-
Notifications
You must be signed in to change notification settings - Fork 28
/
Copy pathmultimaster.c
5551 lines (4997 loc) · 156 KB
/
multimaster.c
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
/*
* multimaster.c
*
* Multimaster based on logical replication
*
*/
#include <unistd.h>
#include <sys/time.h>
#include <time.h>
#include "postgres.h"
#include "funcapi.h"
#include "fmgr.h"
#include "miscadmin.h"
#include "pg_socket.h"
#include "libpq-fe.h"
#include "lib/stringinfo.h"
#include "libpq/pqformat.h"
#include "common/username.h"
#include "postmaster/postmaster.h"
#include "postmaster/bgworker.h"
#include "storage/lwlock.h"
#include "storage/s_lock.h"
#include "storage/spin.h"
#include "storage/lmgr.h"
#include "storage/shmem.h"
#include "storage/ipc.h"
#include "storage/procarray.h"
#include "access/xlogdefs.h"
#include "access/xact.h"
#include "access/xtm.h"
#include "access/transam.h"
#include "access/subtrans.h"
#include "access/commit_ts.h"
#include "access/xlog.h"
#include "storage/proc.h"
#include "executor/executor.h"
#include "access/twophase.h"
#include "utils/guc.h"
#include "utils/hsearch.h"
#include "utils/timeout.h"
#include "utils/tqual.h"
#include "utils/array.h"
#include "utils/builtins.h"
#include "utils/memutils.h"
#include "commands/dbcommands.h"
#include "commands/extension.h"
#include "commands/sequence.h"
#include "postmaster/autovacuum.h"
#include "storage/pmsignal.h"
#include "storage/proc.h"
#include "utils/syscache.h"
#include "utils/lsyscache.h"
#include "replication/walsender.h"
#include "replication/walsender_private.h"
#include "replication/slot.h"
#include "replication/message.h"
#include "port/atomics.h"
#include "tcop/utility.h"
#include "nodes/makefuncs.h"
#include "access/htup_details.h"
#include "catalog/indexing.h"
#include "catalog/namespace.h"
#include "catalog/pg_constraint_fn.h"
#include "catalog/pg_proc.h"
#include "pglogical_output/hooks.h"
#include "parser/analyze.h"
#include "parser/parse_relation.h"
#include "parser/parse_type.h"
#include "parser/parse_func.h"
#include "catalog/pg_class.h"
#include "catalog/pg_type.h"
#include "tcop/pquery.h"
#include "lib/ilist.h"
#include "multimaster.h"
#include "ddd.h"
#include "state.h"
typedef struct {
TransactionId xid; /* local transaction ID */
GlobalTransactionId gtid; /* global transaction ID assigned by coordinator of transaction */
bool isTwoPhase; /* user level 2PC */
bool isReplicated; /* transaction on replica */
bool isDistributed; /* transaction performed INSERT/UPDATE/DELETE and has to be replicated to other nodes */
bool isPrepared; /* transaction is prepared at first stage of 2PC */
bool isSuspended; /* prepared transaction is suspended because coordinator node is switch to offline */
bool isTransactionBlock; /* is transaction block */
bool containsDML; /* transaction contains DML statements */
bool isActive; /* transaction is active (nActiveTransaction counter is incremented) */
XidStatus status; /* transaction status */
csn_t snapshot; /* transaction snapshot */
csn_t csn; /* CSN */
pgid_t gid; /* global transaction identifier (used by 2pc) */
} MtmCurrentTrans;
typedef enum
{
MTM_STATE_LOCK_ID
} MtmLockIds;
#define MTM_SHMEM_SIZE (128*1024*1024)
#define MTM_HASH_SIZE 100003
#define MTM_MAP_SIZE MTM_HASH_SIZE
