forked from bloomberg/comdb2
-
Notifications
You must be signed in to change notification settings - Fork 0
/
sc_alter_table.c
1301 lines (1088 loc) · 42.9 KB
/
sc_alter_table.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
/*
Copyright 2015 Bloomberg Finance L.P.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
#include <unistd.h>
#include <crc32c.h>
#include <memory_sync.h>
#include "schemachange.h"
#include "sc_alter_table.h"
#include "logmsg.h"
#include "sc_global.h"
#include "sc_schema.h"
#include "sc_struct.h"
#include "sc_csc2.h"
#include "sc_util.h"
#include "sc_logic.h"
#include "sc_records.h"
#include "analyze.h"
#include "comdb2_atomic.h"
#include "views.h"
#include "macc_glue.h"
static int prepare_sc_plan(struct schema_change_type *s, int old_changed,
struct dbtable *db, struct dbtable *newdb,
struct scplan *theplan)
{
int changed = old_changed;
/* don't create schema change plan if we're not going to use it */
if (s->force_rebuild) {
s->use_plan = gbl_use_plan = 0;
} else {
int rc = create_schema_change_plan(s, db, newdb, theplan);
if (rc != 0) {
sc_printf(s, "not using plan. error in plan module.\n");
changed = SC_TAG_CHANGE;
} else if (!s->use_plan || s->force_rebuild || s->use_old_blobs_on_rebuild) {
sc_printf(s, "not using plan. full rebuild\n");
s->use_plan = gbl_use_plan = 0;
changed = SC_TAG_CHANGE;
} else {
sc_printf(s, "Using plan.\n");
newdb->plan = theplan;
if (newdb->plan->dta_plan) changed = SC_TAG_CHANGE;
}
s->retry_bad_genids = 0;
}
return changed;
}
static int prepare_changes(struct schema_change_type *s, struct dbtable *db,
struct dbtable *newdb, struct scplan *theplan,
struct scinfo *scinfo)
{
int changed = ondisk_schema_changed(s->tablename, newdb, stderr, s);
if (DBPAD_OR_DBSTORE_ERR(changed)) {
/* we want to capture cases when "alter" is used
to add a field to a table that has no dbstore or
isnull specified,
or change the size of a byte array with no dbpad specified
It is still possible to do so by using "fastinit"
with the new schema instead of "alter"
*/
if (IS_FASTINIT(s)) {
changed = SC_TAG_CHANGE;
}
}
if (changed < 0) {
/* some errors during constraint verifications */
backout_schemas(newdb->tablename);
resume_threads(thedb); /* can now restart stopped threads */
/* these checks should be present in dryrun_int as well */
if (changed == SC_BAD_NEW_FIELD) {
sc_client_error(s, "cannot add new field without dbstore or null");
} else if (changed == SC_BAD_INDEX_CHANGE) {
sc_client_error(s, "cannot change index referenced by other tables");
} else if (changed == SC_BAD_DBPAD) {
sc_client_error(s, "cannot change size of byte array without dbpad");
} else if (changed == SC_BAD_DBSTORE_FUNC_NOT_NULL) {
sc_client_error(s, "column must be nullable to use a function as its default value");
}
sc_errf(s, "Failed to process schema!\n");
return -1;
}
/* force change if we need to add/remove headers (or anything else which
* rebuilds) */
if (db->odh != s->headers) {
changed = SC_TAG_CHANGE;
}
/* bump the version for inplace-updates. */
if (!scinfo->olddb_inplace_updates && s->ip_updates) {
changed = SC_TAG_CHANGE;
}
if (s->compress != scinfo->olddb_compress) {
s->force_dta_rebuild = 1;
}
if (s->compress_blobs != scinfo->olddb_compress_blobs) {
s->force_blob_rebuild = 1;
}
/* force a change if instant schema-change is being enabled. reset version
* to 0. */
if (!scinfo->olddb_instant_sc && s->instant_sc) {
changed = SC_TAG_CHANGE;
}
/* force a change if requested. reset version to 0. */
if (s->force_rebuild) {
changed = SC_TAG_CHANGE;
}
/* rebuild if ran out of version numbers */
if (newdb->schema_version >= MAXVER && newdb->instant_schema_change) {
sc_printf(s, "exhausted versions. forcing rebuild at %d\n",
newdb->schema_version);
if (s->dryrun)
sbuf2printf(s->sb, ">table version: %d. will have to rebuild.\n",
newdb->schema_version);
s->force_rebuild = 1;
changed = SC_TAG_CHANGE;
}
return changed;
}
static void adjust_version(int changed, struct scinfo *scinfo,
struct schema_change_type *s, struct dbtable *db,
struct dbtable *newdb)
{
/* if we don't want to merely bump the version, reset it to 0. */
if (changed == SC_TAG_CHANGE && newdb->instant_schema_change) {
newdb->schema_version = 0;
}
/* if we're disabling instant sc, reset the version to 0. */
else if (db->instant_schema_change && !newdb->instant_schema_change) {
/* instant schema change is being removed */
newdb->schema_version = 0;
}
/* if we are enabling instant sc, set the version to 1. */
else if (db->odh && !db->instant_schema_change &&
newdb->instant_schema_change && !IS_FASTINIT(s)) {
/* old db had odh but not instant schema change.
