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
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
|
/*****************************************************************************
Copyright (c) 1996, 2017, Oracle and/or its affiliates. All Rights Reserved.
Copyright (c) 2017, MariaDB Corporation.
This program is free software; you can redistribute it and/or modify it under
the terms of the GNU General Public License as published by the Free Software
Foundation; version 2 of the License.
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 General Public License for more details.
You should have received a copy of the GNU General Public License along with
this program; if not, write to the Free Software Foundation, Inc.,
51 Franklin Street, Suite 500, Boston, MA 02110-1335 USA
*****************************************************************************/
/**************************************************//**
@file trx/trx0purge.cc
Purge old versions
Created 3/26/1996 Heikki Tuuri
*******************************************************/
#include "ha_prototypes.h"
#include "trx0purge.h"
#include "fsp0fsp.h"
#include "fut0fut.h"
#include "mach0data.h"
#include "mtr0log.h"
#include "os0thread.h"
#include "que0que.h"
#include "read0read.h"
#include "row0purge.h"
#include "row0upd.h"
#include "srv0mon.h"
#include "fsp0sysspace.h"
#include "srv0srv.h"
#include "srv0start.h"
#include "sync0sync.h"
#include "trx0rec.h"
#include "trx0roll.h"
#include "trx0rseg.h"
#include "trx0trx.h"
/** Maximum allowable purge history length. <=0 means 'infinite'. */
ulong srv_max_purge_lag = 0;
/** Max DML user threads delay in micro-seconds. */
ulong srv_max_purge_lag_delay = 0;
/** The global data structure coordinating a purge */
purge_sys_t* purge_sys;
/** A dummy undo record used as a return value when we have a whole undo log
which needs no purge */
trx_undo_rec_t trx_purge_dummy_rec;
#ifdef UNIV_DEBUG
my_bool srv_purge_view_update_only_debug;
#endif /* UNIV_DEBUG */
/** Sentinel value */
const TrxUndoRsegs TrxUndoRsegsIterator::NullElement(UINT64_UNDEFINED);
/** Constructor */
TrxUndoRsegsIterator::TrxUndoRsegsIterator()
:
m_trx_undo_rsegs(NullElement),
m_iter(m_trx_undo_rsegs.end())
{
}
/** Sets the next rseg to purge in purge_sys.
@return whether anything is to be purged */
inline
bool
TrxUndoRsegsIterator::set_next()
{
mutex_enter(&purge_sys->pq_mutex);
/* Only purge consumes events from the priority queue, user
threads only produce the events. */
/* Check if there are more rsegs to process in the
current element. */
if (m_iter != m_trx_undo_rsegs.end()) {
/* We are still processing rollback segment from
the same transaction and so expected transaction
number shouldn't increase. Undo increment of
expected trx_no done by caller assuming rollback
segments from given transaction are done. */
purge_sys->iter.trx_no = (*m_iter)->last_trx_no;
} else if (!purge_sys->purge_queue.empty()) {
/* Read the next element from the queue.
Combine elements if they have same transaction number.
This can happen if a transaction shares redo rollback segment
with another transaction that has already added it to purge
queue and former transaction also needs to schedule non-redo
rollback segment for purge. */
m_trx_undo_rsegs = NullElement;
purge_pq_t& purge_queue = purge_sys->purge_queue;
while (!purge_queue.empty()) {
if (m_trx_undo_rsegs.get_trx_no() == UINT64_UNDEFINED) {
m_trx_undo_rsegs = purge_queue.top();
} else if (purge_queue.top().get_trx_no() ==
m_trx_undo_rsegs.get_trx_no()) {
m_trx_undo_rsegs.append(
purge_queue.top());
} else {
break;
}
purge_queue.pop();
}
m_iter = m_trx_undo_rsegs.begin();
} else {
/* Queue is empty, reset iterator. */
m_trx_undo_rsegs = NullElement;
m_iter = m_trx_undo_rsegs.end();
mutex_exit(&purge_sys->pq_mutex);
purge_sys->rseg = NULL;
return false;
}
purge_sys->rseg = *m_iter++;
mutex_exit(&purge_sys->pq_mutex);
ut_a(purge_sys->rseg != NULL);
mutex_enter(&purge_sys->rseg->mutex);
ut_a(purge_sys->rseg->last_page_no != FIL_NULL);
ut_ad(purge_sys->rseg->last_trx_no == m_trx_undo_rsegs.get_trx_no());
/* We assume in purge of externally stored fields that space id is
in the range of UNDO tablespace space ids */
ut_a(purge_sys->rseg->space == TRX_SYS_SPACE
|| srv_is_undo_tablespace(purge_sys->rseg->space));
ut_a(purge_sys->iter.trx_no <= purge_sys->rseg->last_trx_no);
purge_sys->iter.trx_no = purge_sys->rseg->last_trx_no;
purge_sys->hdr_offset = purge_sys->rseg->last_offset;
purge_sys->hdr_page_no = purge_sys->rseg->last_page_no;
mutex_exit(&purge_sys->rseg->mutex);
return(true);
}
/** Build a purge 'query' graph. The actual purge is performed by executing
this query graph.