#define MIN_WAIT_TIMEOUT 1000
#define MAX_WAIT_TIMEOUT 100000
#define MAX_WAIT_LOOPS 10000 // 1000000
#define STATUS_POLL_DELAY USECS_PER_SEC
void _PG_init(void);
void _PG_fini(void);
PG_MODULE_MAGIC;
PG_FUNCTION_INFO_V1(mtm_start_replication);
PG_FUNCTION_INFO_V1(mtm_stop_replication);
PG_FUNCTION_INFO_V1(mtm_stop_node);
PG_FUNCTION_INFO_V1(mtm_add_node);
PG_FUNCTION_INFO_V1(mtm_poll_node);
PG_FUNCTION_INFO_V1(mtm_recover_node);
PG_FUNCTION_INFO_V1(mtm_resume_node);
PG_FUNCTION_INFO_V1(mtm_get_snapshot);
PG_FUNCTION_INFO_V1(mtm_get_csn);
PG_FUNCTION_INFO_V1(mtm_get_trans_by_gid);
PG_FUNCTION_INFO_V1(mtm_get_trans_by_xid);
PG_FUNCTION_INFO_V1(mtm_get_last_csn);
PG_FUNCTION_INFO_V1(mtm_get_nodes_state);
PG_FUNCTION_INFO_V1(mtm_get_cluster_state);
PG_FUNCTION_INFO_V1(mtm_collect_cluster_info);
PG_FUNCTION_INFO_V1(mtm_make_table_local);
PG_FUNCTION_INFO_V1(mtm_dump_lock_graph);
PG_FUNCTION_INFO_V1(mtm_inject_2pc_error);
PG_FUNCTION_INFO_V1(mtm_check_deadlock);
PG_FUNCTION_INFO_V1(mtm_referee_poll);
PG_FUNCTION_INFO_V1(mtm_broadcast_table);
PG_FUNCTION_INFO_V1(mtm_copy_table);
static Snapshot MtmGetSnapshot(Snapshot snapshot);
static void MtmInitialize(void);
static void MtmXactCallback(XactEvent event, void *arg);
static void MtmBeginTransaction(MtmCurrentTrans* x);
static void MtmPrePrepareTransaction(MtmCurrentTrans* x);
static void MtmPostPrepareTransaction(MtmCurrentTrans* x);
static void MtmAbortPreparedTransaction(MtmCurrentTrans* x);
static void MtmPreCommitPreparedTransaction(MtmCurrentTrans* x);
static void MtmEndTransaction(MtmCurrentTrans* x, bool commit);
static bool MtmTwoPhaseCommit(MtmCurrentTrans* x);
static TransactionId MtmGetOldestXmin(Relation rel, bool ignoreVacuum);
static bool MtmXidInMVCCSnapshot(TransactionId xid, Snapshot snapshot);
static TransactionId MtmAdjustOldestXid(TransactionId xid);
static bool MtmDetectGlobalDeadLock(PGPROC* proc);
static void MtmAddSubtransactions(MtmTransState* ts, TransactionId* subxids, int nSubxids);
static char const* MtmGetName(void);
static size_t MtmGetTransactionStateSize(void);
static void MtmSerializeTransactionState(void* ctx);
static void MtmDeserializeTransactionState(void* ctx);
static void MtmInitializeSequence(int64* start, int64* step);
static void* MtmCreateSavepointContext(void);
static void MtmRestoreSavepointContext(void* ctx);
static void MtmReleaseSavepointContext(void* ctx);
static void MtmSetRemoteFunction(char const* list, void* extra);
static void MtmCheckClusterLock(void);
static void MtmCheckSlots(void);
static void MtmAddSubtransactions(MtmTransState* ts, TransactionId *subxids, int nSubxids);
static void MtmShmemStartup(void);
static BgwPool* MtmPoolConstructor(void);
static bool MtmRunUtilityStmt(PGconn* conn, char const* sql, char **errmsg);
static void MtmBroadcastUtilityStmt(char const* sql, bool ignoreError, int forceOnNode);
static void MtmProcessDDLCommand(char const* queryString, bool transactional);
static void MtmLockCluster(void);
static void MtmUnlockCluster(void);
MtmState* Mtm;
VacuumStmt* MtmVacuumStmt;
IndexStmt* MtmIndexStmt;
DropStmt* MtmDropStmt;
void* MtmTablespaceStmt; /* CREATE/DELETE tablespace */
MemoryContext MtmApplyContext;
MtmConnectionInfo* MtmConnections;
HTAB* MtmXid2State;
HTAB* MtmGid2State;
