-
Notifications
You must be signed in to change notification settings - Fork 4k
/
Copy pathpartition_handler.cc
2965 lines (2599 loc) · 101 KB
/
partition_handler.cc
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 (c) 2005, 2018, Oracle and/or its affiliates. All rights reserved.
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License, version 2.0,
as published by the Free Software Foundation.
This program is also distributed with certain software (including
but not limited to OpenSSL) that is licensed under separate terms,
as designated in a particular file or component or in included license
documentation. The authors of MySQL hereby grant you an additional
permission to link the program and your derivative works with the
separately licensed software that they have included with MySQL.
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, version 2.0, 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 St, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "sql/partitioning/partition_handler.h"
#include <fcntl.h>
#include <limits.h>
#include <stdarg.h>
#include <stdio.h>
#include <new>
#include <utility>
#include "binary_log_types.h"
#include "lex_string.h"
#include "m_ctype.h"
#include "m_string.h"
#include "map_helpers.h"
#include "my_bitmap.h"
#include "my_byteorder.h"
#include "my_compiler.h"
#include "my_dbug.h"
#include "my_loglevel.h"
#include "my_macros.h"
#include "my_psi_config.h"
#include "my_sqlcommand.h"
#include "myisam.h" // MI_MAX_MSG_BUF
#include "mysql/components/services/log_builtins.h"
#include "mysql/components/services/mysql_mutex_bits.h"
#include "mysql/components/services/psi_memory_bits.h"
#include "mysql/components/services/psi_mutex_bits.h"
#include "mysql/plugin.h"
#include "mysql/psi/mysql_memory.h"
#include "mysql/psi/psi_base.h"
#include "mysql/service_mysql_alloc.h"
#include "mysql_com.h"
#include "mysqld_error.h"
#include "sql/derror.h"
#include "sql/discrete_interval.h"
#include "sql/field.h"
#include "sql/key.h" // key_rec_cmp
#include "sql/log.h"
#include "sql/partition_element.h"
#include "sql/partition_info.h" // NOT_A_PARTITION_ID
#include "sql/protocol.h"
#include "sql/protocol_classic.h"
#include "sql/set_var.h"
#include "sql/sql_alter.h"
#include "sql/sql_class.h" // THD
#include "sql/sql_const.h"
#include "sql/sql_lex.h"
#include "sql/sql_list.h"
#include "sql/sql_partition.h" // LIST_PART_ENTRY, part_id_range
#include "sql/system_variables.h"
#include "sql/table.h" // TABLE_SHARE
#include "sql/thr_malloc.h"
#include "sql_string.h"
#include "template_utils.h"
#include "thr_mutex.h"
namespace dd {
class Table;
} // namespace dd
// In sql_class.cc:
int thd_binlog_format(const MYSQL_THD thd);
/** operation names for the enum_part_operation. */
static const char *opt_op_name[] = {
"optimize", "analyze", "check", "repair",
"assign_to_keycache", "preload_keys"};
static PSI_memory_key key_memory_Partition_share;
static PSI_memory_key key_memory_partition_sort_buffer;
static PSI_memory_key key_memory_Partition_admin;
#ifdef HAVE_PSI_INTERFACE
PSI_mutex_key key_partition_auto_inc_mutex;
static PSI_memory_info all_partitioning_memory[] = {
{&key_memory_Partition_share, "Partition_share", 0, 0, PSI_DOCUMENT_ME},
{&key_memory_partition_sort_buffer, "partition_sort_buffer", 0, 0,
PSI_DOCUMENT_ME},
{&key_memory_Partition_admin, "Partition_admin", 0, 0, PSI_DOCUMENT_ME}};
static PSI_mutex_info all_partitioning_mutex[] = {
{&key_partition_auto_inc_mutex, "Partition_share::auto_inc_mutex", 0, 0,
PSI_DOCUMENT_ME}};
#endif
void partitioning_init() {
#ifdef HAVE_PSI_INTERFACE
int count;
count = static_cast<int>(array_elements(all_partitioning_memory));
mysql_memory_register("sql", all_partitioning_memory, count);
count = static_cast<int>(array_elements(all_partitioning_mutex));
mysql_mutex_register("sql", all_partitioning_mutex, count);
#endif
}
/*
Implementation of Partition_share class.