@param[in,out] sess the purge session
@return own: the query graph */
static
que_t*
trx_purge_graph_build(sess_t* sess)
{
ut_a(srv_n_purge_threads > 0);
/* A purge transaction is not a real transaction, we use a transaction
here only because the query threads code requires it. It is otherwise
quite unnecessary. We should get rid of it eventually. */
trx_t* trx = sess->trx;
ut_ad(trx->sess == sess);
trx->id = 0;
trx->start_time = ut_time();
trx->state = TRX_STATE_ACTIVE;
trx->op_info = "purge trx";
mem_heap_t* heap = mem_heap_create(512);
que_fork_t* fork = que_fork_create(
NULL, NULL, QUE_FORK_PURGE, heap);
fork->trx = trx;
for (ulint i = 0; i < srv_n_purge_threads; ++i) {
que_thr_t* thr = que_thr_create(fork, heap, NULL);
thr->child = row_purge_node_create(thr, heap);
}
return(fork);
}
/** Construct the purge system. */
purge_sys_t::purge_sys_t()
: sess(sess_open()), latch(), event(os_event_create(0)),
n_stop(0), running(false), state(PURGE_STATE_INIT),
query(trx_purge_graph_build(sess)),
view(), n_submitted(0), n_completed(0),
iter(), limit(),
#ifdef UNIV_DEBUG
done(),
#endif /* UNIV_DEBUG */
next_stored(false), rseg(NULL),
page_no(0), offset(0), hdr_page_no(0), hdr_offset(0),
rseg_iter(), purge_queue(), pq_mutex(), undo_trunc()
{
ut_ad(!purge_sys);
rw_lock_create(trx_purge_latch_key, &latch, SYNC_PURGE_LATCH);
mutex_create(LATCH_ID_PURGE_SYS_PQ, &pq_mutex);
}
/** Destruct the purge system. */
purge_sys_t::~purge_sys_t()
{
ut_ad(this == purge_sys);
que_graph_free(query);
ut_a(sess->trx->id == 0);
sess->trx->state = TRX_STATE_NOT_STARTED;
sess_close(sess);
view.close();
rw_lock_free(&latch);
/* rw_lock_free() already called latch.~rw_lock_t(); tame the
debug assertions when the destructor will be called once more. */
ut_ad(latch.magic_n == 0);
ut_d(latch.magic_n = RW_LOCK_MAGIC_N);
mutex_free(&pq_mutex);
os_event_destroy(event);
}
/*================ UNDO LOG HISTORY LIST =============================*/
/********************************************************************//**
Adds the update undo log as the first log in the history list. Removes the
update undo log segment from the rseg slot if it is too big for reuse. */
void
trx_purge_add_update_undo_to_history(
/*=================================*/
trx_t* trx, /*!< in: transaction */
page_t* undo_page, /*!< in: update undo log header page,
x-latched */
mtr_t* mtr) /*!< in: mtr */
{
trx_undo_t* undo = trx->rsegs.m_redo.update_undo;
trx_rseg_t* rseg = undo->rseg;
trx_rsegf_t* rseg_header = trx_rsegf_get(
rseg->space, rseg->page_no, mtr);
trx_ulogf_t* undo_header = undo_page + undo->hdr_offset;
if (undo->state != TRX_UNDO_CACHED) {
ulint hist_size;
#ifdef UNIV_DEBUG
trx_usegf_t* seg_header = undo_page + TRX_UNDO_SEG_HDR;
#endif /* UNIV_DEBUG */
/* The undo log segment will not be reused */
if (UNIV_UNLIKELY(undo->id >= TRX_RSEG_N_SLOTS)) {
ib::fatal() << "undo->id is " << undo->id;
}
trx_rsegf_set_nth_undo(rseg_header, undo->id, FIL_NULL, mtr);
MONITOR_DEC(MONITOR_NUM_UNDO_SLOT_USED);
hist_size = mtr_read_ulint(
rseg_header + TRX_RSEG_HISTORY_SIZE, MLOG_4BYTES, mtr);
ut_ad(undo->size == flst_get_len(
seg_header + TRX_UNDO_PAGE_LIST));
mlog_write_ulint(
rseg_header + TRX_RSEG_HISTORY_SIZE,
hist_size + undo->size, MLOG_4BYTES, mtr);
}
/* Before any transaction-generating background threads or the
purge have been started, recv_recovery_rollback_active() can
start transactions in row_merge_drop_temp_indexes() and
fts_drop_orphaned_tables(), and roll back recovered transactions.
Arbitrary user transactions may be executed when all the undo log
related background processes (including purge) are disabled due to
innodb_force_recovery=2 or innodb_force_recovery=3.
DROP TABLE may be executed at any innodb_force_recovery level.
After the purge thread has been given permission to exit,
in fast shutdown, we may roll back transactions (trx->undo_no==0)
in THD::cleanup() invoked from unlink_thd(), and we may also
continue to execute user transactions. */
ut_ad(srv_undo_sources
|| ((srv_startup_is_before_trx_rollback_phase
|| trx_rollback_or_clean_is_active)
&& purge_sys->state == PURGE_STATE_INIT)
|| (srv_force_recovery >= SRV_FORCE_NO_BACKGROUND
&& purge_sys->state == PURGE_STATE_DISABLED)
|| ((trx->undo_no == 0 || trx->in_mysql_trx_list
|| trx->persistent_stats)
&& srv_fast_shutdown));
/* Add the log as the first in the history list */
flst_add_first(rseg_header + TRX_RSEG_HISTORY,
undo_header + TRX_UNDO_HISTORY_NODE, mtr);
my_atomic_addlint(&trx_sys->rseg_history_len, 1);
/* Write the trx number to the undo log header */
mlog_write_ull(undo_header + TRX_UNDO_TRX_NO, trx->no, mtr);
/* Write information about delete markings to the undo log header */
if (!undo->del_marks) {
mlog_write_ulint(undo_header + TRX_UNDO_DEL_MARKS, FALSE,
MLOG_2BYTES, mtr);
}
if (rseg->last_page_no == FIL_NULL) {
rseg->last_page_no = undo->hdr_page_no;
rseg->last_offset = undo->hdr_offset;
rseg->last_trx_no = trx->no;
rseg->last_del_marks = undo->del_marks;
}
}
/** Remove undo log header from the history list.
@param[in,out] rseg_hdr rollback segment header
@param[in] log_hdr undo log segment header
@param[in,out] mtr mini transaction. */
static
void
trx_purge_remove_log_hdr(
trx_rsegf_t* rseg_hdr,
trx_ulogf_t* log_hdr,
mtr_t* mtr)
{
flst_remove(rseg_hdr + TRX_RSEG_HISTORY,
log_hdr + TRX_UNDO_HISTORY_NODE, mtr);
my_atomic_addlint(&trx_sys->rseg_history_len, -1);
}
/** Free an undo log segment, and remove the header from the history list.
@param[in,out] rseg rollback segment
@param[in] hdr_addr file address of log_hdr */
static
void
trx_purge_free_segment(trx_rseg_t* rseg, fil_addr_t hdr_addr)
{
mtr_t mtr;
trx_rsegf_t* rseg_hdr;
trx_ulogf_t* log_hdr;
trx_usegf_t* seg_hdr;
ulint seg_size;
ulint hist_size;
bool marked = false;
for (;;) {
page_t* undo_page;
mtr_start(&mtr);
mutex_enter(&rseg->mutex);
rseg_hdr = trx_rsegf_get(rseg->space, rseg->page_no, &mtr);
undo_page = trx_undo_page_get(
page_id_t(rseg->space, hdr_addr.page), &mtr);
seg_hdr = undo_page + TRX_UNDO_SEG_HDR;
log_hdr = undo_page + hdr_addr.boffset;
/* Mark the last undo log totally purged, so that if the
system crashes, the tail of the undo log will not get accessed
again. The list of pages in the undo log tail gets inconsistent
during the freeing of the segment, and therefore purge should
not try to access them again. */
if (!marked) {
marked = true;
mlog_write_ulint(
log_hdr + TRX_UNDO_DEL_MARKS, FALSE,
MLOG_2BYTES, &mtr);
}
if (fseg_free_step_not_header(
seg_hdr + TRX_UNDO_FSEG_HEADER, false, &mtr)) {
break;
}
mutex_exit(&rseg->mutex);
mtr_commit(&mtr);
}
/* The page list may now be inconsistent, but the length field
stored in the list base node tells us how big it was before we
started the freeing. */
seg_size = flst_get_len(seg_hdr + TRX_UNDO_PAGE_LIST);
/* We may free the undo log segment header page; it must be freed
within the same mtr as the undo log header is removed from the
history list: otherwise, in case of a database crash, the segment
could become inaccessible garbage in the file space. */
trx_purge_remove_log_hdr(rseg_hdr, log_hdr, &mtr);
do {
/* Here we assume that a file segment with just the header
page can be freed in a few steps, so that the buffer pool
is not flooded with bufferfixed pages: see the note in
fsp0fsp.cc. */
} while (!fseg_free_step(seg_hdr + TRX_UNDO_FSEG_HEADER, false, &mtr));
hist_size = mtr_read_ulint(rseg_hdr + TRX_RSEG_HISTORY_SIZE,
MLOG_4BYTES, &mtr);
ut_ad(hist_size >= seg_size);
mlog_write_ulint(rseg_hdr + TRX_RSEG_HISTORY_SIZE,
hist_size - seg_size, MLOG_4BYTES, &mtr);
ut_ad(rseg->curr_size >= seg_size);
rseg->curr_size -= seg_size;
mutex_exit(&(rseg->mutex));
mtr_commit(&mtr);
}
/** Remove unnecessary history data from a rollback segment.