* the physical records will have version 0 &
* will correspond to the latest csc2 version. start
* instant version at 2 and treat physical version 0
* as version 1 */
newdb->schema_version = 1;
}
/* if only index or constraints have changed don't bump version */
int ondisk_changed = compare_tag(s->tablename, ".ondisk", NULL);
if (ondisk_changed != SC_NO_CHANGE /* something changed in .ondisk */
|| IS_FASTINIT(s) /* fastinit */
|| !s->use_plan /* full rebuild due to some prob */
|| s->force_rebuild /* full rebuild requested */
|| (!scinfo->olddb_instant_sc &&
s->instant_sc) /* bump version if enabled instant sc */
) {
++newdb->schema_version;
} else if (ondisk_changed == SC_NO_CHANGE /* nothing changed ondisk */
&& changed == SC_TAG_CHANGE /* plan says it did change */
&& newdb->schema_version == 0 && newdb->plan &&
newdb->plan->dta_plan == -1 && newdb->plan->plan_convert == 1) {
++newdb->schema_version;
}
}
static int prepare_version_for_dbs_without_instant_sc(tran_type *tran,
struct dbtable *db,
struct dbtable *newdb)
{
int rc;
if (db->odh && !db->instant_schema_change && newdb->instant_schema_change &&
newdb->schema_version == 2) {
/* Old db had ODH but not instant schema change.
* The physical records will have version 0 &
* will correspond to the latest csc2 version. Start
* instant version at 2 and treat physical version 0
* as version 1 */
struct schema *ver_one;
if ((rc = prepare_table_version_one(tran, db, &ver_one))) return rc;
replace_tag_schema(newdb, ver_one);
}
return SC_OK;
}
static int switch_versions_with_plan(void *tran, struct dbtable *db,
struct dbtable *newdb)
{
int rc, bdberr;
int blobno, ixnum;
unsigned long long file_versions[MAXINDEX];
if (newdb->plan->dta_plan == -1) {
/*set main data metapointer to new data file*/
rc = bdb_commit_temp_file_version_data(newdb->handle, tran,
0 /*dtanum*/, &bdberr);
if (rc) return SC_BDB_ERROR;
}
/* get all out old file versions */
bzero(file_versions, sizeof(file_versions));
for (blobno = 0; blobno < MAXBLOBS + 1; blobno++) {
bdb_get_file_version_data(db->handle, tran, blobno,
file_versions + blobno, &bdberr);
}
for (blobno = 0; blobno < newdb->numblobs; blobno++) {
if (newdb->plan->blob_plan[blobno] == blobno) {
/* do nothing with this blob */
} else if (newdb->plan->blob_plan[blobno] >= 0) {
logmsg(LOGMSG_INFO, "bdb_file_version_change_dtanum:"
" %d -> %d : data %d now points to %016llx\n",
newdb->plan->blob_plan[blobno] + 1, blobno + 1, blobno + 1,
file_versions[newdb->plan->blob_plan[blobno] + 1]);
rc = bdb_new_file_version_data(
db->handle, tran, blobno + 1,
file_versions[newdb->plan->blob_plan[blobno] + 1], &bdberr);
/* we're reusing the old blob */
if (rc) return SC_BDB_ERROR;
} else if (newdb->plan->blob_plan[blobno] == -1) {
/* we made a new blob */
rc = bdb_commit_temp_file_version_data(newdb->handle, tran,
blobno + 1, &bdberr);
if (rc) return SC_BDB_ERROR;
}
}
/* get all out old file versions */
bzero(file_versions, sizeof(file_versions));
for (ixnum = 0; ixnum < MAXINDEX; ixnum++) {
rc = bdb_get_file_version_index(db->handle, tran, ixnum,
file_versions + ixnum, &bdberr);
}
for (ixnum = 0; ixnum < newdb->nix; ixnum++) {
if (newdb->plan->ix_plan[ixnum] == ixnum) {
logmsg(LOGMSG_INFO, "ix %d being left alone\n", ixnum);
/* do nothing with this index */
} else if (newdb->plan->ix_plan[ixnum] >= 0) {