*/
Partition_share::Partition_share()
: auto_inc_initialized(false),
auto_inc_mutex(NULL),
next_auto_inc_val(0),
partition_names(NULL) {}
Partition_share::~Partition_share() {
if (auto_inc_mutex) {
mysql_mutex_destroy(auto_inc_mutex);
my_free(auto_inc_mutex);
}
if (partition_names) {
my_free(partition_names);
}
}
/**
Initialize auto increment mutex in share.
@return Operation status.
@retval true Failure (out of memory).
@retval false Success.
*/
bool Partition_share::init_auto_inc_mutex(
TABLE_SHARE *table_share MY_ATTRIBUTE((unused))) {
DBUG_ENTER("Partition_share::init_auto_inc_mutex");
DBUG_ASSERT(!auto_inc_mutex);
#ifndef DBUG_OFF
if (table_share->tmp_table == NO_TMP_TABLE) {
mysql_mutex_assert_owner(&table_share->LOCK_ha_data);
}
#endif /* DBUG_OFF */
auto_inc_mutex = static_cast<mysql_mutex_t *>(my_malloc(
key_memory_Partition_share, sizeof(*auto_inc_mutex), MYF(MY_WME)));
if (!auto_inc_mutex) {
DBUG_RETURN(true);
}
mysql_mutex_init(key_partition_auto_inc_mutex, auto_inc_mutex,
MY_MUTEX_INIT_FAST);
DBUG_RETURN(false);
}
/**
Release reserved auto increment values not used.
@param thd Thread.
@param table_share Table Share
@param next_insert_id Next insert id (first non used auto inc value).
@param max_reserved End of reserved auto inc range.
*/
void Partition_share::release_auto_inc_if_possible(
THD *thd, TABLE_SHARE *table_share MY_ATTRIBUTE((unused)),
const ulonglong next_insert_id, const ulonglong max_reserved) {
DBUG_ASSERT(auto_inc_mutex);
#ifndef DBUG_OFF
if (table_share->tmp_table == NO_TMP_TABLE) {
mysql_mutex_assert_owner(auto_inc_mutex);
}
#endif /* DBUG_OFF */
/*
If the current auto_increment values is lower than the reserved value (1)
and the reserved value was reserved by this thread (2), then we can
lower the reserved value.
However, we cannot lower the value if there are forced/non generated
values from 'SET INSERT_ID = forced_val' (3). */
if (next_insert_id < next_auto_inc_val && // (1)
max_reserved >= next_auto_inc_val && // (2)
thd->auto_inc_intervals_forced.maximum() < next_insert_id) // (3)
{
next_auto_inc_val = next_insert_id;
}
}
/**
Populate the partition_name_hash in part_share.
*/
bool Partition_share::populate_partition_name_hash(partition_info *part_info) {
uint tot_names;
uint num_subparts = part_info->num_subparts;
DBUG_ENTER("Partition_share::populate_partition_name_hash");
DBUG_ASSERT(!part_info->is_sub_partitioned() || num_subparts);
if (num_subparts == 0) {
num_subparts = 1;
}
/*
TABLE_SHARE::LOCK_ha_data must been locked before calling this function.
This ensures only one thread/table instance will execute this.