@param[in,out] rseg rollback segment
@param[in] limit truncate offset */
static
void
trx_purge_truncate_rseg_history(trx_rseg_t* rseg, const purge_iter_t* limit)
{
fil_addr_t hdr_addr;
fil_addr_t prev_hdr_addr;
trx_rsegf_t* rseg_hdr;
page_t* undo_page;
trx_ulogf_t* log_hdr;
trx_usegf_t* seg_hdr;
mtr_t mtr;
trx_id_t undo_trx_no;
mtr_start(&mtr);
ut_ad(rseg->is_persistent());
mutex_enter(&(rseg->mutex));
rseg_hdr = trx_rsegf_get(rseg->space, rseg->page_no, &mtr);
hdr_addr = trx_purge_get_log_from_hist(
flst_get_last(rseg_hdr + TRX_RSEG_HISTORY, &mtr));
loop:
if (hdr_addr.page == FIL_NULL) {
mutex_exit(&(rseg->mutex));
mtr_commit(&mtr);
return;
}
undo_page = trx_undo_page_get(page_id_t(rseg->space, hdr_addr.page),
&mtr);
log_hdr = undo_page + hdr_addr.boffset;
undo_trx_no = mach_read_from_8(log_hdr + TRX_UNDO_TRX_NO);
if (undo_trx_no >= limit->trx_no) {
/* limit space_id should match the rollback segment
space id to avoid freeing of the page belongs to
different rollback segment for the same trx_no. */
if (undo_trx_no == limit->trx_no
&& rseg->space == limit->undo_rseg_space) {
trx_undo_truncate_start(
rseg, hdr_addr.page,
hdr_addr.boffset, limit->undo_no);
}
mutex_exit(&(rseg->mutex));
mtr_commit(&mtr);
return;
}
prev_hdr_addr = trx_purge_get_log_from_hist(
flst_get_prev_addr(log_hdr + TRX_UNDO_HISTORY_NODE, &mtr));
seg_hdr = undo_page + TRX_UNDO_SEG_HDR;
if ((mach_read_from_2(seg_hdr + TRX_UNDO_STATE) == TRX_UNDO_TO_PURGE)
&& (mach_read_from_2(log_hdr + TRX_UNDO_NEXT_LOG) == 0)) {
/* We can free the whole log segment */
mutex_exit(&(rseg->mutex));
mtr_commit(&mtr);
/* calls the trx_purge_remove_log_hdr()
inside trx_purge_free_segment(). */
trx_purge_free_segment(rseg, hdr_addr);
} else {
/* Remove the log hdr from the rseg history. */
trx_purge_remove_log_hdr(rseg_hdr, log_hdr, &mtr);
mutex_exit(&(rseg->mutex));
mtr_commit(&mtr);
}
mtr_start(&mtr);
mutex_enter(&(rseg->mutex));
rseg_hdr = trx_rsegf_get(rseg->space, rseg->page_no, &mtr);
hdr_addr = prev_hdr_addr;
goto loop;
}
/** UNDO log truncate logger. Needed to track state of truncate during crash.
An auxiliary redo log file undo_<space_id>_trunc.log will created while the
truncate of the UNDO is in progress. This file is required during recovery
to complete the truncate. */
namespace undo {
/** Populate log file name based on space_id
@param[in] space_id id of the undo tablespace.
@return DB_SUCCESS or error code */
dberr_t populate_log_file_name(
ulint space_id,
char*& log_file_name)
{
ulint log_file_name_sz =
strlen(srv_log_group_home_dir) + 22 + 1 /* NUL */
+ strlen(undo::s_log_prefix)
+ strlen(undo::s_log_ext);
log_file_name = new (std::nothrow) char[log_file_name_sz];
if (log_file_name == 0) {
return(DB_OUT_OF_MEMORY);
}
memset(log_file_name, 0, log_file_name_sz);
strcpy(log_file_name, srv_log_group_home_dir);
ulint log_file_name_len = strlen(log_file_name);
if (log_file_name[log_file_name_len - 1]
!= OS_PATH_SEPARATOR) {
log_file_name[log_file_name_len]
= OS_PATH_SEPARATOR;
log_file_name_len = strlen(log_file_name);
}
snprintf(log_file_name + log_file_name_len,
log_file_name_sz - log_file_name_len,
"%s%lu_%s", undo::s_log_prefix,
(ulong) space_id, s_log_ext);
return(DB_SUCCESS);
}
/** Create the truncate log file.
@param[in] space_id id of the undo tablespace to truncate.
@return DB_SUCCESS or error code. */
dberr_t init(ulint space_id)
{
dberr_t err;
char* log_file_name;
/* Step-1: Create the log file name using the pre-decided
prefix/suffix and table id of undo tablepsace to truncate. */
err = populate_log_file_name(space_id, log_file_name);
if (err != DB_SUCCESS) {
return(err);
}
/* Step-2: Create the log file, open it and write 0 to
indicate init phase. */
bool ret;
os_file_t handle = os_file_create(
innodb_log_file_key, log_file_name, OS_FILE_CREATE,
OS_FILE_NORMAL, OS_LOG_FILE, srv_read_only_mode, &ret);
if (!ret) {
delete[] log_file_name;
return(DB_IO_ERROR);
}
ulint sz = UNIV_PAGE_SIZE;
void* buf = ut_zalloc_nokey(sz + UNIV_PAGE_SIZE);
if (buf == NULL) {
os_file_close(handle);
delete[] log_file_name;
return(DB_OUT_OF_MEMORY);
}
byte* log_buf = static_cast<byte*>(
ut_align(buf, UNIV_PAGE_SIZE));
IORequest request(IORequest::WRITE);
err = os_file_write(
request, log_file_name, handle, log_buf, 0, sz);
os_file_flush(handle);
os_file_close(handle);
ut_free(buf);
delete[] log_file_name;
return(err);
}
/** Mark completion of undo truncate action by writing magic number to
the log file and then removing it from the disk.
If we are going to remove it from disk then why write magic number ?