logmsg(LOGMSG_INFO, "bdb_file_version_change_ixnum:"
" %d -> %d : ix %d now points to %016llx\n",
newdb->plan->ix_plan[ixnum], ixnum, ixnum,
file_versions[newdb->plan->ix_plan[ixnum]]);
rc = bdb_new_file_version_index(
db->handle, tran, ixnum,
file_versions[newdb->plan->ix_plan[ixnum]], &bdberr);
/* we're re-using the old index */
if (rc) return SC_BDB_ERROR;
} else if (newdb->plan->ix_plan[ixnum] == -1) {
/* we rebuilt this index */
rc = bdb_commit_temp_file_version_index(newdb->handle, tran, ixnum,
&bdberr);
if (rc) return SC_BDB_ERROR;
}
}
return SC_OK;
}
static void backout(struct dbtable *db)
{
backout_schemas(db->tablename);
live_sc_off(db);
}
static inline int wait_to_resume(struct schema_change_type *s)
{
int rc = 0;
if (s->resume) {
int stm = BDB_ATTR_GET(thedb->bdb_attr, SC_RESTART_SEC);
if (stm <= 0)
return 0;
logmsg(LOGMSG_WARN, "%s: Schema change will resume in %d seconds\n",
__func__, stm);
while (stm) {
sleep(1);
stm--;
/* give a chance for sc to stop */
if (get_stopsc(__func__, __LINE__)) {
sc_errf(s, "master downgrading\n");
return SC_MASTER_DOWNGRADE;
}
}
logmsg(LOGMSG_WARN, "%s: Schema change resuming.\n", __func__);
}
return rc;
}
int gbl_test_scindex_deadlock = 0;
int gbl_test_sc_resume_race = 0;
int gbl_readonly_sc = 0;
/*********** Outer Business logic for schemachanges ************************/
static void check_for_idx_rename(struct dbtable *newdb, struct dbtable *olddb)
{
if (!newdb || !newdb->plan)
return;
for (int ixnum = 0; ixnum < newdb->nix; ixnum++) {
struct schema *newixs = newdb->ixschema[ixnum];
int oldixnum = newdb->plan->ix_plan[ixnum];
if (oldixnum < 0 || oldixnum >= olddb->nix)
continue;
struct schema *oldixs = olddb->ixschema[oldixnum];
if (!oldixs)
continue;
int offset = get_offset_of_keyname(newixs->csctag);
if (get_offset_of_keyname(oldixs->csctag) > 0) {
logmsg(LOGMSG_USER, "WARN: Oldix has .NEW. in idx name: %s\n",
oldixs->csctag);
return;
}
if (newdb->plan->ix_plan[ixnum] >= 0 &&
strcmp(newixs->csctag + offset, oldixs->csctag) != 0) {
char namebuf1[128];
char namebuf2[128];
form_new_style_name(namebuf1, sizeof(namebuf1), newixs,
newixs->csctag + offset, newdb->tablename);
form_new_style_name(namebuf2, sizeof(namebuf2), oldixs,
oldixs->csctag, olddb->tablename);
logmsg(LOGMSG_INFO,
"ix %d changing name so INSERTING into sqlite_stat* "
"idx='%s' where tbl='%s' and idx='%s' \n",
ixnum, newixs->csctag + offset, newdb->tablename,
oldixs->csctag);
add_idx_stats(newdb->tablename, namebuf2, namebuf1);
}
}
}
static int do_merge_table(struct ireq *iq, struct schema_change_type *s,
tran_type *tran);
static int optionsChanged(struct schema_change_type *sc, struct scinfo *scinfo){
if(sc->headers != scinfo->olddb_odh ||
sc->ip_updates != scinfo->olddb_inplace_updates ||
sc->instant_sc != scinfo->olddb_instant_sc ||
sc->compress_blobs != scinfo->olddb_compress_blobs ||
sc->compress != scinfo->olddb_compress) {
return 1;
}
return 0;
}
int do_alter_table(struct ireq *iq, struct schema_change_type *s,
tran_type *tran)
{
struct dbtable *db;
int rc;
int bdberr = 0;
struct dbtable *newdb;
int datacopy_odh = 0;
int changed;
int i;
char new_prefix[32];
struct scinfo scinfo;
struct errstat err = {0};
if (s->partition.type == PARTITION_MERGE)
return do_merge_table(iq, s, tran);
#ifdef DEBUG_SC
logmsg(LOGMSG_INFO, "do_alter_table() %s\n", s->resume ? "resuming" : "");