*/
#ifndef DBUG_OFF
if (part_info->table->s->tmp_table == NO_TMP_TABLE) {
mysql_mutex_assert_owner(&part_info->table->s->LOCK_ha_data);
}
#endif
if (partition_name_hash != nullptr) {
DBUG_RETURN(false);
}
tot_names = part_info->num_parts;
if (part_info->is_sub_partitioned()) {
tot_names += part_info->num_parts * num_subparts;
}
partition_names = static_cast<const uchar **>(my_malloc(
key_memory_Partition_share,
part_info->get_tot_partitions() * sizeof(*partition_names), MYF(MY_WME)));
if (!partition_names) {
DBUG_RETURN(true);
}
partition_name_hash.reset(
new collation_unordered_map<std::string,
unique_ptr_my_free<PART_NAME_DEF>>(
system_charset_info, key_memory_Partition_share));
List_iterator<partition_element> part_it(part_info->partitions);
uint i = 0;
do {
partition_element *part_elem = part_it++;
DBUG_ASSERT(part_elem->part_state == PART_NORMAL);
if (part_elem->part_state == PART_NORMAL) {
if (insert_partition_name_in_hash(part_elem->partition_name,
i * num_subparts, false))
goto err;
if (part_info->is_sub_partitioned()) {
List_iterator<partition_element> subpart_it(part_elem->subpartitions);
partition_element *sub_elem;
uint j = 0;
do {
sub_elem = subpart_it++;
if (insert_partition_name_in_hash(sub_elem->partition_name,
i * num_subparts + j, true))
goto err;
} while (++j < num_subparts);
}
}
} while (++i < part_info->num_parts);
for (const auto &key_and_value : *partition_name_hash) {
PART_NAME_DEF *part_def = key_and_value.second.get();
if (part_def->is_subpart == part_info->is_sub_partitioned()) {
partition_names[part_def->part_id] = part_def->partition_name;
}
}
DBUG_RETURN(false);
err:
partition_name_hash.reset();
my_free(partition_names);
partition_names = NULL;
DBUG_RETURN(true);
}
/**
Insert a partition name in the partition_name_hash.
@param name Name of partition
@param part_id Partition id (number)
@param is_subpart Set if the name belongs to a subpartition
@return Operation status
@retval true Failure
@retval false Success
*/
bool Partition_share::insert_partition_name_in_hash(const char *name,
uint part_id,
bool is_subpart) {
PART_NAME_DEF *part_def;
char *part_name;
uint part_name_length;
DBUG_ENTER("Partition_share::insert_partition_name_in_hash");
/*
Calculate and store the length here, to avoid doing it when
searching the hash.
*/
part_name_length = static_cast<uint>(strlen(name));
/*
Must use memory that lives as long as table_share.
Freed in the Partition_share destructor.
Since we use my_multi_malloc, then my_free(part_def) will also free
part_name, as a part of my_hash_free.
*/
if (!my_multi_malloc(key_memory_Partition_share, MY_WME, &part_def,
sizeof(PART_NAME_DEF), &part_name, part_name_length + 1,
NULL)) {
DBUG_RETURN(true);
}
memcpy(part_name, name, part_name_length + 1);
part_def->partition_name = pointer_cast<uchar *>(part_name);
part_def->length = part_name_length;
part_def->part_id = part_id;
part_def->is_subpart = is_subpart;
DBUG_RETURN(
!partition_name_hash
->emplace(part_name, unique_ptr_my_free<PART_NAME_DEF>(part_def))
.second);
}
const char *Partition_share::get_partition_name(size_t part_id) const {
if (partition_names == NULL) {
return NULL;
}
return reinterpret_cast<const char *>(partition_names[part_id]);
}
int Partition_handler::truncate_partition(dd::Table *table_def) {
handler *file = get_handler();
if (!file) {
return HA_ERR_WRONG_COMMAND;
}
DBUG_ASSERT(file->table_share->tmp_table != NO_TMP_TABLE ||
file->m_lock_type == F_WRLCK);
file->mark_trx_read_write();
return truncate_partition_low(table_def);
}
int Partition_handler::exchange_partition(const char *partition_path,
const char *swap_table_path,
uint part_id,
dd::Table *part_table_def,
dd::Table *swap_table_def) {
handler *file = get_handler();
if (!file) {
return HA_ERR_WRONG_COMMAND;
}
DBUG_ASSERT(file->table_share->tmp_table != NO_TMP_TABLE ||
file->m_lock_type != F_UNLCK);
file->mark_trx_read_write();
return exchange_partition_low(partition_path, swap_table_path, part_id,
part_table_def, swap_table_def);
}
/*
Implementation of Partition_helper class.