This is to safeguard from unlink (file-system) anomalies that will keep
the link to the file even after unlink action is successfull and
ref-count = 0.
@param[in] space_id id of the undo tablespace to truncate.*/
void done(
ulint space_id)
{
dberr_t err;
char* log_file_name;
/* Step-1: Create the log file name using the pre-decided
prefix/suffix and table id of undo tablepsace to truncate. */
err = populate_log_file_name(space_id, log_file_name);
if (err != DB_SUCCESS) {
return;
}
/* Step-2: Open log file and write magic number to
indicate done phase. */
bool ret;
os_file_t handle =
os_file_create_simple_no_error_handling(
innodb_log_file_key, log_file_name,
OS_FILE_OPEN, OS_FILE_READ_WRITE,
srv_read_only_mode, &ret);
if (!ret) {
os_file_delete(innodb_log_file_key, log_file_name);
delete[] log_file_name;
return;
}
ulint sz = UNIV_PAGE_SIZE;
void* buf = ut_zalloc_nokey(sz + UNIV_PAGE_SIZE);
if (buf == NULL) {
os_file_close(handle);
os_file_delete(innodb_log_file_key, log_file_name);
delete[] log_file_name;
return;
}
byte* log_buf = static_cast<byte*>(
ut_align(buf, UNIV_PAGE_SIZE));
mach_write_to_4(log_buf, undo::s_magic);
IORequest request(IORequest::WRITE);
err = os_file_write(
request, log_file_name, handle, log_buf, 0, sz);
ut_ad(err == DB_SUCCESS);
os_file_flush(handle);
os_file_close(handle);
ut_free(buf);
os_file_delete(innodb_log_file_key, log_file_name);
delete[] log_file_name;
}
/** Check if TRUNCATE_DDL_LOG file exist.
@param[in] space_id id of the undo tablespace.
@return true if exist else false. */
bool is_log_present(
ulint space_id)
{
dberr_t err;
char* log_file_name;
/* Step-1: Populate log file name. */
err = populate_log_file_name(space_id, log_file_name);
if (err != DB_SUCCESS) {
return(false);
}
/* Step-2: Check for existence of the file. */
bool exist;
os_file_type_t type;
os_file_status(log_file_name, &exist, &type);
/* Step-3: If file exists, check it for presence of magic
number. If found, then delete the file and report file
doesn't exist as presence of magic number suggest that
truncate action was complete. */
if (exist) {
bool ret;
os_file_t handle =
os_file_create_simple_no_error_handling(
innodb_log_file_key, log_file_name,
OS_FILE_OPEN, OS_FILE_READ_WRITE,
srv_read_only_mode, &ret);
if (!ret) {
os_file_delete(innodb_log_file_key,
log_file_name);
delete[] log_file_name;
return(false);
}
ulint sz = UNIV_PAGE_SIZE;
void* buf = ut_zalloc_nokey(sz + UNIV_PAGE_SIZE);
if (buf == NULL) {
os_file_close(handle);
os_file_delete(innodb_log_file_key,
log_file_name);
delete[] log_file_name;
return(false);
}
byte* log_buf = static_cast<byte*>(
ut_align(buf, UNIV_PAGE_SIZE));
IORequest request(IORequest::READ);
dberr_t err;
err = os_file_read(request, handle, log_buf, 0, sz);
os_file_close(handle);
if (err != DB_SUCCESS) {
ib::info()
<< "Unable to read '"
<< log_file_name << "' : "
<< ut_strerr(err);
os_file_delete(
innodb_log_file_key, log_file_name);
ut_free(buf);
delete[] log_file_name;
return(false);
}
ulint magic_no = mach_read_from_4(log_buf);
ut_free(buf);
if (magic_no == undo::s_magic) {
/* Found magic number. */
os_file_delete(innodb_log_file_key,
log_file_name);
delete[] log_file_name;
return(false);
}
}
delete[] log_file_name;
return(exist);
}
};
/** Iterate over all the UNDO tablespaces and check if any of the UNDO
tablespace qualifies for TRUNCATE (size > threshold).
@param[in,out] undo_trunc undo truncate tracker */
static
void
trx_purge_mark_undo_for_truncate(
undo::Truncate* undo_trunc)
{
/* Step-1: If UNDO Tablespace
- already marked for truncate (OR)
- truncate disabled
return immediately else search for qualifying tablespace. */
if (undo_trunc->is_marked() || !srv_undo_log_truncate) {
return;
}
/* Step-2: Validation/Qualification checks
a. At-least 2 UNDO tablespaces so even if one UNDO tablespace
is being truncated server can continue to operate.
b. At-least 2 persistent UNDO logs (besides the default rseg-0)
b. At-least 1 UNDO tablespace size > threshold. */
if (srv_undo_tablespaces_active < 2 || srv_undo_logs < 3) {
return;
}
/* Avoid bias selection and so start the scan from immediate next
of last selected UNDO tablespace for truncate. */
ulint space_id = undo_trunc->get_scan_start();
for (ulint i = 1; i <= srv_undo_tablespaces_active; i++) {
if (fil_space_get_size(space_id)
> (srv_max_undo_log_size / srv_page_size)) {
/* Tablespace qualifies for truncate. */
undo_trunc->mark(space_id);
undo::Truncate::add_space_to_trunc_list(space_id);
break;
}
space_id = ((space_id + 1) % (srv_undo_tablespaces_active + 1));
if (space_id == 0) {
/* Note: UNDO tablespace ids starts from 1. */
++space_id;
}
}
/* Couldn't make any selection. */
if (!undo_trunc->is_marked()) {
return;
}
DBUG_LOG("undo",
"marking for truncate UNDO tablespace "
<< undo_trunc->get_marked_space_id());
/* Step-3: Iterate over all the rsegs of selected UNDO tablespace
and mark them temporarily unavailable for allocation.*/
for (ulint i = 0; i < TRX_SYS_N_RSEGS; ++i) {
if (trx_rseg_t* rseg = trx_sys->rseg_array[i]) {
ut_ad(rseg->is_persistent());
if (rseg->space == undo_trunc->get_marked_space_id()) {
/* Once set this rseg will not be allocated
to new booting transaction but we will wait
for existing active transaction to finish. */
rseg->skip_allocation = true;
undo_trunc->add_rseg_to_trunc(rseg);
}
}
}
}
undo::undo_spaces_t undo::Truncate::s_spaces_to_truncate;
/** Cleanse purge queue to remove the rseg that reside in undo-tablespace
marked for truncate.