#endif
gbl_use_plan = 1;
gbl_sc_last_writer_time = 0;
db = get_dbtable_by_name(s->tablename);
if (db == NULL) {
sc_errf(s, "Table not found:%s\n", s->tablename);
return SC_TABLE_DOESNOT_EXIST;
}
if (s->resume == SC_PREEMPT_RESUME) {
newdb = db->sc_to;
changed = s->schema_change;
goto convert_records;
}
set_schemachange_options_tran(s, db, &scinfo, tran);
if (optionsChanged(s,&scinfo)) {
sc_printf(s, "table option(s) have changed. Forcing a rebuild\n");
sc_printf(s," new options -> \n");
sc_printf(s," headers: %d\n", s->headers);
sc_printf(s," ip_updates: %d\n", s->ip_updates);
sc_printf(s," instant_sc: %d\n", s->instant_sc);
sc_printf(s," compress: %d\n", s->compress);
sc_printf(s," compress_blobs: %d\n", s->compress_blobs);
sc_printf(s," --------------------------------------------------\n");
sc_printf(s," old options -> \n");
sc_printf(s," headers: %d\n", scinfo.olddb_odh);
sc_printf(s," ip_updates: %d\n", scinfo.olddb_inplace_updates);
sc_printf(s," instant_sc: %d\n", scinfo.olddb_instant_sc);
sc_printf(s," compress: %d\n", scinfo.olddb_compress);
sc_printf(s," compress_blobs: %d\n", scinfo.olddb_compress_blobs);
s->force_rebuild = 1;
}
if ((rc = check_option_coherency(s, db, &scinfo))) return rc;
sc_printf(s, "starting schema update with seed %0#16llx\n",
flibc_ntohll(iq->sc_seed));
int local_lock = 0;
if (!iq->sc_locked) {
wrlock_schema_lk();
local_lock = 1;
}
Pthread_mutex_lock(&csc2_subsystem_mtx);
newdb = create_new_dbtable(thedb, s->tablename, s->newcsc2, db->dbnum,
1 /* sc_alt_name */,
(s->same_schema) ? 1 : 0, 0, &err);
if (!newdb) {
sc_client_error(s, "%s", err.errstr);
Pthread_mutex_unlock(&csc2_subsystem_mtx);
if (local_lock)
unlock_schema_lk();
return SC_INTERNAL_ERROR;
}
newdb->dtastripe = gbl_dtastripe; // we have only one setting currently
newdb->odh = s->headers;
/* don't lose precious flags like this */
newdb->instant_schema_change = s->headers && s->instant_sc;
newdb->inplace_updates = s->headers && s->ip_updates;
newdb->iq = iq;
newdb->schema_version = get_csc2_version(newdb->tablename);
if ((rc = sql_syntax_check(iq, newdb))) {
Pthread_mutex_unlock(&csc2_subsystem_mtx);
if (local_lock)
unlock_schema_lk();
backout(newdb);
cleanup_newdb(newdb);
Pthread_mutex_unlock(&csc2_subsystem_mtx);
sc_errf(s, "Sqlite syntax check failed\n");
return SC_CSC2_ERROR;
} else {
sc_printf(s, "Sqlite syntax check succeeded\n");
}
newdb->ix_blob = newdb->schema->ix_blob;
Pthread_mutex_unlock(&csc2_subsystem_mtx);
if ((iq == NULL || iq->tranddl <= 1) &&
verify_constraints_exist(NULL, newdb, newdb, s) != 0) {
if (local_lock)
unlock_schema_lk();
backout(newdb);
cleanup_newdb(newdb);
sc_errf(s, "Failed to process schema!\n");
return -1;
}
s->schema_change = changed =
prepare_changes(s, db, newdb, &s->plan, &scinfo);
if (changed == SC_UNKNOWN_ERROR) {
backout(newdb);
if (local_lock)
unlock_schema_lk();
cleanup_newdb(newdb);
sc_errf(s, "Internal error");
return SC_INTERNAL_ERROR;
}
adjust_version(changed, &scinfo, s, db, newdb);
s->schema_change = changed =
prepare_sc_plan(s, changed, db, newdb, &s->plan);
print_schemachange_info(s, db, newdb);
/*************** open tables ********************************************/
/* create temporary tables. to try to avoid strange issues always
* use a unqiue prefix. this avoids multiple histories for these
* new. files in our logs.
