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
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
|
/* Copyright 2009 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/>.
*
* 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::kNetwork
#include "mongo/platform/basic.h"
#include "mongo/client/dbclient_rs.h"
#include <memory>
#include <utility>
#include "mongo/bson/util/builder.h"
#include "mongo/client/connpool.h"
#include "mongo/client/dbclientcursor.h"
#include "mongo/client/global_conn_pool.h"
#include "mongo/client/read_preference.h"
#include "mongo/client/replica_set_monitor.h"
#include "mongo/client/sasl_client_authenticate.h"
#include "mongo/db/dbmessage.h"
#include "mongo/db/jsobj.h"
#include "mongo/rpc/metadata/server_selection_metadata.h"
#include "mongo/util/log.h"
namespace mongo {
using std::shared_ptr;
using std::unique_ptr;
using std::endl;
using std::map;
using std::set;
using std::string;
using std::vector;
namespace {
/*
* Set of commands that can be used with $readPreference
*/
set<string> _secOkCmdList;
class PopulateReadPrefSecOkCmdList {
public:
PopulateReadPrefSecOkCmdList() {
_secOkCmdList.insert("aggregate");
_secOkCmdList.insert("collStats");
_secOkCmdList.insert("count");
_secOkCmdList.insert("distinct");
_secOkCmdList.insert("dbStats");
_secOkCmdList.insert("explain");
_secOkCmdList.insert("find");
_secOkCmdList.insert("geoNear");
_secOkCmdList.insert("geoSearch");
_secOkCmdList.insert("group");
}
} _populateReadPrefSecOkCmdList;
/**
* Maximum number of retries to make for auto-retry logic when performing a slave ok operation.
*/
const size_t MAX_RETRY = 3;
/**
* Extracts the read preference settings from the query document. Note that this method
* assumes that the query is ok for secondaries so it defaults to
* ReadPreference::SecondaryPreferred when nothing is specified. Supports the following
* format:
*
* Format A (official format):
* { query: <actual query>, $readPreference: <read pref obj> }
*
* Format B (unofficial internal format from mongos):
* { <actual query>, $queryOptions: { $readPreference: <read pref obj> }}
*
* @param query the raw query document
*
* @return the read preference setting if a read preference exists, otherwise the default read
* preference of Primary_Only. If the tags field was not present, it will contain one
* empty tag document {} which matches any tag.
*
* @throws AssertionException if the read preference object is malformed
*/
ReadPreferenceSetting* _extractReadPref(const BSONObj& query, int queryOptions) {
if (Query::hasReadPreference(query)) {
BSONElement readPrefElement;
if (query.hasField(Query::ReadPrefField.name())) {
readPrefElement = query[Query::ReadPrefField.name()];
} else {
readPrefElement = query["$queryOptions"][Query::ReadPrefField.name()];
}
uassert(16381, "$readPreference should be an object", readPrefElement.isABSONObj());
const BSONObj& prefDoc = readPrefElement.Obj();
auto readPrefSetting = uassertStatusOK(ReadPreferenceSetting::fromBSON(prefDoc));
return new ReadPreferenceSetting(std::move(readPrefSetting));
}
// Default read pref is primary only or secondary preferred with slaveOK
ReadPreference pref = queryOptions & QueryOption_SlaveOk
? mongo::ReadPreference::SecondaryPreferred
: mongo::ReadPreference::PrimaryOnly;
return new ReadPreferenceSetting(pref, TagSet());
}
} // namespace
// --------------------------------
// ----- DBClientReplicaSet ---------
// --------------------------------
bool DBClientReplicaSet::_authPooledSecondaryConn = true;
DBClientReplicaSet::DBClientReplicaSet(const string& name,
const vector<HostAndPort>& servers,
StringData applicationName,
double so_timeout,
MongoURI uri)
: _setName(name),
_applicationName(applicationName.toString()),
_so_timeout(so_timeout),
_uri(std::move(uri)) {
if (uri.isValid()) {
_rsm = ReplicaSetMonitor::createIfNeeded(_uri);
} else {
_rsm = ReplicaSetMonitor::createIfNeeded(name,
set<HostAndPort>(servers.begin(), servers.end()));
}
}
DBClientReplicaSet::~DBClientReplicaSet() {
if (_lastSlaveOkConn.get() == _master.get()) {
_lastSlaveOkConn.release();
}
}
ReplicaSetMonitorPtr DBClientReplicaSet::_getMonitor() {
// If you can't get a ReplicaSetMonitor then this connection isn't valid
uassert(16340,
str::stream() << "No replica set monitor active and no cached seed "
"found for set: "
<< _setName,
_rsm);
return _rsm;
}
// This can't throw an exception because it is called in the destructor of ScopedDbConnection
string DBClientReplicaSet::getServerAddress() const {
if (!_rsm) {
warning() << "Trying to get server address for DBClientReplicaSet, but no "
"ReplicaSetMonitor exists for "
<< _setName;
return str::stream() << _setName << "/";
}
return _rsm->getServerAddress();
}
HostAndPort DBClientReplicaSet::getSuspectedPrimaryHostAndPort() const {
if (!_master) {
return HostAndPort();
}
return _master->getServerHostAndPort();
}
void DBClientReplicaSet::setRequestMetadataWriter(rpc::RequestMetadataWriter writer) {
// Set the hooks in both our sub-connections and in ourselves.