*/
Partition_helper::Partition_helper(handler *main_handler)
: m_handler(main_handler),
m_part_info(),
m_tot_parts(),
m_last_part(),
m_err_rec(),
m_ordered(),
m_ordered_scan_ongoing(),
m_ordered_rec_buffer(),
m_queue() {}
Partition_helper::~Partition_helper() {
DBUG_ASSERT(m_ordered_rec_buffer == NULL);
DBUG_ASSERT(m_key_not_found_partitions.bitmap == NULL);
}
/**
Set partition info.
To be called from Partition_handler.
@param part_info Partition info to use.
@param early True if called when part_info only created and parsed,
but not setup, checked or fixed.
*/
void Partition_helper::set_part_info_low(partition_info *part_info,
bool early) {
/*
ha_partition will set m_tot_parts from the .par file during creating
the new handler.
And this call can be earlier than the partition_default_handling(),
so get_tot_partitions() may return zero.
*/
if (m_tot_parts == 0 && (m_part_info == NULL || !early)) {
m_tot_parts = part_info->get_tot_partitions();
}
m_part_info = part_info;
m_is_sub_partitioned = m_part_info->is_sub_partitioned();
}
/**
Initialize the partitioning helper for use after the table is opened.
@param part_share Partitioning share (used for auto increment).
@return Operation status.
@retval false for success otherwise true.
*/
bool Partition_helper::open_partitioning(Partition_share *part_share) {
m_table = get_table();
DBUG_ASSERT(m_part_info == m_table->part_info);
m_part_share = part_share;
m_tot_parts = m_part_info->get_tot_partitions();
if (bitmap_init(&m_key_not_found_partitions, NULL, m_tot_parts, false)) {
return true;
}
bitmap_clear_all(&m_key_not_found_partitions);
m_key_not_found = false;
m_is_sub_partitioned = m_part_info->is_sub_partitioned();
m_auto_increment_lock = false;
m_auto_increment_safe_stmt_log_lock = false;
m_pkey_is_clustered = m_handler->primary_key_is_clustered();
m_part_spec.start_part = NOT_A_PARTITION_ID;
m_part_spec.end_part = NOT_A_PARTITION_ID;
m_index_scan_type = PARTITION_NO_INDEX_SCAN;
m_start_key.key = NULL;
m_start_key.length = 0;
m_scan_value = 3;
m_reverse_order = false;
m_curr_key_info[0] = NULL;
m_curr_key_info[1] = NULL;
m_curr_key_info[2] = NULL;
m_top_entry = NO_CURRENT_PART_ID;
m_ref_usage = REF_NOT_USED;
m_rec_length = m_table->s->reclength;
return false;
}
void Partition_helper::close_partitioning() {
bitmap_free(&m_key_not_found_partitions);
DBUG_ASSERT(!m_ordered_rec_buffer);
destroy_record_priority_queue();
}
void Partition_helper::lock_auto_increment() {
/* lock already taken */
if (m_auto_increment_safe_stmt_log_lock) return;
DBUG_ASSERT(!m_auto_increment_lock);
if (m_table->s->tmp_table == NO_TMP_TABLE) {
m_auto_increment_lock = true;
m_part_share->lock_auto_inc();
}
}
/****************************************************************************
MODULE change record
****************************************************************************/
/**
Insert a row to the partitioned table.
@param buf The row in MySQL Row Format.
@return Operation status.
@retval 0 Success
@retval != 0 Error code
*/
int Partition_helper::ph_write_row(uchar *buf) {
uint32 part_id;
int error;
longlong func_value;
bool have_auto_increment =
m_table->next_number_field && buf == m_table->record[0];
THD *thd = get_thd();
sql_mode_t saved_sql_mode = thd->variables.sql_mode;
bool saved_auto_inc_field_not_null = m_table->auto_increment_field_not_null;
#ifndef DBUG_OFF
my_bitmap_map *old_map;
#endif /* DBUG_OFF */
DBUG_ENTER("Partition_helper::ph_write_row");
DBUG_ASSERT(buf == m_table->record[0]);
/*
If we have an auto_increment column and we are writing a changed row
or a new row, then update the auto_increment value in the record.
*/
if (have_auto_increment) {
error = m_handler->update_auto_increment();
/*
If we have failed to set the auto-increment value for this row,
it is highly likely that we will not be able to insert it into
the correct partition. We must check and fail if neccessary.