@param[in,out] undo_trunc undo truncate tracker */
static
void
trx_purge_cleanse_purge_queue(
undo::Truncate* undo_trunc)
{
mutex_enter(&purge_sys->pq_mutex);
typedef std::vector<TrxUndoRsegs> purge_elem_list_t;
purge_elem_list_t purge_elem_list;
/* Remove rseg instances that are in the purge queue before we start
truncate of corresponding UNDO truncate. */
while (!purge_sys->purge_queue.empty()) {
purge_elem_list.push_back(purge_sys->purge_queue.top());
purge_sys->purge_queue.pop();
}
ut_ad(purge_sys->purge_queue.empty());
for (purge_elem_list_t::iterator it = purge_elem_list.begin();
it != purge_elem_list.end();
++it) {
for (TrxUndoRsegs::iterator it2 = it->begin();
it2 != it->end();
++it2) {
if ((*it2)->space
== undo_trunc->get_marked_space_id()) {
it->erase(it2);
break;
}
}
if (it->size()) {
/* size != 0 suggest that there exist other rsegs that
needs processing so add this element to purge queue.
Note: Other rseg could be non-redo rsegs. */
purge_sys->purge_queue.push(*it);
}
}
mutex_exit(&purge_sys->pq_mutex);
}
/** Iterate over selected UNDO tablespace and check if all the rsegs
that resides in the tablespace are free.
@param[in] limit truncate_limit
@param[in,out] undo_trunc undo truncate tracker */
static
void
trx_purge_initiate_truncate(
purge_iter_t* limit,
undo::Truncate* undo_trunc)
{
/* Step-1: Early check to findout if any of the the UNDO tablespace
is marked for truncate. */
if (!undo_trunc->is_marked()) {
/* No tablespace marked for truncate yet. */
return;
}
/* Step-2: Scan over each rseg and ensure that it doesn't hold any
active undo records. */
bool all_free = true;
for (ulint i = 0; i < undo_trunc->rsegs_size() && all_free; ++i) {
trx_rseg_t* rseg = undo_trunc->get_ith_rseg(i);
mutex_enter(&rseg->mutex);
if (rseg->trx_ref_count > 0) {
/* This rseg is still being held by an active
transaction. */
all_free = false;
mutex_exit(&rseg->mutex);
continue;
}
ut_ad(rseg->trx_ref_count == 0);
ut_ad(rseg->skip_allocation);
ulint size_of_rsegs = rseg->curr_size;
if (size_of_rsegs == 1) {
mutex_exit(&rseg->mutex);
continue;
} else {
/* There could be cached undo segment. Check if records
in these segments can be purged. Normal purge history
will not touch these cached segment. */
ulint cached_undo_size = 0;
for (trx_undo_t* undo =
UT_LIST_GET_FIRST(rseg->update_undo_cached);
undo != NULL && all_free;
undo = UT_LIST_GET_NEXT(undo_list, undo)) {
if (limit->trx_no < undo->trx_id) {
all_free = false;
} else {
cached_undo_size += undo->size;
}
}
for (trx_undo_t* undo =
UT_LIST_GET_FIRST(rseg->insert_undo_cached);
undo != NULL && all_free;
undo = UT_LIST_GET_NEXT(undo_list, undo)) {
if (limit->trx_no < undo->trx_id) {
all_free = false;
} else {
cached_undo_size += undo->size;
}
}
ut_ad(size_of_rsegs >= (cached_undo_size + 1));
if (size_of_rsegs > (cached_undo_size + 1)) {
/* There are pages besides cached pages that
still hold active data. */
all_free = false;
}
}
mutex_exit(&rseg->mutex);
}
if (!all_free) {
/* rseg still holds active data.*/
return;
}
/* Step-3: Start the actual truncate.
a. log-checkpoint
b. Write the DDL log to protect truncate action from CRASH
c. Remove rseg instance if added to purge queue before we
initiate truncate.
d. Execute actual truncate
e. Remove the DDL log. */
/* After truncate if server crashes then redo logging done for this
undo tablespace might not stand valid as tablespace has been
truncated. */
log_make_checkpoint_at(LSN_MAX, TRUE);
const ulint space_id = undo_trunc->get_marked_space_id();
ib::info() << "Truncating UNDO tablespace " << space_id;
#ifdef UNIV_DEBUG
dberr_t err =
#endif /* UNIV_DEBUG */
undo_trunc->start_logging(space_id);
ut_ad(err == DB_SUCCESS);
DBUG_EXECUTE_IF("ib_undo_trunc_before_truncate",
ib::info() << "ib_undo_trunc_before_truncate";
DBUG_SUICIDE(););
trx_purge_cleanse_purge_queue(undo_trunc);
if (!trx_undo_truncate_tablespace(undo_trunc)) {
/* Note: In case of error we don't enable the rsegs
and neither unmark the tablespace so the tablespace
continue to remain inactive. */
ib::error() << "Failed to truncate UNDO tablespace "
<< space_id;
return;
}
if (purge_sys->rseg != NULL
&& purge_sys->rseg->last_page_no == FIL_NULL) {
/* If purge_sys->rseg is pointing to rseg that was recently
truncated then move to next rseg element.
Note: Ideally purge_sys->rseg should be NULL because purge
should complete processing of all the records but there is
purge_batch_size that can force the purge loop to exit before
all the records are purged and in this case purge_sys->rseg
could point to a valid rseg waiting for next purge cycle. */
purge_sys->next_stored = false;
purge_sys->rseg = NULL;
}
DBUG_EXECUTE_IF("ib_undo_trunc_before_ddl_log_end",
ib::info() << "ib_undo_trunc_before_ddl_log_end";
DBUG_SUICIDE(););
log_make_checkpoint_at(LSN_MAX, TRUE);
undo_trunc->done_logging(space_id);
/* Completed truncate. Now it is safe to re-use the tablespace. */
for (ulint i = 0; i < undo_trunc->rsegs_size(); ++i) {
trx_rseg_t* rseg = undo_trunc->get_ith_rseg(i);
rseg->skip_allocation = false;
}
ib::info() << "Truncated UNDO tablespace " << space_id;
undo_trunc->reset();
undo::Truncate::clear_trunc_list();
DBUG_EXECUTE_IF("ib_undo_trunc_trunc_done",
ib::info() << "ib_undo_trunc_trunc_done";
DBUG_SUICIDE(););
}
/********************************************************************//**
Removes unnecessary history data from rollback segments. NOTE that when this
function is called, the caller must not have any latches on undo log pages! */
static
void
trx_purge_truncate_history(
/*========================*/
purge_iter_t* limit, /*!< in: truncate limit */
const ReadView* view) /*!< in: purge view */
{
ut_ad(trx_purge_check_limit());
/* We play safe and set the truncate limit at most to the purge view
low_limit number, though this is not necessary */
if (limit->trx_no >= view->low_limit_no()) {
limit->trx_no = view->low_limit_no();
limit->undo_no = 0;
limit->undo_rseg_space = ULINT_UNDEFINED;
}
ut_ad(limit->trx_no <= purge_sys->view.low_limit_no());
for (ulint i = 0; i < TRX_SYS_N_RSEGS; ++i) {
trx_rseg_t* rseg = trx_sys->rseg_array[i];
if (rseg != NULL) {
ut_a(rseg->id == i);
trx_purge_truncate_rseg_history(rseg, limit);
}
}
/* UNDO tablespace truncate. We will try to truncate as much as we
can (greedy approach). This will ensure when the server is idle we
try and truncate all the UNDO tablespaces. */
for (ulint i = srv_undo_tablespaces_active; i--; ) {
trx_purge_mark_undo_for_truncate(&purge_sys->undo_trunc);
trx_purge_initiate_truncate(limit, &purge_sys->undo_trunc);
}
}
/***********************************************************************//**