if (_master) {
_master->setRequestMetadataWriter(writer);
}
if (_lastSlaveOkConn.get()) {
_lastSlaveOkConn->setRequestMetadataWriter(writer);
}
DBClientWithCommands::setRequestMetadataWriter(std::move(writer));
}
void DBClientReplicaSet::setReplyMetadataReader(rpc::ReplyMetadataReader reader) {
// Set the hooks in both our sub-connections and in ourselves.
if (_master) {
_master->setReplyMetadataReader(reader);
}
if (_lastSlaveOkConn.get()) {
_lastSlaveOkConn->setReplyMetadataReader(reader);
}
DBClientWithCommands::setReplyMetadataReader(std::move(reader));
}
int DBClientReplicaSet::getMinWireVersion() {
return _getMonitor()->getMinWireVersion();
}
int DBClientReplicaSet::getMaxWireVersion() {
return _getMonitor()->getMaxWireVersion();
}
// A replica set connection is never disconnected, since it controls its own reconnection
// logic.
//
// Has the side effect of proactively clearing any cached connections which have been
// disconnected in the background.
bool DBClientReplicaSet::isStillConnected() {
if (_master && !_master->isStillConnected()) {
resetMaster();
// Don't notify monitor of bg failure, since it's not clear how long ago it happened
}
if (_lastSlaveOkConn.get() && !_lastSlaveOkConn->isStillConnected()) {
resetSlaveOkConn();
// Don't notify monitor of bg failure, since it's not clear how long ago it happened
}
return true;
}
namespace {
bool _isSecondaryCommand(StringData commandName, const BSONObj& commandArgs) {
if (_secOkCmdList.count(commandName.toString())) {
return true;
}
if (commandName == "mapReduce" || commandName == "mapreduce") {
if (!commandArgs.hasField("out")) {
return false;
}
BSONElement outElem(commandArgs["out"]);
if (outElem.isABSONObj() && outElem["inline"].ok()) {
return true;
}
}
return false;
}
// Internal implementation of isSecondaryQuery, takes previously-parsed read preference
bool _isSecondaryQuery(const string& ns,
const BSONObj& queryObj,
const ReadPreferenceSetting& readPref) {
// If the read pref is primary only, this is not a secondary query
if (readPref.pref == ReadPreference::PrimaryOnly)
return false;
if (ns.find(".$cmd") == string::npos) {
return true;
}
// This is a command with secondary-possible read pref
// Only certain commands are supported for secondary operation.
BSONObj actualQueryObj;
if (strcmp(queryObj.firstElement().fieldName(), "query") == 0) {
actualQueryObj = queryObj["query"].embeddedObject();
} else {
actualQueryObj = queryObj;
}
StringData commandName = actualQueryObj.firstElementFieldName();
return _isSecondaryCommand(commandName, actualQueryObj);
}
} // namespace
bool DBClientReplicaSet::isSecondaryQuery(const string& ns,
const BSONObj& queryObj,
int queryOptions) {
unique_ptr<ReadPreferenceSetting> readPref(_extractReadPref(queryObj, queryOptions));
return _isSecondaryQuery(ns, queryObj, *readPref);
}
DBClientConnection* DBClientReplicaSet::checkMaster() {
ReplicaSetMonitorPtr monitor = _getMonitor();
HostAndPort h = monitor->getMasterOrUassert();
if (h == _masterHost && _master) {
// a master is selected. let's just make sure connection didn't die
if (!_master->isFailed())
return _master.get();
monitor->failedHost(
_masterHost, {ErrorCodes::fromInt(40332), "Last known master host cannot be reached"});
h = monitor->getMasterOrUassert(); // old master failed, try again.