*
* since the prefix doesn't matter and bdb needs to be able to unappend
* it, we let bdb choose the prefix */
/* ignore failures, there shouln't be any and we'd just have a
* truncated prefix anyway */
bdb_get_new_prefix(new_prefix, sizeof(new_prefix), &bdberr);
rc = open_temp_db_resume(iq, newdb, new_prefix, s->resume, 0, tran);
if (rc) {
/* todo: clean up db */
sc_errf(s, "failed opening new db\n");
change_schemas_recover(s->tablename);
if (local_lock)
unlock_schema_lk();
return -1;
}
if (verify_new_temp_sc_db(db, newdb, tran)) {
sc_errf(s, "some of the newdb's file versions are the same or less "
"than some of db's.\n"
"we will delete the newdb's file version entries from "
"llmeta and quit.\n"
"the new master should be able to resume this sc starting "
"over with a fresh newdb\n");
if ((rc = bdb_del_file_versions(newdb->handle, tran, &bdberr)) ||
bdberr != BDBERR_NOERROR) {
sc_errf(s, "failed to clear newdb's file versions, hopefully "
"the new master can sort it out\n");
}
if (local_lock)
unlock_schema_lk();
clean_exit();
}
if (local_lock)
unlock_schema_lk();
/* we can resume sql threads at this point */
/* must do this before rebuilding, otherwise we'll have the wrong
* blobstripe_genid. */
transfer_db_settings(db, newdb);
get_db_datacopy_odh_tran(db, &datacopy_odh, tran);
if (s->force_rebuild || /* we're first to set */
newdb->instant_schema_change) /* we're doing instant sc*/
{
datacopy_odh = 1;
}
/* we set compression /odh options in bdb only here.
for full operation they also need to be set in the meta tables.
however the new db gets its meta table assigned further down,
so we can't set meta options until we're there. */
set_bdb_option_flags(newdb, s->headers, s->ip_updates,
newdb->instant_schema_change, newdb->schema_version,
s->compress, s->compress_blobs, datacopy_odh);
/* set sc_genids, 0 them if we are starting a new schema change, or
* restore them to their previous values if we are resuming */
if (init_sc_genids(newdb, s)) {
sc_errf(s, "failed initilizing sc_genids\n");
delete_temp_table(iq, newdb);
change_schemas_recover(s->tablename);
return -1;
}
Pthread_rwlock_wrlock(&db->sc_live_lk);
db->sc_from = s->db = db;
db->sc_to = s->newdb = newdb;
db->sc_abort = 0;
db->sc_downgrading = 0;
db->doing_conversion = 1; /* live_sc_off will unset it */
Pthread_rwlock_unlock(&db->sc_live_lk);
convert_records:
assert(db->sc_from == db && s->db == db);
assert(db->sc_to == newdb && s->newdb == newdb);
assert(db->doing_conversion == 1);
if (s->resume && IS_ALTERTABLE(s)) {
if (gbl_test_sc_resume_race && !get_stopsc(__func__, __LINE__)) {
logmsg(LOGMSG_INFO, "%s:%d sleeping 5s for sc_resume test\n",
__func__, __LINE__);
sleep(5);
}
if (gbl_sc_resume_start > 0)
ATOMIC_ADD32(gbl_sc_resume_start, -1);
}
MEMORY_SYNC;
/* give a chance for sc to stop */
if (get_stopsc(__func__, __LINE__)) {
sc_errf(s, "master downgrading\n");
change_schemas_recover(s->tablename);
return SC_MASTER_DOWNGRADE;
}
reset_sc_stat();
rc = wait_to_resume(s);
if (rc || get_stopsc(__func__, __LINE__)) {
sc_errf(s, "master downgrading\n");
return SC_MASTER_DOWNGRADE;
}
int prev_preempted = s->preempted;
if (s->preempted == SC_ACTION_PAUSE) {
sc_client_error(s, "SCHEMACHANGE PAUSED");