static HTAB* MtmRemoteFunctions;
static HTAB* MtmLocalTables;
static bool MtmIsRecoverySession;
static MtmCurrentTrans MtmTx;
static dlist_head MtmLsnMapping = DLIST_STATIC_INIT(MtmLsnMapping);
static TransactionManager MtmTM =
{
PgTransactionIdGetStatus,
PgTransactionIdSetTreeStatus,
MtmGetSnapshot,
PgGetNewTransactionId,
MtmGetOldestXmin,
PgTransactionIdIsInProgress,
PgGetGlobalTransactionId,
MtmXidInMVCCSnapshot,
MtmDetectGlobalDeadLock,
MtmGetName,
MtmGetTransactionStateSize,
MtmSerializeTransactionState,
MtmDeserializeTransactionState,
MtmInitializeSequence,
MtmCreateSavepointContext,
MtmRestoreSavepointContext,
MtmReleaseSavepointContext
};
char const* const MtmNodeStatusMnem[] =
{
"Disabled",
"Recovery",
"Recovered",
"Online"
};
char const* const MtmTxnStatusMnem[] =
{
"InProgress",
"Committed",
"Aborted",
"Unknown"
};
bool MtmDoReplication;
char* MtmDatabaseName;
char* MtmDatabaseUser;
Oid MtmDatabaseId;
bool MtmBackgroundWorker;
int MtmNodes;
int MtmNodeId;
int MtmReplicationNodeId;
int MtmArbiterPort;
int MtmNodeDisableDelay;
int MtmTransSpillThreshold;
int MtmMaxNodes;
int MtmHeartbeatSendTimeout;
int MtmHeartbeatRecvTimeout;
int MtmMin2PCTimeout;
int MtmMax2PCRatio;
bool MtmUseDtm;
bool MtmUseRDMA;
bool MtmPreserveCommitOrder;
bool MtmVolksWagenMode; /* Pretend to be normal postgres. This means skip some NOTICE's and use local sequences */
bool MtmMajorNode;
char* MtmRefereeConnStr;
static char* MtmConnStrs;
static char* MtmRemoteFunctionsList;
static char* MtmClusterName;
static int MtmQueueSize;
static int MtmWorkers;
static int MtmVacuumDelay;
static int MtmMinRecoveryLag;
static int MtmMaxRecoveryLag;
static int MtmGcPeriod;
static bool MtmIgnoreTablesWithoutPk;
static int MtmLockCount;
static bool MtmBreakConnection;
static bool MtmBypass;
static bool MtmClusterLocked;
static bool MtmInsideTransaction;
static bool MtmReferee;
static bool MtmMonotonicSequences;
static void const* MtmDDLStatement;
static ExecutorStart_hook_type PreviousExecutorStartHook;
static ExecutorFinish_hook_type PreviousExecutorFinishHook;
static ProcessUtility_hook_type PreviousProcessUtilityHook;
static shmem_startup_hook_type PreviousShmemStartupHook;
static seq_nextval_hook_t PreviousSeqNextvalHook;
static void MtmExecutorStart(QueryDesc *queryDesc, int eflags);
static void MtmExecutorFinish(QueryDesc *queryDesc);
static void MtmProcessUtility(Node *parsetree, const char *queryString,
ProcessUtilityContext context, ParamListInfo params,
DestReceiver *dest, char *completionTag);
static void MtmSeqNextvalHook(Oid seqid, int64 next);
static bool MtmAtExitHookRegistered = false;
/*
* Release multimaster main lock if been hold.
* This function is called when backend is terminated because of critical error or when error is catched
* by FINALLY block
*/
void MtmReleaseLocks(void)
{
MtmResetTransaction();
if (MtmInsideTransaction)
{
MtmLock(LW_EXCLUSIVE);
Assert(Mtm->nRunningTransactions > 0);
Mtm->nRunningTransactions -= 1;
MtmInsideTransaction = false;
MtmUnlock();
}
if (MtmClusterLocked) {
MtmUnlockCluster();
}
}
/*
* -------------------------------------------
* Synchronize access to MTM structures.
* Using LWLock seems to be more efficient (at our benchmarks)
* Multimaster uses trash of 2N+1 lwlocks, where N is number of nodes.