}
_masterHost = h;
MongoURI masterUri;
if (!_uri.isValid())
masterUri = MongoURI(ConnectionString(_masterHost));
else
masterUri = _uri.cloneURIForServer(_masterHost);
string errmsg;
DBClientConnection* newConn = NULL;
boost::optional<double> socketTimeout;
if (_so_timeout > 0.0)
socketTimeout = _so_timeout;
try {
// Needs to perform a dynamic_cast because we need to set the replSet
// callback. We should eventually not need this after we remove the
// callback.
newConn = dynamic_cast<DBClientConnection*>(
masterUri.connect(_applicationName, errmsg, socketTimeout));
} catch (const AssertionException& ex) {
errmsg = ex.toString();
}
if (newConn == NULL || !errmsg.empty()) {
const std::string message = str::stream() << "can't connect to new replica set master ["
<< _masterHost.toString() << "]"
<< (errmsg.empty() ? "" : ", err: ") << errmsg;
monitor->failedHost(_masterHost, {ErrorCodes::fromInt(40333), message});
uasserted(ErrorCodes::FailedToSatisfyReadPreference, message);
}
resetMaster();
_masterHost = h;
_master.reset(newConn);
_master->setParentReplSetName(_setName);
_master->setRequestMetadataWriter(getRequestMetadataWriter());
_master->setReplyMetadataReader(getReplyMetadataReader());
_authConnection(_master.get());
return _master.get();
}
bool DBClientReplicaSet::checkLastHost(const ReadPreferenceSetting* readPref) {
// Can't use a cached host if we don't have one.
if (!_lastSlaveOkConn.get() || _lastSlaveOkHost.empty()) {
return false;
}
// Don't pin if the readPrefs differ.
if (!_lastReadPref || !_lastReadPref->equals(*readPref)) {
return false;
}
// Make sure we don't think the host is down.
if (_lastSlaveOkConn->isFailed() || !_getMonitor()->isHostUp(_lastSlaveOkHost)) {
_invalidateLastSlaveOkCache(
{ErrorCodes::fromInt(40334), "Last slave connection is no longer available"});
return false;
}
return true;
}
void DBClientReplicaSet::_authConnection(DBClientConnection* conn) {
for (map<string, BSONObj>::const_iterator i = _auths.begin(); i != _auths.end(); ++i) {
try {
conn->auth(i->second);
} catch (const UserException&) {
warning() << "cached auth failed for set: " << _setName
<< " db: " << i->second[saslCommandUserDBFieldName].str()
<< " user: " << i->second[saslCommandUserFieldName].str() << endl;
}
}
}
void DBClientReplicaSet::logoutAll(DBClientConnection* conn) {
for (map<string, BSONObj>::const_iterator i = _auths.begin(); i != _auths.end(); ++i) {
BSONObj response;
try {
conn->logout(i->first, response);
} catch (const UserException& ex) {
warning() << "Failed to logout: " << conn->getServerAddress() << " on db: " << i->first
<< causedBy(redact(ex));
}
}
}
DBClientConnection& DBClientReplicaSet::masterConn() {
return *checkMaster();
}
DBClientConnection& DBClientReplicaSet::slaveConn() {
shared_ptr<ReadPreferenceSetting> readPref(
new ReadPreferenceSetting(ReadPreference::SecondaryPreferred, TagSet()));
DBClientConnection* conn = selectNodeUsingTags(readPref);
uassert(16369,
str::stream() << "No good nodes available for set: " << _getMonitor()->getName(),
conn != NULL);
return *conn;
}
bool DBClientReplicaSet::connect() {
// Returns true if there are any up hosts.
const ReadPreferenceSetting anyUpHost(ReadPreference::Nearest, TagSet());
return _getMonitor()->getHostOrRefresh(anyUpHost).isOK();
}
static bool isAuthenticationException(const DBException& ex) {
return ex.getCode() == ErrorCodes::AuthenticationFailed;
}
void DBClientReplicaSet::_auth(const BSONObj& params) {
// We prefer to authenticate against a primary, but otherwise a secondary is ok too
// Empty tag matches every secondary
shared_ptr<ReadPreferenceSetting> readPref(
new ReadPreferenceSetting(ReadPreference::PrimaryPreferred, TagSet()));
LOG(3) << "dbclient_rs authentication of " << _getMonitor()->getName() << endl;
// NOTE that we retry MAX_RETRY + 1 times, since we're always primary preferred we don't
// fallback to the primary.