add_ongoing_alter(s);
return SC_PAUSED;
} else if (s->preempted == SC_ACTION_ABORT) {
sc_client_error(s, "SCHEMACHANGE ABORTED");
rc = SC_ABORTED;
goto errout;
}
add_ongoing_alter(s);
/* skip converting records for fastinit and planned schema change
* that doesn't require rebuilding anything. */
if ((!newdb->plan || newdb->plan->plan_convert) ||
changed == SC_CONSTRAINT_CHANGE) {
if (!s->live)
gbl_readonly_sc = 1;
rc = convert_all_records(db, newdb, newdb->sc_genids, s);
if (rc == 1) rc = 0;
} else
rc = 0;
remove_ongoing_alter(s);
if (s->preempted != prev_preempted || rc == SC_PREEMPTED) {
sc_client_error(s, "SCHEMACHANGE PREEMPTED");
return SC_PREEMPTED;
} else if (get_stopsc(__func__, __LINE__) || rc == SC_MASTER_DOWNGRADE)
rc = SC_MASTER_DOWNGRADE;
else if (rc)
rc = SC_CONVERSION_FAILED;
if (gbl_test_scindex_deadlock) {
logmsg(LOGMSG_INFO, "%s: sleeping for 30s\n", __func__);
sleep(30);
logmsg(LOGMSG_INFO, "%s: slept 30s\n", __func__);
}
if (s->convert_sleep > 0) {
sc_printf(s, "[%s] Sleeping after conversion for %d...\n",
db->tablename, s->convert_sleep);
logmsg(LOGMSG_INFO, "Sleeping after conversion for %d...\n",
s->convert_sleep);
sleep(s->convert_sleep);
sc_printf(s, "[%s] ...slept for %d\n", db->tablename, s->convert_sleep);
}
errout:
if (rc && rc != SC_MASTER_DOWNGRADE) {
/* For live schema change, MUST do this before trying to remove
* the .new tables or you get crashes */
if (gbl_sc_abort || db->sc_abort || iq->sc_should_abort ||
rc == SC_ABORTED) {
sc_errf(s, "convert_all_records aborted\n");
} else {
sc_errf(s, "convert_all_records failed\n");
}
live_sc_off(db);
for (i = 0; i < gbl_dtastripe; i++) {
sc_errf(s, " > [%s] stripe %2d was at 0x%016llx\n", s->tablename,
i, newdb->sc_genids[i]);
}
while (s->logical_livesc) {
usleep(200);
}
if (db->sc_live_logical) {
bdb_clear_logical_live_sc(db->handle, 1 /* lock table */);
db->sc_live_logical = 0;
}
backout_constraint_pointers(newdb, db);
delete_temp_table(iq, newdb);
change_schemas_recover(s->tablename);
return rc;
}
newdb->iq = NULL;
/* check for rename outside of taking schema lock */
/* handle renaming sqlite_stat1 entries for idx */
check_for_idx_rename(s->newdb, s->db);
return SC_OK;
}
static int do_merge_table(struct ireq *iq, struct schema_change_type *s,
tran_type *tran)
{
struct dbtable *db;
struct dbtable *newdb;
int i;
int rc;
struct scinfo scinfo;
#ifdef DEBUG_SC
logmsg(LOGMSG_INFO, "do_alter_table() %s\n", s->resume ? "resuming" : "");
#endif
gbl_sc_last_writer_time = 0;
db = get_dbtable_by_name(s->tablename);
if (db == NULL) {
sc_errf(s, "Table not found:%s\n", s->tablename);
return SC_TABLE_DOESNOT_EXIST;
}
newdb = s->newdb;
/* NOTE: add prepopulates newdb->csc2_schema, but alter does not
* we need this to be able to call populate_db_with_alt_schema
* later on
*/
if (!newdb->csc2_schema) {
assert(bdb_have_llmeta() && s->kind == SC_ALTERTABLE);
int ver;
ver = get_csc2_version_tran(db->tablename, tran);
if (ver > 0) {
get_csc2_file_tran(db->tablename, ver, &newdb->csc2_schema,
&newdb->csc2_schema_len, tran);
} else {
sc_client_error(s, "Cannot get csc2 for the table %s",
db->tablename);
return -1;
}
}
if (s->resume == SC_PREEMPT_RESUME) {
newdb = db->sc_to;
goto convert_records;