*/
if (error) DBUG_RETURN(error);
/*
Don't allow generation of auto_increment value the partitions handler.
If a partitions handler would change the value, then it might not
match the partition any longer.
This can occur if 'SET INSERT_ID = 0; INSERT (NULL)',
So allow this by adding 'MODE_NO_AUTO_VALUE_ON_ZERO' to sql_mode.
The partitions handler::next_insert_id must always be 0. Otherwise
we need to forward release_auto_increment, or reset it for all
partitions.
*/
if (m_table->next_number_field->val_int() == 0) {
m_table->auto_increment_field_not_null = true;
thd->variables.sql_mode |= MODE_NO_AUTO_VALUE_ON_ZERO;
}
}
#ifndef DBUG_OFF
/* Temporary mark the partitioning fields as readable. */
old_map = dbug_tmp_use_all_columns(m_table, m_table->read_set);
#endif /* DBUG_OFF */
error = m_part_info->get_partition_id(m_part_info, &part_id, &func_value);
#ifndef DBUG_OFF
dbug_tmp_restore_column_map(m_table->read_set, old_map);
#endif /* DBUG_OFF */
if (unlikely(error)) {
m_part_info->err_value = func_value;
goto exit;
}
if (!m_part_info->is_partition_locked(part_id)) {
DBUG_PRINT("info", ("Write to non-locked partition %u (func_value: %ld)",
part_id, (long)func_value));
error = HA_ERR_NOT_IN_LOCK_PARTITIONS;
goto exit;
}
m_last_part = part_id;
DBUG_PRINT("info", ("Insert in partition %d", part_id));
error = write_row_in_part(part_id, buf);
if (have_auto_increment && !m_table->s->next_number_keypart) {
set_auto_increment_if_higher();
}
exit:
thd->variables.sql_mode = saved_sql_mode;
m_table->auto_increment_field_not_null = saved_auto_inc_field_not_null;
DBUG_RETURN(error);
}
/**
Update an existing row in the partitioned table.
Yes, update_row() does what you expect, it updates a row. old_data will
have the previous row record in it, while new_data will have the newest
data in it.
Keep in mind that the server can do updates based on ordering if an
ORDER BY clause was used. Consecutive ordering is not guaranteed.
If the new record belongs to a different partition than the old record
then it will be inserted into the new partition and deleted from the old.
new_data is always record[0]
old_data is always record[1]
@param old_data The old record in MySQL Row Format.
@param new_data The new record in MySQL Row Format.
@return Operation status.
@retval 0 Success
@retval != 0 Error code
*/
int Partition_helper::ph_update_row(const uchar *old_data, uchar *new_data) {
uint32 new_part_id, old_part_id;
int error = 0;
longlong func_value;
DBUG_ENTER("Partition_helper::ph_update_row");
m_err_rec = NULL;
// Need to read partition-related columns, to locate the row's partition:
DBUG_ASSERT(
bitmap_is_subset(&m_part_info->full_part_field_set, m_table->read_set));
if ((error = get_parts_for_update(old_data, new_data, m_table->record[0],
m_part_info, &old_part_id, &new_part_id,
&func_value))) {
DBUG_RETURN(error);
}
if (!bitmap_is_set(&(m_part_info->lock_partitions), new_part_id)) {
error = HA_ERR_NOT_IN_LOCK_PARTITIONS;
DBUG_RETURN(error);
}
/*
The protocol for updating a row is:
1) position the handler (cursor) on the row to be updated,
either through the last read row (rnd or index) or by rnd_pos.
2) call update_row with both old and new full records as arguments.
This means that m_last_part should already be set to actual partition
where the row was read from. And if that is not the same as the
calculated part_id we found a misplaced row, we return an error to
notify the user that something is broken in the row distribution
between partitions! Since we don't check all rows on read, we return an
error instead of correcting m_last_part, to make the user aware of the
problem!
Notice that HA_READ_BEFORE_WRITE_REMOVAL does not require this protocol,
so this is not supported for this engine.