Status lastNodeStatus = Status::OK();
for (size_t retry = 0; retry < MAX_RETRY + 1; retry++) {
try {
DBClientConnection* conn = selectNodeUsingTags(readPref);
if (conn == NULL) {
break;
}
conn->auth(params);
// Cache the new auth information since we now validated it's good
_auths[params[saslCommandUserDBFieldName].str()] = params.getOwned();
// Ensure the only child connection open is the one we authenticated against - other
// child connections may not have full authentication information.
// NOTE: _lastSlaveOkConn may or may not be the same as _master
dassert(_lastSlaveOkConn.get() == conn || _master.get() == conn);
if (conn != _lastSlaveOkConn.get()) {
resetSlaveOkConn();
}
if (conn != _master.get()) {
resetMaster();
}
return;
} catch (const DBException& ex) {
// We care if we can't authenticate (i.e. bad password) in credential params.
if (isAuthenticationException(ex)) {
throw;
}
const Status status = ex.toStatus();
lastNodeStatus = {status.code(),
str::stream() << "can't authenticate against replica set node "
<< _lastSlaveOkHost
<< ": "
<< status.reason()};
_invalidateLastSlaveOkCache(lastNodeStatus);
}
}
if (lastNodeStatus.isOK()) {
StringBuilder assertMsgB;
assertMsgB << "Failed to authenticate, no good nodes in " << _getMonitor()->getName();
uasserted(ErrorCodes::NodeNotFound, assertMsgB.str());
} else {
uasserted(lastNodeStatus.code(), lastNodeStatus.reason());
}
}
void DBClientReplicaSet::logout(const string& dbname, BSONObj& info) {
DBClientConnection* priConn = checkMaster();
priConn->logout(dbname, info);
_auths.erase(dbname);
/* Also logout the cached secondary connection. Note that this is only
* needed when we actually have something cached and is last known to be
* working.
*/
if (_lastSlaveOkConn.get() != NULL && !_lastSlaveOkConn->isFailed()) {
try {
BSONObj dummy;
_lastSlaveOkConn->logout(dbname, dummy);
} catch (const DBException&) {
// Make sure we can't use this connection again.
verify(_lastSlaveOkConn->isFailed());
}
}
}
// ------------- simple functions -----------------
void DBClientReplicaSet::insert(const string& ns, BSONObj obj, int flags) {
checkMaster()->insert(ns, obj, flags);
}
void DBClientReplicaSet::insert(const string& ns, const vector<BSONObj>& v, int flags) {
checkMaster()->insert(ns, v, flags);
}
void DBClientReplicaSet::remove(const string& ns, Query obj, int flags) {
checkMaster()->remove(ns, obj, flags);
}
void DBClientReplicaSet::update(const string& ns, Query query, BSONObj obj, int flags) {
return checkMaster()->update(ns, query, obj, flags);
}
unique_ptr<DBClientCursor> DBClientReplicaSet::query(const string& ns,
Query query,
int nToReturn,
int nToSkip,
const BSONObj* fieldsToReturn,
int queryOptions,
int batchSize) {
shared_ptr<ReadPreferenceSetting> readPref(_extractReadPref(query.obj, queryOptions));
if (_isSecondaryQuery(ns, query.obj, *readPref)) {
LOG(3) << "dbclient_rs query using secondary or tagged node selection in "
<< _getMonitor()->getName() << ", read pref is " << readPref->toBSON()
<< " (primary : "
<< (_master.get() != NULL ? _master->getServerAddress() : "[not cached]")
<< ", lastTagged : " << (_lastSlaveOkConn.get() != NULL
? _lastSlaveOkConn->getServerAddress()
: "[not cached]")
<< ")" << endl;
string lastNodeErrMsg;
for (size_t retry = 0; retry < MAX_RETRY; retry++) {
try {
DBClientConnection* conn = selectNodeUsingTags(readPref);
if (conn == NULL) {
break;
}
unique_ptr<DBClientCursor> cursor = conn->query(
ns, query, nToReturn, nToSkip, fieldsToReturn, queryOptions, batchSize);
return checkSlaveQueryResult(std::move(cursor));
} catch (const DBException& ex) {
const Status status = ex.toStatus();
lastNodeErrMsg = str::stream() << "can't query replica set node "
<< _lastSlaveOkHost << ": " << status.reason();
_invalidateLastSlaveOkCache({status.code(), lastNodeErrMsg});
}
}
StringBuilder assertMsg;
assertMsg << "Failed to do query, no good nodes in " << _getMonitor()->getName();
if (!lastNodeErrMsg.empty()) {
assertMsg << ", last error: " << lastNodeErrMsg;
}
uasserted(16370, assertMsg.str());
}
LOG(3) << "dbclient_rs query to primary node in " << _getMonitor()->getName() << endl;
return checkMaster()->query(
ns, query, nToReturn, nToSkip, fieldsToReturn, queryOptions, batchSize);
}
BSONObj DBClientReplicaSet::findOne(const string& ns,
const Query& query,