Updates the last not yet purged history log info in rseg when we have purged
a whole undo log. Advances also purge_sys->purge_trx_no past the purged log. */
static
void
trx_purge_rseg_get_next_history_log(
/*================================*/
trx_rseg_t* rseg, /*!< in: rollback segment */
ulint* n_pages_handled)/*!< in/out: number of UNDO pages
handled */
{
page_t* undo_page;
trx_ulogf_t* log_hdr;
fil_addr_t prev_log_addr;
trx_id_t trx_no;
ibool del_marks;
mtr_t mtr;
mutex_enter(&(rseg->mutex));
ut_a(rseg->last_page_no != FIL_NULL);
purge_sys->iter.trx_no = rseg->last_trx_no + 1;
purge_sys->iter.undo_no = 0;
purge_sys->iter.undo_rseg_space = ULINT_UNDEFINED;
purge_sys->next_stored = false;
mtr_start(&mtr);
undo_page = trx_undo_page_get_s_latched(
page_id_t(rseg->space, rseg->last_page_no), &mtr);
log_hdr = undo_page + rseg->last_offset;
/* Increase the purge page count by one for every handled log */
(*n_pages_handled)++;
prev_log_addr = trx_purge_get_log_from_hist(
flst_get_prev_addr(log_hdr + TRX_UNDO_HISTORY_NODE, &mtr));
if (prev_log_addr.page == FIL_NULL) {
/* No logs left in the history list */
rseg->last_page_no = FIL_NULL;
mutex_exit(&(rseg->mutex));
mtr_commit(&mtr);
return;
}
mutex_exit(&rseg->mutex);
mtr_commit(&mtr);
/* Read the trx number and del marks from the previous log header */
mtr_start(&mtr);
log_hdr = trx_undo_page_get_s_latched(page_id_t(rseg->space,
prev_log_addr.page),
&mtr)
+ prev_log_addr.boffset;
trx_no = mach_read_from_8(log_hdr + TRX_UNDO_TRX_NO);
del_marks = mach_read_from_2(log_hdr + TRX_UNDO_DEL_MARKS);
mtr_commit(&mtr);
mutex_enter(&(rseg->mutex));
rseg->last_page_no = prev_log_addr.page;
rseg->last_offset = prev_log_addr.boffset;
rseg->last_trx_no = trx_no;
rseg->last_del_marks = del_marks;
TrxUndoRsegs elem(rseg->last_trx_no);
elem.push_back(rseg);
/* Purge can also produce events, however these are already ordered
in the rollback segment and any user generated event will be greater
than the events that Purge produces. ie. Purge can never produce
events from an empty rollback segment. */
mutex_enter(&purge_sys->pq_mutex);
purge_sys->purge_queue.push(elem);
mutex_exit(&purge_sys->pq_mutex);
mutex_exit(&rseg->mutex);
}
/** Position the purge sys "iterator" on the undo record to use for purging. */
static
void
trx_purge_read_undo_rec()
{
ulint offset;
ulint page_no;
ib_uint64_t undo_no;
ulint undo_rseg_space;
purge_sys->hdr_offset = purge_sys->rseg->last_offset;
page_no = purge_sys->hdr_page_no = purge_sys->rseg->last_page_no;
if (purge_sys->rseg->last_del_marks) {
mtr_t mtr;
trx_undo_rec_t* undo_rec = NULL;
mtr_start(&mtr);
undo_rec = trx_undo_get_first_rec(
purge_sys->rseg->space,
purge_sys->hdr_page_no,
purge_sys->hdr_offset, RW_S_LATCH, &mtr);
if (undo_rec != NULL) {
offset = page_offset(undo_rec);
undo_no = trx_undo_rec_get_undo_no(undo_rec);
undo_rseg_space = purge_sys->rseg->space;
page_no = page_get_page_no(page_align(undo_rec));
} else {
offset = 0;
undo_no = 0;
undo_rseg_space = ULINT_UNDEFINED;
}
mtr_commit(&mtr);
} else {
offset = 0;
undo_no = 0;
undo_rseg_space = ULINT_UNDEFINED;
}
purge_sys->offset = offset;
purge_sys->page_no = page_no;
purge_sys->iter.undo_no = undo_no;
purge_sys->iter.undo_rseg_space = undo_rseg_space;
purge_sys->next_stored = true;
}
/***********************************************************************//**
Chooses the next undo log to purge and updates the info in purge_sys. This
function is used to initialize purge_sys when the next record to purge is
not known, and also to update the purge system info on the next record when
purge has handled the whole undo log for a transaction. */
static
void
trx_purge_choose_next_log(void)
/*===========================*/
{
ut_ad(!purge_sys->next_stored);
if (purge_sys->rseg_iter.set_next()) {
trx_purge_read_undo_rec();
} else {
/* There is nothing to do yet. */
os_thread_yield();
}
}
/***********************************************************************//**
Gets the next record to purge and updates the info in the purge system.
@return copy of an undo log record or pointer to the dummy undo log record */
static
trx_undo_rec_t*
trx_purge_get_next_rec(
/*===================*/
ulint* n_pages_handled,/*!< in/out: number of UNDO pages
handled */
mem_heap_t* heap) /*!< in: memory heap where copied */
{
trx_undo_rec_t* rec;
trx_undo_rec_t* rec_copy;
trx_undo_rec_t* rec2;
page_t* undo_page;
page_t* page;
ulint offset;
ulint page_no;
ulint space;
mtr_t mtr;
ut_ad(purge_sys->next_stored);
ut_ad(purge_sys->iter.trx_no < purge_sys->view.low_limit_no());
space = purge_sys->rseg->space;
page_no = purge_sys->page_no;
offset = purge_sys->offset;
if (offset == 0) {
/* It is the dummy undo log record, which means that there is
no need to purge this undo log */
trx_purge_rseg_get_next_history_log(
purge_sys->rseg, n_pages_handled);
/* Look for the next undo log and record to purge */
trx_purge_choose_next_log();
return(&trx_purge_dummy_rec);
}
mtr_start(&mtr);
undo_page = trx_undo_page_get_s_latched(page_id_t(space, page_no),
&mtr);
rec = undo_page + offset;
rec2 = rec;
for (;;) {
ulint type;
trx_undo_rec_t* next_rec;
ulint cmpl_info;
/* Try first to find the next record which requires a purge
operation from the same page of the same undo log */
next_rec = trx_undo_page_get_next_rec(
rec2, purge_sys->hdr_page_no, purge_sys->hdr_offset);
if (next_rec == NULL) {
rec2 = trx_undo_get_next_rec(
rec2, purge_sys->hdr_page_no,
purge_sys->hdr_offset, &mtr);
break;
}
rec2 = next_rec;
type = trx_undo_rec_get_type(rec2);
if (type == TRX_UNDO_DEL_MARK_REC) {
break;
}
cmpl_info = trx_undo_rec_get_cmpl_info(rec2);
if (trx_undo_rec_get_extern_storage(rec2)) {
break;
}
if ((type == TRX_UNDO_UPD_EXIST_REC)
&& !(cmpl_info & UPD_NODE_NO_ORD_CHANGE)) {
break;
}
}
if (rec2 == NULL) {
mtr_commit(&mtr);
trx_purge_rseg_get_next_history_log(
purge_sys->rseg, n_pages_handled);
/* Look for the next undo log and record to purge */
trx_purge_choose_next_log();
mtr_start(&mtr);
undo_page = trx_undo_page_get_s_latched(
page_id_t(space, page_no), &mtr);
rec = undo_page + offset;
} else {
page = page_align(rec2);
purge_sys->offset = rec2 - page;
purge_sys->page_no = page_get_page_no(page);
purge_sys->iter.undo_no = trx_undo_rec_get_undo_no(rec2);
purge_sys->iter.undo_rseg_space = space;
if (undo_page != page) {
/* We advance to a new page of the undo log: */
(*n_pages_handled)++;
}
}
rec_copy = trx_undo_rec_copy(rec, heap);
mtr_commit(&mtr);
return(rec_copy);
}
/********************************************************************//**
Fetches the next undo log record from the history list to purge. It must be
released with the corresponding release function.