*/
if (old_part_id != m_last_part) {
m_err_rec = old_data;
DBUG_RETURN(HA_ERR_ROW_IN_WRONG_PARTITION);
}
m_last_part = new_part_id;
if (new_part_id == old_part_id) {
DBUG_PRINT("info", ("Update in partition %d", new_part_id));
error = update_row_in_part(new_part_id, old_data, new_data);
} else {
Field *saved_next_number_field = m_table->next_number_field;
/*
Don't allow generation of auto_increment value for update.
table->next_number_field is never set on UPDATE.
But is set for INSERT ... ON DUPLICATE KEY UPDATE,
and since update_row() does not generate or update an auto_inc value,
we cannot have next_number_field set when moving a row
to another partition with write_row(), since that could
generate/update the auto_inc value.
This gives the same behavior for partitioned vs non partitioned tables.
*/
m_table->next_number_field = NULL;
DBUG_PRINT("info", ("Update from partition %d to partition %d", old_part_id,
new_part_id));
error = write_row_in_part(new_part_id, new_data);
m_table->next_number_field = saved_next_number_field;
if (!error) {
error = delete_row_in_part(old_part_id, old_data);
}
}
/*
if updating an auto_increment column, update
m_part_share->next_auto_inc_val if needed.
(not to be used if auto_increment on secondary field in a multi-column
index)
mysql_update does not set table->next_number_field, so we use
table->found_next_number_field instead.
Also checking that the field is marked in the write set.
*/
if (m_table->found_next_number_field && new_data == m_table->record[0] &&
!m_table->s->next_number_keypart &&
bitmap_is_set(m_table->write_set,
m_table->found_next_number_field->field_index)) {
set_auto_increment_if_higher();
}
DBUG_RETURN(error);
}
/**
Delete an existing row in the partitioned table.
This will delete a row. buf will contain a copy of the row to be deleted.
The server will call this right after the current row has been read
(from either a previous rnd_xxx() or index_xxx() call).
If you keep a pointer to the last row or can access a primary key it will
make doing the deletion quite a bit easier.
Keep in mind that the server does no guarentee consecutive deletions.
ORDER BY clauses can be used.
buf is either record[0] or record[1]
@param buf The record in MySQL Row Format.
@return Operation status.
@retval 0 Success
@retval != 0 Error code
*/
int Partition_helper::ph_delete_row(const uchar *buf) {
int error;
uint part_id;
DBUG_ENTER("Partition_helper::ph_delete_row");
m_err_rec = NULL;
DBUG_ASSERT(
bitmap_is_subset(&m_part_info->full_part_field_set, m_table->read_set));
if ((error = get_part_for_delete(buf, m_table->record[0], m_part_info,
&part_id))) {
DBUG_RETURN(error);
}
if (!m_part_info->is_partition_locked(part_id)) {
DBUG_RETURN(HA_ERR_NOT_IN_LOCK_PARTITIONS);
}
/*
The protocol for deleting a row is:
1) position the handler (cursor) on the row to be deleted,
either through the last read row (rnd or index) or by rnd_pos.
2) call delete_row with the full record as argument.
This means that m_last_part should already be set to actual partition
where the row was read from. And if that is not the same as the
calculated part_id we found a misplaced row, we return an error to
notify the user that something is broken in the row distribution
between partitions! Since we don't check all rows on read, we return an
error instead of forwarding the delete to the correct (m_last_part)
partition!
Notice that HA_READ_BEFORE_WRITE_REMOVAL does not require this protocol,
so this is not supported for this engine.
TODO: change the assert in InnoDB into an error instead and make this one
an assert instead and remove the get_part_for_delete()!
*/
if (part_id != m_last_part) {
m_err_rec = buf;
DBUG_RETURN(HA_ERR_ROW_IN_WRONG_PARTITION);
}
/* Should never call delete_row on a partition which is not read */
DBUG_ASSERT(m_part_info->is_partition_used(part_id));
m_last_part = part_id;
error = delete_row_in_part(part_id, buf);
DBUG_RETURN(error);
}
/**
Get a range of auto increment values.
Can only be used if the auto increment field is the first field in an index.