const BSONObj* fieldsToReturn,
int queryOptions) {
shared_ptr<ReadPreferenceSetting> readPref(_extractReadPref(query.obj, queryOptions));
if (_isSecondaryQuery(ns, query.obj, *readPref)) {
LOG(3) << "dbclient_rs findOne using secondary or tagged node selection in "
<< _getMonitor()->getName() << ", read pref is " << readPref->toBSON()
<< " (primary : "
<< (_master.get() != NULL ? _master->getServerAddress() : "[not cached]")
<< ", lastTagged : " << (_lastSlaveOkConn.get() != NULL
? _lastSlaveOkConn->getServerAddress()
: "[not cached]")
<< ")" << endl;
string lastNodeErrMsg;
for (size_t retry = 0; retry < MAX_RETRY; retry++) {
try {
DBClientConnection* conn = selectNodeUsingTags(readPref);
if (conn == NULL) {
break;
}
return conn->findOne(ns, query, fieldsToReturn, queryOptions);
} catch (const DBException& ex) {
const Status status = ex.toStatus();
lastNodeErrMsg = str::stream() << "can't findone replica set node "
<< _lastSlaveOkHost.toString() << ": "
<< status.reason();
_invalidateLastSlaveOkCache({status.code(), lastNodeErrMsg});
}
}
StringBuilder assertMsg;
assertMsg << "Failed to call findOne, no good nodes in " << _getMonitor()->getName();
if (!lastNodeErrMsg.empty()) {
assertMsg << ", last error: " << lastNodeErrMsg;
}
uasserted(16379, assertMsg.str());
}
LOG(3) << "dbclient_rs findOne to primary node in " << _getMonitor()->getName() << endl;
return checkMaster()->findOne(ns, query, fieldsToReturn, queryOptions);
}
void DBClientReplicaSet::killCursor(long long cursorID) {
// we should never call killCursor on a replica set connection
// since we don't know which server it belongs to
// can't assume master because of slave ok
// and can have a cursor survive a master change
verify(0);
}
void DBClientReplicaSet::isntMaster() {
// Can't use _getMonitor because that will create a new monitor from the cached seed if the
// monitor doesn't exist.
_rsm->failedHost(
_masterHost,
{ErrorCodes::NotMaster, str::stream() << "got not master for: " << _masterHost});
resetMaster();
}
unique_ptr<DBClientCursor> DBClientReplicaSet::checkSlaveQueryResult(
unique_ptr<DBClientCursor> result) {
if (result.get() == NULL)
return result;
BSONObj error;
bool isError = result->peekError(&error);
if (!isError)
return result;
// We only check for "not master or secondary" errors here
// If the error code here ever changes, we need to change this code also
BSONElement code = error["code"];
if (code.isNumber() && code.Int() == ErrorCodes::NotMasterOrSecondary) {
isntSecondary();
throw DBException(str::stream() << "slave " << _lastSlaveOkHost.toString()
<< " is no longer secondary",
14812);
}
return result;
}
void DBClientReplicaSet::isntSecondary() {
// Failover to next slave
_getMonitor()->failedHost(
_lastSlaveOkHost,
{ErrorCodes::NotMasterOrSecondary,
str::stream() << "slave no longer has secondary status: " << _lastSlaveOkHost});
resetSlaveOkConn();
}
DBClientConnection* DBClientReplicaSet::selectNodeUsingTags(
shared_ptr<ReadPreferenceSetting> readPref) {
if (checkLastHost(readPref.get())) {
LOG(3) << "dbclient_rs selecting compatible last used node " << _lastSlaveOkHost;
return _lastSlaveOkConn.get();
}
ReplicaSetMonitorPtr monitor = _getMonitor();
auto selectedNodeStatus = monitor->getHostOrRefresh(*readPref);
if (!selectedNodeStatus.isOK()) {
LOG(3) << "dbclient_rs no compatible node found"
<< causedBy(redact(selectedNodeStatus.getStatus()));
return nullptr;
}
const HostAndPort selectedNode = std::move(selectedNodeStatus.getValue());
// We are now about to get a new connection from the pool, so cleanup
// the current one and release it back to the pool.
resetSlaveOkConn();
_lastReadPref = readPref;
_lastSlaveOkHost = selectedNode;
// Primary connection is special because it is the only connection that is
// versioned in mongos. Therefore, we have to make sure that this object
// maintains only one connection to the primary and use that connection
// every time we need to talk to the primary.
if (monitor->isPrimary(selectedNode)) {
checkMaster();
LOG(3) << "dbclient_rs selecting primary node " << selectedNode << endl;
_lastSlaveOkConn.reset(_master.get());
return _master.get();
}
// Needs to perform a dynamic_cast because we need to set the replSet
// callback. We should eventually not need this after we remove the
// callback.
