const NamespaceString requestNs(ns);
UpdateRequest request(requestNs);
request.setQuery(b.done());
request.setUpdates(o);
request.setUpsert();
UpdateLifecycleImpl updateLifecycle(true, requestNs);
request.setLifecycle(&updateLifecycle);
update(txn, db, request, &debug);
}
}
else if ( *opType == 'u' ) {
opCounters->gotUpdate();
OpDebug debug;
BSONObj updateCriteria = o2;
const bool upsert = valueB || convertUpdateToUpsert;
uassert(ErrorCodes::NoSuchKey, str::stream() <<
"Failed to apply update due to missing _id: " << op.toString(),
updateCriteria.hasField("_id"));
const NamespaceString requestNs(ns);
UpdateRequest request(requestNs);
request.setQuery(updateCriteria);
request.setUpdates(o);
request.setUpsert(upsert);
UpdateLifecycleImpl updateLifecycle(true, requestNs);
request.setLifecycle(&updateLifecycle);
UpdateResult ur = update(txn, db, request, &debug);
if( ur.numMatched == 0 ) {
if( ur.modifiers ) {
if( updateCriteria.nFields() == 1 ) {
// was a simple { _id : ... } update criteria
string msg = str::stream() << "failed to apply update: " << op.toString();
error() << msg;
return Status(ErrorCodes::OperationFailed, msg);
}
// Need to check to see if it isn't present so we can exit early with a
// failure. Note that adds some overhead for this extra check in some cases,
// such as an updateCriteria
// of the form
// { _id:..., { x : {$size:...} }
// thus this is not ideal.
if (collection == NULL ||
(indexCatalog->haveIdIndex(txn) &&
Helpers::findById(txn, collection, updateCriteria).isNull()) ||
// capped collections won't have an _id index
(!indexCatalog->haveIdIndex(txn) &&
Helpers::findOne(txn, collection, updateCriteria, false).isNull())) {
string msg = str::stream() << "couldn't find doc: " << op.toString();
error() << msg;
return Status(ErrorCodes::OperationFailed, msg);
}
// Otherwise, it's present; zero objects were updated because of additional specifiers
// in the query for idempotence
}
else {
// this could happen benignly on an oplog duplicate replay of an upsert
// (because we are idempotent),
// if an regular non-mod update fails the item is (presumably) missing.
if( !upsert ) {
string msg = str::stream() << "update of non-mod failed: " << op.toString();
error() << msg;
return Status(ErrorCodes::OperationFailed, msg);
}
}
}
}
else if ( *opType == 'd' ) {
opCounters->gotDelete();
uassert(ErrorCodes::NoSuchKey, str::stream() <<
"Failed to apply delete due to missing _id: " << op.toString(),
o.hasField("_id"));
if (opType[1] == 0) {
deleteObjects(txn, db, ns, o, PlanExecutor::YIELD_MANUAL, /*justOne*/ valueB);
}
else
verify( opType[1] == 'b' ); // "db" advertisement
}
else if ( *opType == 'n' ) {
// no op
}
else {
throw MsgAssertionException( 14825 , ErrorMsg("error in applyOperation : unknown opType ", *opType) );
}
// AuthorizationManager's logOp method registers a RecoveryUnit::Change
// and to do so we need to have begun a UnitOfWork
WriteUnitOfWork wuow(txn);
getGlobalAuthorizationManager()->logOp(
txn,
opType,
ns,
o,
fieldO2.isABSONObj() ? &o2 : NULL);
wuow.commit();
return Status::OK();
}
Status applyCommand_inlock(OperationContext* txn, const BSONObj& op) {
const char *names[] = { "o", "ns", "op" };
BSONElement fields[3];
op.getFields(3, names, fields);
BSONElement& fieldO = fields[0];
BSONElement& fieldNs = fields[1];
BSONElement& fieldOp = fields[2];
const char* opType = fieldOp.valuestrsafe();
invariant(*opType == 'c'); // only commands are processed here
BSONObj o;
if (fieldO.isABSONObj()) {
o = fieldO.embeddedObject();
}
const char* ns = fieldNs.valuestrsafe();
// Applying commands in repl is done under Global W-lock, so it is safe to not
// perform the current DB checks after reacquiring the lock.
invariant(txn->lockState()->isW());
bool done = false;
while (!done) {
ApplyOpMetadata curOpToApply = opsMap.find(o.firstElementFieldName())->second;
Status status = Status::OK();
try {
status = curOpToApply.applyFunc(txn, ns, o);
}
catch (...) {
status = exceptionToStatus();
}
switch (status.code()) {
case ErrorCodes::WriteConflict: {
// Need to throw this up to a higher level where it will be caught and the
// operation retried.
throw WriteConflictException();
}
case ErrorCodes::BackgroundOperationInProgressForDatabase: {
Lock::TempRelease release(txn->lockState());
BackgroundOperation::awaitNoBgOpInProgForDb(nsToDatabaseSubstring(ns));
txn->recoveryUnit()->abandonSnapshot();
break;
}
case ErrorCodes::BackgroundOperationInProgressForNamespace: {
Lock::TempRelease release(txn->lockState());
Command* cmd = Command::findCommand(o.firstElement().fieldName());
invariant(cmd);
BackgroundOperation::awaitNoBgOpInProgForNs(cmd->parseNs(nsToDatabase(ns), o));
txn->recoveryUnit()->abandonSnapshot();
break;
}
default:
if (_oplogCollectionName == masterSlaveOplogName) {
error() << "Failed command " << o << " on " << nsToDatabaseSubstring(ns)
<< " with status " << status << " during oplog application";
}
else if (curOpToApply.acceptableErrors.find(status.code())
== curOpToApply.acceptableErrors.end()) {
error() << "Failed command " << o << " on " << nsToDatabaseSubstring(ns)
<< " with status " << status << " during oplog application";
return status;
}
// fallthrough
case ErrorCodes::OK:
done = true;
break;
}
}
// AuthorizationManager's logOp method registers a RecoveryUnit::Change
// and to do so we need to have begun a UnitOfWork
WriteUnitOfWork wuow(txn);
getGlobalAuthorizationManager()->logOp(txn, opType, ns, o, nullptr);
wuow.commit();
return Status::OK();
}
void waitUpToOneSecondForTimestampChange(const Timestamp& referenceTime) {
stdx::unique_lock<stdx::mutex> lk(newOpMutex);
while (referenceTime == getLastSetTimestamp()) {
if (!newTimestampNotifier.timed_wait(lk, boost::posix_time::seconds(1)))
return;
}
}
void setNewTimestamp(const Timestamp& newTime) {
stdx::lock_guard<stdx::mutex> lk(newOpMutex);
setGlobalTimestamp(newTime);
newTimestampNotifier.notify_all();
}
OpTime extractOpTime(const BSONObj& op) {
const Timestamp ts = op["ts"].timestamp();
const long long term = op["t"].numberLong(); // Default to 0 if it's absent
return OpTime(ts, term);
}
void initTimestampFromOplog(OperationContext* txn, const std::string& oplogNS) {
DBDirectClient c(txn);
BSONObj lastOp = c.findOne(oplogNS,
Query().sort(reverseNaturalObj),
NULL,
QueryOption_SlaveOk);
if (!lastOp.isEmpty()) {
LOG(1) << "replSet setting last Timestamp";
setNewTimestamp(lastOp[ "ts" ].timestamp());
}
}
void oplogCheckCloseDatabase(OperationContext* txn, Database* db) {
invariant(txn->lockState()->isW());
_localDB = nullptr;
_localOplogCollection = nullptr;
}
} // namespace repl
} // namespace mongo
|