* locks[0] is used to synchronize access to multimaster state,
* locks[1..N] are used to provide exclusive access to replication session for each node
* locks[N+1..2*N] are used to synchronize access to distributed lock graph at each node
* -------------------------------------------
*/
// #define DEBUG_MTM_LOCK 1
#if DEBUG_MTM_LOCK
static timestamp_t MtmLockLastReportTime;
static timestamp_t MtmLockElapsedWaitTime;
static timestamp_t MtmLockMaxWaitTime;
static size_t MtmLockHitCount;
#endif
void MtmLock(LWLockMode mode)
{
if (!MtmAtExitHookRegistered) {
atexit(MtmReleaseLocks);
MtmAtExitHookRegistered = true;
}
if (MtmLockCount != 0 && Mtm->lastLockHolder == MyProcPid) {
MtmLockCount += 1;
}
else
{
#if DEBUG_MTM_LOCK
timestamp_t start, stop;
start = MtmGetSystemTime();
#endif
if (MyProc == NULL) { /* Can not wait if have no PGPROC. It can happen at process exit. TODO: without lock we can get race condition and corrupt Mtm state */
return;
}
LWLockAcquire((LWLockId)&Mtm->locks[MTM_STATE_LOCK_ID], mode);
#if DEBUG_MTM_LOCK
stop = MtmGetSystemTime();
MtmLockElapsedWaitTime += stop - start;
if (stop - start > MtmLockMaxWaitTime) {
MtmLockMaxWaitTime = stop - start;
}
MtmLockHitCount += 1;
if (stop - MtmLockLastReportTime > USECS_PER_SEC) {
MTM_LOG1("%d: average lock wait time %lld usec, maximal lock wait time: %lld usec",
MyProcPid, MtmLockElapsedWaitTime/MtmLockHitCount, MtmLockMaxWaitTime);
MtmLockLastReportTime = stop;
MtmLockMaxWaitTime = 0;
MtmLockElapsedWaitTime = 0;
MtmLockHitCount = 0;
}
#endif
if (mode == LW_EXCLUSIVE) {
Assert(MtmLockCount == 0);
Assert(MyProcPid != 0);
Mtm->lastLockHolder = MyProcPid;
Assert(MyProcPid);
MtmLockCount = 1;
} else {
MtmLockCount = 0;
}
}
}
void MtmUnlock(void)
{
if (MtmLockCount != 0 && --MtmLockCount != 0) {
Assert(Mtm->lastLockHolder == MyProcPid);
return;
}
Mtm->lastLockHolder = 0;
/* If we have no PGPROC, then lock was not obtained. */
if (MyProc != NULL)
LWLockRelease((LWLockId)&Mtm->locks[MTM_STATE_LOCK_ID]);
}
void MtmDeepUnlock(void)
{
if (MtmLockCount == 0)
return;
Assert(Mtm->lastLockHolder == MyProcPid);
MtmLockCount = 0;
Mtm->lastLockHolder = 0;
/* If we have no PGPROC, then lock was not obtained. */
if (MyProc != NULL)
LWLockRelease((LWLockId)&Mtm->locks[MTM_STATE_LOCK_ID]);
}
void MtmLockNode(int nodeId, LWLockMode mode)
{
Assert(nodeId > 0 && nodeId <= MtmMaxNodes*2);
LWLockAcquire((LWLockId)&Mtm->locks[nodeId], mode);
}
bool MtmTryLockNode(int nodeId, LWLockMode mode)
{
return LWLockConditionalAcquire((LWLockId)&Mtm->locks[nodeId], mode);
}
void MtmUnlockNode(int nodeId)
{
Assert(nodeId > 0 && nodeId <= MtmMaxNodes*2);
LWLockRelease((LWLockId)&Mtm->locks[nodeId]);
}
/*
* -------------------------------------------
* System time manipulation functions
* -------------------------------------------
*/
timestamp_t MtmGetSystemTime(void)
{
struct timeval tv;
gettimeofday(&tv, NULL);
return (timestamp_t)tv.tv_sec*USECS_PER_SEC + tv.tv_usec;
}
/*
* Get adjusted system time: taking in account time shift
*/
timestamp_t MtmGetCurrentTime(void)
{
return MtmGetSystemTime() + Mtm->timeShift;
}
void MtmSleep(timestamp_t interval)
{
struct timespec ts;
struct timespec rem;
ts.tv_sec = interval/USECS_PER_SEC;
ts.tv_nsec = interval%USECS_PER_SEC*1000;
while (nanosleep(&ts, &rem) < 0) {
Assert(errno == EINTR);
CHECK_FOR_INTERRUPTS();
ts = rem;
}
}
/**
* Return ascending unique timestamp which is used as CSN
*/
csn_t MtmAssignCSN()
{
csn_t csn = MtmGetCurrentTime();
if (csn <= Mtm->csn) {
csn = ++Mtm->csn;
} else {
Mtm->csn = csn;
}
return csn;
}
/**
* "Adjust" system clock if we receive message from future
*/
csn_t MtmSyncClock(csn_t global_csn)
{
csn_t local_csn;
while ((local_csn = MtmAssignCSN()) < global_csn) {
Mtm->timeShift += global_csn - local_csn;
}
return local_csn;
}
/*
* Distribute transaction manager functions
*/
static char const* MtmGetName(void)
{
return MULTIMASTER_NAME;
}
static size_t
MtmGetTransactionStateSize(void)
{
return sizeof(MtmTx);
}
static void
MtmSerializeTransactionState(void* ctx)
{
memcpy(ctx, &MtmTx, sizeof(MtmTx));
}
static void
MtmDeserializeTransactionState(void* ctx)
{
memcpy(&MtmTx, ctx, sizeof(MtmTx));
}
static void
MtmInitializeSequence(int64* start, int64* step)
{
if (MtmVolksWagenMode)
{
*start = 1;
*step = 1;
}
else
{
*start = MtmNodeId;
*step = MtmMaxNodes;
}
}
static void* MtmCreateSavepointContext(void)
{
return (void*)(size_t)MtmTx.containsDML;
}
static void MtmRestoreSavepointContext(void* ctx)
{
MtmTx.containsDML = ctx != NULL;
}
static void MtmReleaseSavepointContext(void* ctx)
{
}
/*
* -------------------------------------------
* Visibility&snapshots
* -------------------------------------------
*/
/*
* Get snapshot of transaction proceed by WAL sender pglogical plugin.