DBClientConnection* newConn = dynamic_cast<DBClientConnection*>(
globalConnPool.get(_lastSlaveOkHost.toString(), _so_timeout));
// Assert here instead of returning NULL since the contract of this method is such
// that returning NULL means none of the nodes were good, which is not the case here.
uassert(16532,
str::stream() << "Failed to connect to " << _lastSlaveOkHost.toString(),
newConn != NULL);
_lastSlaveOkConn.reset(newConn);
_lastSlaveOkConn->setParentReplSetName(_setName);
_lastSlaveOkConn->setRequestMetadataWriter(getRequestMetadataWriter());
_lastSlaveOkConn->setReplyMetadataReader(getReplyMetadataReader());
if (_authPooledSecondaryConn) {
_authConnection(_lastSlaveOkConn.get());
} else {
// Mongos pooled connections are authenticated through
// ShardingConnectionHook::onCreate().
}
LOG(3) << "dbclient_rs selecting node " << _lastSlaveOkHost << endl;
return _lastSlaveOkConn.get();
}
void DBClientReplicaSet::say(Message& toSend, bool isRetry, string* actualServer) {
if (!isRetry)
_lazyState = LazyState();
const int lastOp = toSend.operation();
if (lastOp == dbQuery) {
// TODO: might be possible to do this faster by changing api
DbMessage dm(toSend);
QueryMessage qm(dm);
shared_ptr<ReadPreferenceSetting> readPref(_extractReadPref(qm.query, qm.queryOptions));
if (_isSecondaryQuery(qm.ns, qm.query, *readPref)) {
LOG(3) << "dbclient_rs say using secondary or tagged node selection in "
<< _getMonitor()->getName() << ", read pref is " << readPref->toBSON()
<< " (primary : "
<< (_master.get() != NULL ? _master->getServerAddress() : "[not cached]")
<< ", lastTagged : " << (_lastSlaveOkConn.get() != NULL
? _lastSlaveOkConn->getServerAddress()
: "[not cached]")
<< ")" << endl;
string lastNodeErrMsg;
for (size_t retry = 0; retry < MAX_RETRY; retry++) {
_lazyState._retries = retry;
try {
DBClientConnection* conn = selectNodeUsingTags(readPref);
if (conn == NULL) {
break;
}
if (actualServer != NULL) {
*actualServer = conn->getServerAddress();
}
conn->say(toSend);
_lazyState._lastOp = lastOp;
_lazyState._secondaryQueryOk = true;
_lazyState._lastClient = conn;
} catch (const DBException& ex) {
const Status status = ex.toStatus();
lastNodeErrMsg = str::stream() << "can't callLazy replica set node "
<< _lastSlaveOkHost.toString() << ": "
<< status.reason();
_invalidateLastSlaveOkCache({status.code(), lastNodeErrMsg});
continue;
}
return;
}
StringBuilder assertMsg;
assertMsg << "Failed to call say, no good nodes in " << _getMonitor()->getName();
if (!lastNodeErrMsg.empty()) {
assertMsg << ", last error: " << lastNodeErrMsg;
}
uasserted(16380, assertMsg.str());
}
}
LOG(3) << "dbclient_rs say to primary node in " << _getMonitor()->getName() << endl;
DBClientConnection* master = checkMaster();
if (actualServer)
*actualServer = master->getServerAddress();
_lazyState._lastOp = lastOp;
_lazyState._secondaryQueryOk = false;
// Don't retry requests to primary since there is only one host to try
_lazyState._retries = MAX_RETRY;
_lazyState._lastClient = master;
master->say(toSend);
return;
}
bool DBClientReplicaSet::recv(Message& m) {
verify(_lazyState._lastClient);
// TODO: It would be nice if we could easily wrap a conn error as a result error
try {
return _lazyState._lastClient->recv(m);
} catch (DBException& e) {
log() << "could not receive data from " << _lazyState._lastClient->toString()
<< causedBy(redact(e));
return false;
}
}
void DBClientReplicaSet::checkResponse(const char* data,
int nReturned,
bool* retry,
string* targetHost) {
// For now, do exactly as we did before, so as not to break things. In general though, we
// should fix this so checkResponse has a more consistent contract.