@return copy of an undo log record or pointer to trx_purge_dummy_rec,
if the whole undo log can skipped in purge; NULL if none left */
static MY_ATTRIBUTE((warn_unused_result))
trx_undo_rec_t*
trx_purge_fetch_next_rec(
/*=====================*/
roll_ptr_t* roll_ptr, /*!< out: roll pointer to undo record */
ulint* n_pages_handled,/*!< in/out: number of UNDO log pages
handled */
mem_heap_t* heap) /*!< in: memory heap where copied */
{
if (!purge_sys->next_stored) {
trx_purge_choose_next_log();
if (!purge_sys->next_stored) {
DBUG_PRINT("ib_purge",
("no logs left in the history list"));
return(NULL);
}
}
if (purge_sys->iter.trx_no >= purge_sys->view.low_limit_no()) {
return(NULL);
}
/* fprintf(stderr, "Thread %lu purging trx %llu undo record %llu\n",
os_thread_get_curr_id(), iter->trx_no, iter->undo_no); */
*roll_ptr = trx_undo_build_roll_ptr(
FALSE, purge_sys->rseg->id,
purge_sys->page_no, purge_sys->offset);
/* The following call will advance the stored values of the
purge iterator. */
return(trx_purge_get_next_rec(n_pages_handled, heap));
}
/*******************************************************************//**
This function runs a purge batch.
@return number of undo log pages handled in the batch */
static
ulint
trx_purge_attach_undo_recs(
/*=======================*/
ulint n_purge_threads,/*!< in: number of purge threads */
purge_sys_t* purge_sys, /*!< in/out: purge instance */
ulint batch_size) /*!< in: no. of pages to purge */
{
que_thr_t* thr;
ulint i = 0;
ulint n_pages_handled = 0;
ulint n_thrs = UT_LIST_GET_LEN(purge_sys->query->thrs);
ut_a(n_purge_threads > 0);
purge_sys->limit = purge_sys->iter;
/* Debug code to validate some pre-requisites and reset done flag. */
for (thr = UT_LIST_GET_FIRST(purge_sys->query->thrs);
thr != NULL && i < n_purge_threads;
thr = UT_LIST_GET_NEXT(thrs, thr), ++i) {
purge_node_t* node;
/* Get the purge node. */
node = (purge_node_t*) thr->child;
ut_a(que_node_get_type(node) == QUE_NODE_PURGE);
ut_a(node->undo_recs == NULL);
ut_a(node->done);
node->done = FALSE;
}
/* There should never be fewer nodes than threads, the inverse
however is allowed because we only use purge threads as needed. */
ut_a(i == n_purge_threads);
/* Fetch and parse the UNDO records. The UNDO records are added
to a per purge node vector. */
thr = UT_LIST_GET_FIRST(purge_sys->query->thrs);
ut_a(n_thrs > 0 && thr != NULL);
ut_ad(trx_purge_check_limit());
i = 0;
for (;;) {
purge_node_t* node;
trx_purge_rec_t* purge_rec;
ut_a(!thr->is_active);
/* Get the purge node. */
node = (purge_node_t*) thr->child;
ut_a(que_node_get_type(node) == QUE_NODE_PURGE);
purge_rec = static_cast<trx_purge_rec_t*>(
mem_heap_zalloc(node->heap, sizeof(*purge_rec)));
/* Track the max {trx_id, undo_no} for truncating the
UNDO logs once we have purged the records. */
if (trx_purge_check_limit()) {
purge_sys->limit = purge_sys->iter;
}
/* Fetch the next record, and advance the purge_sys->iter. */
purge_rec->undo_rec = trx_purge_fetch_next_rec(
&purge_rec->roll_ptr, &n_pages_handled, node->heap);
if (purge_rec->undo_rec != NULL) {
if (node->undo_recs == NULL) {
node->undo_recs = ib_vector_create(
ib_heap_allocator_create(node->heap),
sizeof(trx_purge_rec_t),
batch_size);
} else {
ut_a(!ib_vector_is_empty(node->undo_recs));
}
ib_vector_push(node->undo_recs, purge_rec);
if (n_pages_handled >= batch_size) {
break;
}
} else {
break;
}
thr = UT_LIST_GET_NEXT(thrs, thr);
if (!(++i % n_purge_threads)) {
thr = UT_LIST_GET_FIRST(purge_sys->query->thrs);
}
ut_a(thr != NULL);
}
ut_ad(trx_purge_check_limit());
return(n_pages_handled);
}
/*******************************************************************//**
Calculate the DML delay required.
@return delay in microseconds or ULINT_MAX */
static
ulint
trx_purge_dml_delay(void)
/*=====================*/
{
/* Determine how much data manipulation language (DML) statements
need to be delayed in order to reduce the lagging of the purge
thread. */
ulint delay = 0; /* in microseconds; default: no delay */
/* If purge lag is set (ie. > 0) then calculate the new DML delay.