* If it is local transaction or replication node is not in participant mask, then return INVALID_CSN.
* Transaction should be skipped by WAL sender in the following cases:
* 1. Transaction was replicated from some other node and it is not a recovery process.
* 2. State of transaction is unknown
* 3. Replication node is not participated in transaction
*/
csn_t MtmDistributedTransactionSnapshot(TransactionId xid, int nodeId, nodemask_t* participantsMask)
{
csn_t snapshot = INVALID_CSN;
*participantsMask = 0;
MtmLock(LW_SHARED);
if (Mtm->status == MTM_ONLINE) {
MtmTransState* ts = (MtmTransState*)hash_search(MtmXid2State, &xid, HASH_FIND, NULL);
if (ts != NULL) {
*participantsMask = ts->participantsMask;
/* If node is disables, then we are in a process of recovery of this node */
if (!ts->isLocal && BIT_CHECK(ts->participantsMask|Mtm->disabledNodeMask, nodeId-1)) {
snapshot = ts->snapshot;
Assert(ts->gtid.node == MtmNodeId || MtmIsRecoverySession);
} else {
MTM_LOG1("Do not send transaction %s (%llu) to node %d participants mask %llx",
ts->gid, (long64)ts->xid, nodeId, ts->participantsMask);
}
}
}
MtmUnlock();
return snapshot;
}
void MtmSetSnapshot(csn_t globalSnapshot)
{
MtmLock(LW_EXCLUSIVE);
MtmSyncClock(globalSnapshot);
MtmTx.snapshot = globalSnapshot;
MtmUnlock();
}
Snapshot MtmGetSnapshot(Snapshot snapshot)
{
snapshot = PgGetSnapshotData(snapshot);
if (XactIsoLevel == XACT_READ_COMMITTED && MtmTx.snapshot != INVALID_CSN) {
MtmTx.snapshot = MtmGetCurrentTime();
if (TransactionIdIsValid(GetCurrentTransactionIdIfAny())) {
LogLogicalMessage("S", (char*)&MtmTx.snapshot, sizeof(MtmTx.snapshot), true);
}
}
RecentGlobalDataXmin = RecentGlobalXmin = Mtm->oldestXid;
return snapshot;
}
TransactionId MtmGetOldestXmin(Relation rel, bool ignoreVacuum)
{
TransactionId xmin = PgGetOldestXmin(NULL, false); /* consider all backends */
if (TransactionIdIsValid(xmin)) {
MtmLock(LW_EXCLUSIVE);
xmin = MtmAdjustOldestXid(xmin);
MtmUnlock();
}
return xmin;
}
bool MtmXidInMVCCSnapshot(TransactionId xid, Snapshot snapshot)
{
#if TRACE_SLEEP_TIME
static timestamp_t firstReportTime;
static timestamp_t prevReportTime;
static timestamp_t totalSleepTime;
static timestamp_t maxSleepTime;
#endif
timestamp_t delay = MIN_WAIT_TIMEOUT;
int i;
#if DEBUG_LEVEL > 1
timestamp_t start = MtmGetSystemTime();
#endif
Assert(xid != InvalidTransactionId);
if (!MtmUseDtm || TransactionIdPrecedes(xid, Mtm->oldestXid)) {
return PgXidInMVCCSnapshot(xid, snapshot);
}
MtmLock(LW_SHARED);
#if TRACE_SLEEP_TIME