if (!retry) {
if (_lazyState._lastClient)
return _lazyState._lastClient->checkResponse(data, nReturned);
else
return checkMaster()->checkResponse(data, nReturned);
}
*retry = false;
if (targetHost && _lazyState._lastClient)
*targetHost = _lazyState._lastClient->getServerAddress();
else if (targetHost)
*targetHost = "";
if (!_lazyState._lastClient)
return;
// nReturned == 1 means that we got one result back, which might be an error
// nReturned == -1 is a sentinel value for "no data returned" aka (usually) network problem
// If neither, this must be a query result so our response is ok wrt the replica set
if (nReturned != 1 && nReturned != -1)
return;
BSONObj dataObj;
if (nReturned == 1)
dataObj = BSONObj(data);
// Check if we should retry here
if (_lazyState._lastOp == dbQuery && _lazyState._secondaryQueryOk) {
// query could potentially go to a secondary, so see if this is an error (or empty) and
// retry if we're not past our retry limit.
if (nReturned == -1 /* no result, maybe network problem */ ||
(hasErrField(dataObj) && !dataObj["code"].eoo() &&
dataObj["code"].Int() == ErrorCodes::NotMasterOrSecondary)) {
if (_lazyState._lastClient == _lastSlaveOkConn.get()) {
isntSecondary();
} else if (_lazyState._lastClient == _master.get()) {
isntMaster();
} else {
warning() << "passed " << redact(dataObj) << " but last rs client "
<< _lazyState._lastClient->toString() << " is not master or secondary"
<< endl;
}
if (_lazyState._retries < static_cast<int>(MAX_RETRY)) {
_lazyState._retries++;
*retry = true;
} else {
log() << "too many retries (" << _lazyState._retries
<< "), could not get data from replica set" << endl;
}
}
} else if (_lazyState._lastOp == dbQuery) {
// if query could not potentially go to a secondary, just mark the master as bad
if (nReturned == -1 /* no result, maybe network problem */ ||
(hasErrField(dataObj) && !dataObj["code"].eoo() &&
dataObj["code"].Int() == ErrorCodes::NotMasterNoSlaveOk)) {
if (_lazyState._lastClient == _master.get()) {
isntMaster();
}
}
}
}
rpc::UniqueReply DBClientReplicaSet::runCommandWithMetadata(StringData database,
StringData command,
const BSONObj& metadata,
const BSONObj& commandArgs) {
auto ret = runCommandWithMetadataAndTarget(database, command, metadata, commandArgs);
return std::move(std::get<0>(ret));
}
std::tuple<rpc::UniqueReply, DBClientWithCommands*>
DBClientReplicaSet::runCommandWithMetadataAndTarget(StringData database,
StringData command,
const BSONObj& metadata,
const BSONObj& commandArgs) {
// This overload exists so we can parse out the read preference and then use server
// selection directly without having to re-parse the raw message.
// TODO: eventually we will want to pass the metadata before serializing it to BSON
// so we don't have to re-parse it, however, that will come with its own set of
// complications (e.g. some kind of base class or concept for MetadataSerializable
// objects). For now we do it the stupid way.
auto ssm = uassertStatusOK(rpc::ServerSelectionMetadata::readFromMetadata(
metadata.getField(rpc::ServerSelectionMetadata::fieldName())));
// If we didn't get a readPref with this query, we assume SecondaryPreferred if secondaryOk
// is true, and PrimaryOnly otherwise. This logic is replicated from _extractReadPref.
auto defaultReadPref = ssm.isSecondaryOk()
? ReadPreferenceSetting(ReadPreference::SecondaryPreferred, TagSet())
: ReadPreferenceSetting(ReadPreference::PrimaryOnly, TagSet::primaryOnly());
auto readPref = ssm.getReadPreference().get_value_or(defaultReadPref);
if (readPref.pref == ReadPreference::PrimaryOnly ||
// If the command is not runnable on a secondary, we run it on the primary
// regardless of the read preference.