}
set_schemachange_options_tran(s, db, &scinfo, tran);
if ((rc = check_option_coherency(s, db, &scinfo))) return rc;
sc_printf(s, "starting table merge with seed %0#16llx\n",
flibc_ntohll(iq->sc_seed));
if ((iq == NULL || iq->tranddl <= 1) && db->n_rev_constraints > 0 &&
!self_referenced_only(db)) {
sc_client_error(s, "Cannot drop a table referenced by a foreign key");
return -1;
}
print_schemachange_info(s, db, newdb);
/* ban old settings */
if (db->dbnum) {
sc_client_error(s, "Cannot comdbg tables");
return -1;
}
/* set sc_genids, 0 them if we are starting a new schema change, or
* restore them to their previous values if we are resuming */
if (init_sc_genids(newdb, s)) {
sc_client_error(s, "Failed to initialize sc_genids");
return -1;
}
Pthread_rwlock_wrlock(&db->sc_live_lk);
db->sc_from = s->db = db;
db->sc_to = s->newdb = newdb;
db->sc_abort = 0;
db->sc_downgrading = 0;
db->doing_conversion = 1; /* live_sc_off will unset it */
Pthread_rwlock_unlock(&db->sc_live_lk);
convert_records:
assert(db->sc_from == db && s->db == db);
assert(db->sc_to == newdb && s->newdb == newdb);
assert(db->doing_conversion == 1);
MEMORY_SYNC;
if (get_stopsc(__func__, __LINE__)) {
sc_errf(s, "master downgrading\n");
return SC_MASTER_DOWNGRADE;
}
rc = wait_to_resume(s);
if (rc || get_stopsc(__func__, __LINE__)) {
sc_errf(s, "master downgrading\n");
return SC_MASTER_DOWNGRADE;
}
/* Merging two scenarios:
*- alter merge: in this case the schema is already populated with NEW tags
*- create merge: we need to populate schema with NEW tags
*/
struct schema *tag = find_tag_schema(newdb, ".NEW..ONDISK");
if (!tag) {
struct errstat err = {0};
rc =
populate_db_with_alt_schema(thedb, newdb, newdb->csc2_schema, &err);
if (rc) {
logmsg(LOGMSG_ERROR, "%s\ncsc2: \"%s\"\n", err.errstr,
newdb->csc2_schema);
sc_client_error(s, "%s", err.errval, err.errstr);
return -1;
}
}
add_ongoing_alter(s);
/* skip converting records for fastinit and planned schema change
* that doesn't require rebuilding anything. */
rc = convert_all_records(db, newdb, newdb->sc_genids, s);
if (rc == 1) rc = 0;
remove_ongoing_alter(s);
if (!tag)
backout_schemas(newdb->tablename);
if (rc == SC_PREEMPTED) {
sc_client_error(s, "SCHEMACHANGE PREEMPTED");
return SC_PREEMPTED;
} else if (get_stopsc(__func__, __LINE__) || rc == SC_MASTER_DOWNGRADE)
rc = SC_MASTER_DOWNGRADE;
else if (rc)
rc = SC_CONVERSION_FAILED;
if (s->convert_sleep > 0) {
sc_printf(s, "[%s] Sleeping after conversion for %d...\n",
db->tablename, s->convert_sleep);
logmsg(LOGMSG_INFO, "Sleeping after conversion for %d...\n",
s->convert_sleep);
sleep(s->convert_sleep);
sc_printf(s, "[%s] ...slept for %d\n", db->tablename, s->convert_sleep);
}
if (rc && rc != SC_MASTER_DOWNGRADE) {
/* For live schema change, MUST do this before trying to remove
* the .new tables or you get crashes */
if (gbl_sc_abort || db->sc_abort || iq->sc_should_abort ||
rc == SC_ABORTED) {
sc_errf(s, "convert_all_records aborted\n");
} else {
sc_errf(s, "convert_all_records failed\n");
}
live_sc_off(db);
for (i = 0; i < gbl_dtastripe; i++) {
sc_errf(s, " > [%s] stripe %2d was at 0x%016llx\n", s->tablename,
i, newdb->sc_genids[i]);
}
while (s->logical_livesc) {
usleep(200);
}