if (firstReportTime == 0) {
firstReportTime = MtmGetCurrentTime();
}
#endif
for (i = 0; i < MAX_WAIT_LOOPS; i++)
{
MtmTransState* ts = (MtmTransState*)hash_search(MtmXid2State, &xid, HASH_FIND, NULL);
if (ts != NULL /*&& ts->status != TRANSACTION_STATUS_IN_PROGRESS*/)
{
if (ts->csn > MtmTx.snapshot) {
MTM_LOG4("%d: tuple with xid=%lld(csn=%lld) is invisible in snapshot %lld",
MyProcPid, (long64)xid, ts->csn, MtmTx.snapshot);
#if DEBUG_LEVEL > 1
if (MtmGetSystemTime() - start > USECS_PER_SEC) {
MTM_ELOG(WARNING, "Backend %d waits for transaction %s (%llu) status %lld usecs", MyProcPid, ts->gid, (long64)xid, MtmGetSystemTime() - start);
}
#endif
MtmUnlock();
return true;
}
if (ts->status == TRANSACTION_STATUS_UNKNOWN)
{
MTM_LOG3("%d: wait for in-doubt transaction %u in snapshot %llu", MyProcPid, xid, MtmTx.snapshot);
MtmUnlock();
#if TRACE_SLEEP_TIME
{
timestamp_t delta, now = MtmGetCurrentTime();
#endif
MtmSleep(delay);
#if TRACE_SLEEP_TIME
delta = MtmGetCurrentTime() - now;
totalSleepTime += delta;
if (delta > maxSleepTime) {
maxSleepTime = delta;
}
if (now > prevReportTime + USECS_PER_SEC*10) {
prevReportTime = now;
if (firstReportTime == 0) {
firstReportTime = now;
} else {
MTM_LOG3("Snapshot sleep %llu of %llu usec (%f%%), maximum=%llu", totalSleepTime, now - firstReportTime, totalSleepTime*100.0/(now - firstReportTime), maxSleepTime);
}
}
}
#endif
if (delay*2 <= MAX_WAIT_TIMEOUT) {
delay *= 2;
}
MtmLock(LW_SHARED);
}
else
{
bool invisible = ts->status != TRANSACTION_STATUS_COMMITTED;
MTM_LOG4("%d: tuple with xid=%lld(csn= %lld) is %s in snapshot %lld",
MyProcPid, (long64)xid, ts->csn, invisible ? "rollbacked" : "committed", MtmTx.snapshot);
MtmUnlock();
#if DEBUG_LEVEL > 1
if (MtmGetSystemTime() - start > USECS_PER_SEC) {
MTM_ELOG(WARNING, "Backend %d waits for %s transaction %s (%llu) %lld usecs", MyProcPid, invisible ? "rollbacked" : "committed",
ts->gid, (long64)xid, MtmGetSystemTime() - start);
}
#endif
return invisible;
}
}
else
{
MTM_LOG4("%d: visibility check is skipped for transaction %llu in snapshot %llu", MyProcPid, (long64)xid, MtmTx.snapshot);
MtmUnlock();
return PgXidInMVCCSnapshot(xid, snapshot);
}
}
MtmUnlock();
#if DEBUG_LEVEL > 1
MTM_ELOG(ERROR, "Failed to get status of XID %llu in %lld usec", (long64)xid, MtmGetSystemTime() - start);
#else
MTM_ELOG(ERROR, "Failed to get status of XID %llu", (long64)xid);
#endif
return true;
}
/*
* There can be different oldest XIDs at different cluster node.
* We collect oldest CSNs from all nodes and choose minimum from them.