!_isSecondaryCommand(command, commandArgs)) {
auto conn = checkMaster();
return std::make_tuple(
conn->runCommandWithMetadata(database, command, metadata, commandArgs), conn);
}
auto rpShared = std::make_shared<ReadPreferenceSetting>(std::move(readPref));
for (size_t retry = 0; retry < MAX_RETRY; ++retry) {
try {
auto* conn = selectNodeUsingTags(rpShared);
if (conn == nullptr) {
break;
}
// We can't move database and command in case this throws
// and we retry.
return std::make_tuple(
conn->runCommandWithMetadata(database, command, metadata, commandArgs), conn);
} catch (const DBException& ex) {
_invalidateLastSlaveOkCache(ex.toStatus());
}
}
uasserted(ErrorCodes::NodeNotFound,
str::stream() << "Could not satisfy $readPreference of '" << readPref.toBSON() << "' "
<< "while attempting to run command "
<< command);
}
bool DBClientReplicaSet::call(Message& toSend,
Message& response,
bool assertOk,
string* actualServer) {
const char* ns = 0;
if (toSend.operation() == dbQuery) {
// TODO: might be possible to do this faster by changing api
DbMessage dm(toSend);
QueryMessage qm(dm);
ns = qm.ns;
shared_ptr<ReadPreferenceSetting> readPref(_extractReadPref(qm.query, qm.queryOptions));
if (_isSecondaryQuery(ns, qm.query, *readPref)) {
LOG(3) << "dbclient_rs call using secondary or tagged node selection in "
<< _getMonitor()->getName() << ", read pref is " << readPref->toBSON()
<< " (primary : "
<< (_master.get() != NULL ? _master->getServerAddress() : "[not cached]")
<< ", lastTagged : " << (_lastSlaveOkConn.get() != NULL
? _lastSlaveOkConn->getServerAddress()
: "[not cached]")
<< ")" << endl;
for (size_t retry = 0; retry < MAX_RETRY; retry++) {
try {
DBClientConnection* conn = selectNodeUsingTags(readPref);
if (conn == NULL) {
return false;
}
if (actualServer != NULL) {
*actualServer = conn->getServerAddress();
}
return conn->call(toSend, response, assertOk, nullptr);
} catch (const DBException& ex) {
if (actualServer)
*actualServer = "";
const Status status = ex.toStatus();
_invalidateLastSlaveOkCache(
{status.code(),
str::stream() << "can't call replica set node " << _lastSlaveOkHost << ": "
<< status.reason()});
}
}
// Was not able to successfully send after max retries
return false;
}
}
LOG(3) << "dbclient_rs call to primary node in " << _getMonitor()->getName() << endl;
DBClientConnection* m = checkMaster();
if (actualServer)
*actualServer = m->getServerAddress();
if (!m->call(toSend, response, assertOk, nullptr))
return false;
if (ns) {
QueryResult::View res = response.singleData().view2ptr();
if (res.getNReturned() == 1) {
BSONObj x(res.data());
if (str::contains(ns, "$cmd")) {
if (isNotMasterErrorString(x["errmsg"]))
isntMaster();
} else {
if (isNotMasterErrorString(getErrField(x)))
isntMaster();
}
}
}
return true;
}
void DBClientReplicaSet::_invalidateLastSlaveOkCache(const Status& status) {
// This is not wrapped in with if (_lastSlaveOkConn && _lastSlaveOkConn->isFailed()) because
// there are certain exceptions that will not make the connection be labeled as failed. For
// example, asserts 13079, 13080, 16386
_getMonitor()->failedHost(_lastSlaveOkHost, status);
resetSlaveOkConn();
}
void DBClientReplicaSet::reset() {
resetSlaveOkConn();
_lazyState._lastClient = NULL;
_lastReadPref.reset();
}
void DBClientReplicaSet::setAuthPooledSecondaryConn(bool setting) {
_authPooledSecondaryConn = setting;
}
void DBClientReplicaSet::resetMaster() {
if (_master.get() == _lastSlaveOkConn.get()) {
_lastSlaveOkConn.release();
_lastSlaveOkHost = HostAndPort();
}
_master.reset();
_masterHost = HostAndPort();
}
void DBClientReplicaSet::resetSlaveOkConn() {
if (_lastSlaveOkConn.get() == _master.get()) {
_lastSlaveOkConn.release();
} else if (_lastSlaveOkConn.get() != NULL) {
if (_authPooledSecondaryConn) {
logoutAll(_lastSlaveOkConn.get());
} else {
// Mongos pooled connections are all authenticated with the same credentials;
// so no need to logout.
}
// If the connection was bad, the pool will clean it up.
globalConnPool.release(_lastSlaveOkHost.toString(), _lastSlaveOkConn.release());
}
_lastSlaveOkHost = HostAndPort();
}
} // namespace mongo
|