if (db->sc_live_logical) {
bdb_clear_logical_live_sc(db->handle, 1 /* lock table */);
db->sc_live_logical = 0;
}
return rc;
}
newdb->iq = NULL;
/* check for rename outside of taking schema lock */
/* handle renaming sqlite_stat1 entries for idx */
check_for_idx_rename(s->newdb, s->db);
return SC_OK;
}
static int finalize_merge_table(struct ireq *iq, struct schema_change_type *s,
tran_type *transac);
#define BACKOUT \
do { \
sc_errf(s, "%s:%d backing out\n", __func__, __LINE__); \
goto backout; \
} while (0);
int finalize_alter_table(struct ireq *iq, struct schema_change_type *s,
tran_type *transac)
{
int rc, bdberr;
struct dbtable *db = s->db;
struct dbtable *newdb = s->newdb;
void *old_bdb_handle, *new_bdb_handle;
int olddb_bthashsz;
if (s->partition.type == PARTITION_MERGE)
return finalize_merge_table(iq, s, transac);
iq->usedb = db;
new_bdb_handle = newdb->handle;
old_bdb_handle = db->handle;
if ((rc = bdb_lock_tablename_write(db->handle, "comdb2_tables", transac)) != 0) {
sc_errf(s, "Error getting comdb2_tables lock: %d\n", rc);
BACKOUT;
}
if ((rc = bdb_lock_table_write(db->handle, transac)) != 0) {
sc_errf(s, "Error getting tablelock: %d\n", rc);
BACKOUT;
}
if (iq && iq->tranddl > 1 &&
verify_constraints_exist(NULL, newdb, newdb, s) != 0) {
sc_errf(s, "error verifying constraints\n");
BACKOUT;
}
if (get_db_bthash_tran(db, &olddb_bthashsz, transac) != 0)
olddb_bthashsz = 0;
sc_printf(s, "[%s] Got table write lock OK\n", s->tablename);
s->got_tablelock = 1;
/* wait for logical redo thread to stop */
while (s->logical_livesc) {
usleep(200);
}
db->sc_to = newdb;
if (db->sc_live_logical)
bdb_clear_logical_live_sc(db->handle, 0 /* already locked */);
if (gbl_sc_abort || db->sc_abort || iq->sc_should_abort) {
sc_errf(s, "Aborting schema change %s\n", s->tablename);
sc_errf(s, "gbl_sc_abort=%d db->sc_abort=%d iq->sc_should_abort=%d\n",
gbl_sc_abort, db->sc_abort, iq->sc_should_abort);
BACKOUT;
}
/* All records converted ok. Whether this is live schema change or
* not, the db is readonly at this point so we can reset the live
* schema change flag. */
sc_printf(s, "[%s] --- All records copied --- \n", s->tablename);
/* Before this handle is closed, lets wait for all the db reads to
* finish*/
/* No insert transactions should happen after this
so lock the table. */
rc = restore_constraint_pointers(db, newdb);
if (rc != 0) {
sc_errf(s, "Error restoring constraing pointers!\n");
BACKOUT;
}
/* from this point on failures should goto either backout if recoverable
* or failure if unrecoverable */
newdb->meta = db->meta;
/* TODO: at this point if a backup is going on, it will be bad */
gbl_sc_commit_count++;
/*begin updating things*/
if (newdb->schema_version == 1) {
/* newdb's version has been reset */
bdberr =
bdb_reset_csc2_version(transac, db->tablename, db->schema_version, 1);
if (bdberr != BDBERR_NOERROR)
BACKOUT;
}
if ((rc = alter_table_sequences(iq, transac, db, newdb))) {
sc_errf(s, "Failed alter_table_sequences: %d\n", rc);
BACKOUT;
}
rc = create_datacopy_array(newdb);
if (rc) {
sc_errf(s, "error initializing datacopy array\n");
return -1;
}
if ((rc = prepare_version_for_dbs_without_instant_sc(transac, db, newdb)))
BACKOUT;
/* load new csc2 data */