* If no such XID can be located, then return previously observed oldest XID
*/
static TransactionId
MtmAdjustOldestXid(TransactionId xid)
{
int i;
csn_t oldestSnapshot = INVALID_CSN;
MtmTransState *prev = NULL;
MtmTransState *ts = (MtmTransState*)hash_search(MtmXid2State, &xid, HASH_FIND, NULL);
MTM_LOG2("%d: MtmAdjustOldestXid(%d): snapshot=%lld, csn=%lld, status=%d", MyProcPid, xid, ts != NULL ? ts->snapshot : 0, ts != NULL ? ts->csn : 0, ts != NULL ? ts->status : -1);
Mtm->gcCount = 0;
if (ts != NULL) {
oldestSnapshot = ts->snapshot;
Assert(oldestSnapshot != INVALID_CSN);
if (Mtm->nodes[MtmNodeId-1].oldestSnapshot < oldestSnapshot) {
Mtm->nodes[MtmNodeId-1].oldestSnapshot = oldestSnapshot;
} else {
oldestSnapshot = Mtm->nodes[MtmNodeId-1].oldestSnapshot;
}
for (i = 0; i < Mtm->nAllNodes; i++) {
if (!BIT_CHECK(Mtm->disabledNodeMask, i)
&& Mtm->nodes[i].oldestSnapshot < oldestSnapshot)
{
oldestSnapshot = Mtm->nodes[i].oldestSnapshot;
}
}
if (oldestSnapshot > MtmVacuumDelay*USECS_PER_SEC) {
oldestSnapshot -= MtmVacuumDelay*USECS_PER_SEC;
} else {
oldestSnapshot = 0;
}
for (ts = Mtm->transListHead;
ts != NULL
&& (ts->status == TRANSACTION_STATUS_ABORTED || ts->status == TRANSACTION_STATUS_COMMITTED)
&& ts->csn < oldestSnapshot
&& !ts->isPinned
&& TransactionIdPrecedes(ts->xid, xid);
prev = ts, ts = ts->next)
{
Assert(!ts->isActive);
if (prev != NULL) {
/* Remove information about too old transactions */
hash_search(MtmXid2State, &prev->xid, HASH_REMOVE, NULL);
hash_search(MtmGid2State, &prev->gid, HASH_REMOVE, NULL);
}
}
if (ts != NULL) {
MTM_LOG2("Adjust(%lld) stop at snashot %lld, xid %lld, pinned=%d, oldestSnaphsot=%lld\n",
(long64)xid, ts->csn, (long64)ts->xid, ts->isPinned, oldestSnapshot);
}
}
if (MtmUseDtm && !MtmVolksWagenMode)
{
if (prev != NULL) {
MTM_LOG2("%d: MtmAdjustOldestXid: oldestXid=%d, prev->xid=%d, prev->status=%s, prev->snapshot=%lld, ts->xid=%d, ts->status=%d, ts->snapshot=%lld, oldestSnapshot=%lld",
MyProcPid, xid, prev->xid, MtmTxnStatusMnem[prev->status], prev->snapshot, (ts ? ts->xid : 0), (ts ? ts->status : -1), (ts ? ts->snapshot : -1), oldestSnapshot);
Mtm->transListHead = prev;
Mtm->oldestXid = xid = prev->xid;
} else if (TransactionIdPrecedes(Mtm->oldestXid, xid)) {
xid = Mtm->oldestXid;
}
} else {
if (prev != NULL) {
MTM_LOG2("%d: MtmAdjustOldestXid: oldestXid=%d, prev->xid=%d, prev->status=%s, prev->snapshot=%lld, ts->xid=%d, ts->status=%d, ts->snapshot=%lld, oldestSnapshot=%lld",
MyProcPid, xid, prev->xid, MtmTxnStatusMnem[prev->status], prev->snapshot, (ts ? ts->xid : 0), (ts ? ts->status : -1), (ts ? ts->snapshot : -1), oldestSnapshot);
Mtm->transListHead = prev;
}
}
if (!MyReplicationSlot) {
MtmCheckSlots();
}
return xid;
}
/*
* -------------------------------------------
* Transaction list manipulation.
* All distributed transactions are linked in L1-list ordered by transaction start time.
* This list is inspected by MtmAdjustOldestXid and transactions which are not used in any snapshot at any node
* are removed from the list and from the hash.