This method is called by update_auto_increment which in turn is called
by the individual handlers as part of write_row. We use the
part_share->next_auto_inc_val, or search all
partitions for the highest auto_increment_value if not initialized or
if auto_increment field is a secondary part of a key, we must search
every partition when holding a mutex to be sure of correctness.
@param[in] increment Increment value.
@param[in] nb_desired_values Number of desired values.
@param[out] first_value First auto inc value reserved
or MAX if failure.
@param[out] nb_reserved_values Number of values reserved.
*/
void Partition_helper ::get_auto_increment_first_field(
ulonglong increment, ulonglong nb_desired_values, ulonglong *first_value,
ulonglong *nb_reserved_values) {
THD *thd = get_thd();
DBUG_ENTER("Partition_helper::get_auto_increment_first_field");
DBUG_PRINT("info",
("inc: %lu desired_values: %lu first_value: %lu", (ulong)increment,
(ulong)nb_desired_values, (ulong)*first_value));
DBUG_ASSERT(increment && nb_desired_values);
/*
next_number_keypart is != 0 if the auto_increment column is a secondary
column in the index (it is allowed in MyISAM)
*/
DBUG_ASSERT(m_table->s->next_number_keypart == 0);
*first_value = 0;
/*
Get a lock for handling the auto_increment in part_share
for avoiding two concurrent statements getting the same number.
*/
lock_auto_increment();
/* Initialize if not already done. */
if (!m_part_share->auto_inc_initialized) {
initialize_auto_increment(false);
}
/*
In a multi-row insert statement like INSERT SELECT and LOAD DATA
where the number of candidate rows to insert is not known in advance
we must hold a lock/mutex for the whole statement if we have statement
based replication. Because the statement-based binary log contains
only the first generated value used by the statement, and slaves assumes
all other generated values used by this statement were consecutive to
this first one, we must exclusively lock the generator until the statement
is done.
*/
int binlog_format = thd_binlog_format(thd);
if (!m_auto_increment_safe_stmt_log_lock &&
thd->lex->sql_command != SQLCOM_INSERT &&
binlog_format != BINLOG_FORMAT_UNSPEC &&
binlog_format != BINLOG_FORMAT_ROW) {
DBUG_PRINT("info", ("locking auto_increment_safe_stmt_log_lock"));
m_auto_increment_safe_stmt_log_lock = true;
}
/* this gets corrected (for offset/increment) in update_auto_increment */
*first_value = m_part_share->next_auto_inc_val;
m_part_share->next_auto_inc_val += nb_desired_values * increment;
if (m_part_share->next_auto_inc_val < *first_value) {
/* Overflow, set to max. */
m_part_share->next_auto_inc_val = ULLONG_MAX;
}
unlock_auto_increment();
DBUG_PRINT("info", ("*first_value: %lu", (ulong)*first_value));
*nb_reserved_values = nb_desired_values;
DBUG_VOID_RETURN;
}
inline void Partition_helper::set_auto_increment_if_higher() {
Field_num *field = static_cast<Field_num *>(m_table->found_next_number_field);
ulonglong nr =
(field->unsigned_flag || field->val_int() > 0) ? field->val_int() : 0;
lock_auto_increment();
if (!m_part_share->auto_inc_initialized) {
initialize_auto_increment(false);
}
/* must hold the mutex when looking/changing m_part_share. */
if (nr >= m_part_share->next_auto_inc_val) {
m_part_share->next_auto_inc_val = nr + 1;
}
unlock_auto_increment();
save_auto_increment(nr);
}
void Partition_helper::ph_release_auto_increment() {
DBUG_ENTER("Partition_helper::ph_release_auto_increment");
if (m_table->s->next_number_keypart) {
release_auto_increment_all_parts();
} else if (m_handler->next_insert_id) {
ulonglong max_reserved = m_handler->auto_inc_interval_for_cur_row.maximum();
lock_auto_increment();
m_part_share->release_auto_inc_if_possible(
get_thd(), m_table->s, m_handler->next_insert_id, max_reserved);
DBUG_PRINT("info", ("part_share->next_auto_inc_val: %lu",
(ulong)m_part_share->next_auto_inc_val));
/* Unlock the multi row statement lock taken in get_auto_increment */
if (m_auto_increment_safe_stmt_log_lock) {
m_auto_increment_safe_stmt_log_lock = false;
DBUG_PRINT("info", ("unlocking auto_increment_safe_stmt_log_lock"));
}
unlock_auto_increment();
}
DBUG_VOID_RETURN;
}
/**
Calculate key hash value from an null terminated array of fields.