Note: we do a dirty read of the trx_sys_t data structure here,
without holding trx_sys->mutex. */
if (srv_max_purge_lag > 0) {
float ratio;
ratio = float(trx_sys->rseg_history_len) / srv_max_purge_lag;
if (ratio > 1.0) {
/* If the history list length exceeds the
srv_max_purge_lag, the data manipulation
statements are delayed by at least 5000
microseconds. */
delay = (ulint) ((ratio - .5) * 10000);
}
if (delay > srv_max_purge_lag_delay) {
delay = srv_max_purge_lag_delay;
}
MONITOR_SET(MONITOR_DML_PURGE_DELAY, delay);
}
return(delay);
}
/*******************************************************************//**
Wait for pending purge jobs to complete. */
static
void
trx_purge_wait_for_workers_to_complete(
/*===================================*/
purge_sys_t* purge_sys) /*!< in: purge instance */
{
ulint n_submitted = purge_sys->n_submitted;
/* Ensure that the work queue empties out. */
while ((ulint) my_atomic_loadlint(&purge_sys->n_completed) != n_submitted) {
if (srv_get_task_queue_length() > 0) {
srv_release_threads(SRV_WORKER, 1);
}
os_thread_yield();
}
/* None of the worker threads should be doing any work. */
ut_a(purge_sys->n_submitted == purge_sys->n_completed);
/* There should be no outstanding tasks as long
as the worker threads are active. */
ut_a(srv_get_task_queue_length() == 0);
}
/*******************************************************************//**
This function runs a purge batch.
@return number of undo log pages handled in the batch */
ulint
trx_purge(
/*======*/
ulint n_purge_threads, /*!< in: number of purge tasks
to submit to the work queue */
ulint batch_size, /*!< in: the maximum number of records
to purge in one batch */
bool truncate) /*!< in: truncate history if true */
{
que_thr_t* thr = NULL;
ulint n_pages_handled;
ut_a(n_purge_threads > 0);
srv_dml_needed_delay = trx_purge_dml_delay();
/* The number of tasks submitted should be completed. */
ut_a(purge_sys->n_submitted == purge_sys->n_completed);
rw_lock_x_lock(&purge_sys->latch);
trx_sys->mvcc->clone_oldest_view(&purge_sys->view);
rw_lock_x_unlock(&purge_sys->latch);
#ifdef UNIV_DEBUG
if (srv_purge_view_update_only_debug) {
return(0);
}
#endif /* UNIV_DEBUG */
/* Fetch the UNDO recs that need to be purged. */
n_pages_handled = trx_purge_attach_undo_recs(
n_purge_threads, purge_sys, batch_size);
/* Do we do an asynchronous purge or not ? */
if (n_purge_threads > 1) {
ulint i = 0;
/* Submit the tasks to the work queue. */
for (i = 0; i < n_purge_threads - 1; ++i) {
thr = que_fork_scheduler_round_robin(
purge_sys->query, thr);
ut_a(thr != NULL);
srv_que_task_enqueue_low(thr);
}
thr = que_fork_scheduler_round_robin(purge_sys->query, thr);
ut_a(thr != NULL);
purge_sys->n_submitted += n_purge_threads - 1;
goto run_synchronously;
/* Do it synchronously. */
} else {
thr = que_fork_scheduler_round_robin(purge_sys->query, NULL);
ut_ad(thr);
run_synchronously:
++purge_sys->n_submitted;
que_run_threads(thr);
my_atomic_addlint(
&purge_sys->n_completed, 1);
if (n_purge_threads > 1) {
trx_purge_wait_for_workers_to_complete(purge_sys);
}
}
ut_a(purge_sys->n_submitted == purge_sys->n_completed);
#ifdef UNIV_DEBUG
rw_lock_x_lock(&purge_sys->latch);
if (purge_sys->limit.trx_no == 0) {
purge_sys->done = purge_sys->iter;
} else {
purge_sys->done = purge_sys->limit;
}
rw_lock_x_unlock(&purge_sys->latch);
#endif /* UNIV_DEBUG */
if (truncate) {
trx_purge_truncate_history(
purge_sys->limit.trx_no
? &purge_sys->limit
: &purge_sys->iter,
&purge_sys->view);
}
MONITOR_INC_VALUE(MONITOR_PURGE_INVOKED, 1);
MONITOR_INC_VALUE(MONITOR_PURGE_N_PAGE_HANDLED, n_pages_handled);
return(n_pages_handled);
}
/*******************************************************************//**
Get the purge state.
@return purge state. */
purge_state_t
trx_purge_state(void)
/*=================*/
{
purge_state_t state;
rw_lock_x_lock(&purge_sys->latch);
state = purge_sys->state;
rw_lock_x_unlock(&purge_sys->latch);
return(state);
}
/*******************************************************************//**
Stop purge and wait for it to stop, move to PURGE_STATE_STOP. */
void
trx_purge_stop(void)
/*================*/
{
rw_lock_x_lock(&purge_sys->latch);
switch (purge_sys->state) {
case PURGE_STATE_INIT:
case PURGE_STATE_DISABLED:
ut_error;
case PURGE_STATE_EXIT:
/* Shutdown must have been initiated during
FLUSH TABLES FOR EXPORT. */
ut_ad(!srv_undo_sources);
unlock:
rw_lock_x_unlock(&purge_sys->latch);
break;
case PURGE_STATE_STOP:
ut_ad(srv_n_purge_threads > 0);
++purge_sys->n_stop;
purge_sys->state = PURGE_STATE_STOP;
if (!purge_sys->running) {
goto unlock;
}
ib::info() << "Waiting for purge to stop";
do {
rw_lock_x_unlock(&purge_sys->latch);
os_thread_sleep(10000);
rw_lock_x_lock(&purge_sys->latch);
} while (purge_sys->running);
goto unlock;
case PURGE_STATE_RUN:
ut_ad(srv_n_purge_threads > 0);
++purge_sys->n_stop;
ib::info() << "Stopping purge";
/* We need to wakeup the purge thread in case it is suspended,
so that it can acknowledge the state change. */
const int64_t sig_count = os_event_reset(purge_sys->event);
purge_sys->state = PURGE_STATE_STOP;
srv_purge_wakeup();
rw_lock_x_unlock(&purge_sys->latch);
/* Wait for purge coordinator to signal that it
is suspended. */
os_event_wait_low(purge_sys->event, sig_count);
}
MONITOR_INC_VALUE(MONITOR_PURGE_STOP_COUNT, 1);
}
/*******************************************************************//**
Resume purge, move to PURGE_STATE_RUN. */
void
trx_purge_run(void)
/*===============*/
{
rw_lock_x_lock(&purge_sys->latch);
switch (purge_sys->state) {
case PURGE_STATE_EXIT:
/* Shutdown must have been initiated during
FLUSH TABLES FOR EXPORT. */
ut_ad(!srv_undo_sources);
break;
case PURGE_STATE_INIT:
case PURGE_STATE_DISABLED:
ut_error;
case PURGE_STATE_RUN:
ut_a(!purge_sys->n_stop);
break;
case PURGE_STATE_STOP:
ut_a(purge_sys->n_stop);
if (--purge_sys->n_stop == 0) {
ib::info() << "Resuming purge";
purge_sys->state = PURGE_STATE_RUN;
}
MONITOR_INC_VALUE(MONITOR_PURGE_RESUME_COUNT, 1);
}
rw_lock_x_unlock(&purge_sys->latch);
srv_purge_wakeup();
}
|