* -------------------------------------------
*/
static void MtmTransactionListAppend(MtmTransState* ts)
{
if (!ts->isEnqueued) {
ts->isEnqueued = true;
ts->next = NULL;
ts->nSubxids = 0;
*Mtm->transListTail = ts;
Mtm->transListTail = &ts->next;
}
}
static void MtmTransactionListInsertAfter(MtmTransState* after, MtmTransState* ts)
{
ts->next = after->next;
after->next = ts;
ts->isEnqueued = true;
if (Mtm->transListTail == &after->next) {
Mtm->transListTail = &ts->next;
}
}
static void MtmAddSubtransactions(MtmTransState* ts, TransactionId* subxids, int nSubxids)
{
int i;
ts->nSubxids = nSubxids;
for (i = 0; i < nSubxids; i++) {
bool found;
MtmTransState* sts;
Assert(TransactionIdIsValid(subxids[i]));
sts = (MtmTransState*)hash_search(MtmXid2State, &subxids[i], HASH_ENTER, &found);
Assert(!found);
sts->isActive = false;
sts->isPinned = false;
sts->status = ts->status;
sts->csn = ts->csn;
sts->votingCompleted = true;
MtmTransactionListInsertAfter(ts, sts);
}
}
void MtmAdjustSubtransactions(MtmTransState* ts)
{
int i;
int nSubxids = ts->nSubxids;
MtmTransState* sts = ts;
for (i = 0; i < nSubxids; i++) {
sts = sts->next;
sts->status = ts->status;
sts->csn = ts->csn;
}
}
/*
* -------------------------------------------
* Transaction control
* -------------------------------------------
*/
static void
MtmXactCallback(XactEvent event, void *arg)
{
switch (event)
{
case XACT_EVENT_START:
MtmBeginTransaction(&MtmTx);
break;
case XACT_EVENT_PRE_PREPARE:
MtmPrePrepareTransaction(&MtmTx);
break;
case XACT_EVENT_POST_PREPARE:
MtmPostPrepareTransaction(&MtmTx);
break;
case XACT_EVENT_ABORT_PREPARED:
MtmAbortPreparedTransaction(&MtmTx);
break;
case XACT_EVENT_PRE_COMMIT_PREPARED:
MtmPreCommitPreparedTransaction(&MtmTx);
break;
case XACT_EVENT_COMMIT:
MtmEndTransaction(&MtmTx, true);
break;
case XACT_EVENT_ABORT:
MtmEndTransaction(&MtmTx, false);
break;
case XACT_EVENT_COMMIT_COMMAND:
if (!MtmTx.isTransactionBlock && !IsSubTransaction()) {
MtmTwoPhaseCommit(&MtmTx);
}
break;
default:
break;
}
}
/*
* Check if this is "normal" user transaction which should be distributed to other nodes
*/
static bool
MtmIsUserTransaction()
{
return !IsAutoVacuumLauncherProcess() &&
IsNormalProcessingMode() &&
MtmDoReplication &&
!am_walsender &&
!MtmBackgroundWorker &&
!IsAutoVacuumWorkerProcess();
}
void
MtmResetTransaction()
{
MtmCurrentTrans* x = &MtmTx;
x->snapshot = INVALID_CSN;
x->xid = InvalidTransactionId;
x->gtid.xid = InvalidTransactionId;
x->isDistributed = false;
x->isPrepared = false;
x->isSuspended = false;
x->isActive = false;
x->isTwoPhase = false;
x->csn = INVALID_CSN;
x->status = TRANSACTION_STATUS_UNKNOWN;
x->gid[0] = '\0';
MtmDDLStatement = NULL;
}
#if 0
static const char* const isoLevelStr[] =
{
"read uncommitted",
"read committed",
"repeatable read",
"serializable"
};
#endif
bool MtmTransIsActive(void)
{
return MtmTx.isActive;
}
static void
MtmBeginTransaction(MtmCurrentTrans* x)
{
if (x->snapshot == INVALID_CSN) {
TransactionId xmin = (Mtm->gcCount >= MtmGcPeriod) ? PgGetOldestXmin(NULL, false) : InvalidTransactionId; /* Get oldest xmin outside critical section */
Assert(!x->isActive);
MtmLock(LW_EXCLUSIVE);
if (TransactionIdIsValid(xmin) && Mtm->gcCount >= MtmGcPeriod) {
MtmAdjustOldestXid(xmin);
}
x->xid = GetCurrentTransactionIdIfAny();
x->isReplicated = MtmIsLogicalReceiver;
x->isDistributed = MtmIsUserTransaction();
x->isPrepared = false;
x->isSuspended = false;
x->isTwoPhase = false;
x->isTransactionBlock = IsTransactionBlock();
/* Application name can be changed using PGAPPNAME environment variable */
if (x->isDistributed && Mtm->status != MTM_ONLINE && strcmp(application_name, MULTIMASTER_ADMIN) != 0
&& strcmp(application_name, MULTIMASTER_BROADCAST_SERVICE) != 0
&& !MtmBypass) {
/* Reject all user's transactions at offline cluster.
* Allow execution of transaction by bg-workers to make it possible to perform recovery.
*/
MtmUnlock();
MTM_ELOG(MtmBreakConnection ? FATAL : ERROR, "Multimaster node is not online: current status %s", MtmNodeStatusMnem[Mtm->status]);
}
x->containsDML = false;
x->gtid.xid = InvalidTransactionId;
x->gid[0] = '\0';
x->status = TRANSACTION_STATUS_IN_PROGRESS;
/*
* Check if there is global multimaster lock preventing new transaction from commit to make a chance to wal-senders to caught-up.
* Allow applying of replicated transactions to avoid deadlock (to caught-up we need active transaction counter to become zero).
* Also allow user to complete explicit 2PC transactions.
*/