Support function for KEY partitioning.
@param field_array An array of the fields in KEY partitioning
@return hash_value calculated
@note Uses the hash function on the character set of the field.
Integer and floating point fields use the binary character set by default.
*/
uint32 Partition_helper::ph_calculate_key_hash_value(Field **field_array) {
ulong nr1 = 1;
ulong nr2 = 4;
bool use_51_hash = (*field_array)->table->part_info->key_algorithm ==
enum_key_algorithm::KEY_ALGORITHM_51;
do {
Field *field = *field_array;
if (use_51_hash) {
switch (field->real_type()) {
case MYSQL_TYPE_TINY:
case MYSQL_TYPE_SHORT:
case MYSQL_TYPE_LONG:
case MYSQL_TYPE_FLOAT:
case MYSQL_TYPE_DOUBLE:
case MYSQL_TYPE_NEWDECIMAL:
case MYSQL_TYPE_TIMESTAMP:
case MYSQL_TYPE_LONGLONG:
case MYSQL_TYPE_INT24:
case MYSQL_TYPE_TIME:
case MYSQL_TYPE_DATETIME:
case MYSQL_TYPE_YEAR:
case MYSQL_TYPE_NEWDATE: {
if (field->is_null()) {
nr1 ^= (nr1 << 1) | 1;
continue;
}
/* Force this to my_hash_sort_bin, which was used in 5.1! */
uint len = field->pack_length();
my_charset_bin.coll->hash_sort(&my_charset_bin, field->ptr, len, &nr1,
&nr2);
/* Done with this field, continue with next one. */
continue;
}
case MYSQL_TYPE_STRING:
case MYSQL_TYPE_VARCHAR:
case MYSQL_TYPE_BIT:
/* Not affected, same in 5.1 and 5.5 */
break;
/*
ENUM/SET uses my_hash_sort_simple in 5.1 (i.e. my_charset_latin1)
and my_hash_sort_bin in 5.5!
*/
case MYSQL_TYPE_ENUM:
case MYSQL_TYPE_SET: {
if (field->is_null()) {
nr1 ^= (nr1 << 1) | 1;
continue;
}
/* Force this to my_hash_sort_bin, which was used in 5.1! */
uint len = field->pack_length();
my_charset_latin1.coll->hash_sort(&my_charset_latin1, field->ptr, len,
&nr1, &nr2);
continue;
}
/* New types in mysql-5.6. */
case MYSQL_TYPE_DATETIME2:
case MYSQL_TYPE_TIME2:
case MYSQL_TYPE_TIMESTAMP2:
/* Not affected, 5.6+ only! */
break;
/* These types should not be allowed for partitioning! */
case MYSQL_TYPE_NULL:
case MYSQL_TYPE_DECIMAL:
case MYSQL_TYPE_DATE:
case MYSQL_TYPE_TINY_BLOB:
case MYSQL_TYPE_MEDIUM_BLOB:
case MYSQL_TYPE_LONG_BLOB:
case MYSQL_TYPE_BLOB:
case MYSQL_TYPE_VAR_STRING:
case MYSQL_TYPE_GEOMETRY:
/* fall through. */
default:
DBUG_ASSERT(0); // New type?
/* Fall through for default hashing (5.5). */
}
/* fall through, use collation based hashing. */
}
field->hash(&nr1, &nr2);
} while (*(++field_array));
return (uint32)nr1;
}
bool Partition_helper::print_partition_error(int error) {
THD *thd = get_thd();
DBUG_ENTER("Partition_helper::print_partition_error");
/* Should probably look for my own errors first */
DBUG_PRINT("enter", ("error: %d", error));
if ((error == HA_ERR_NO_PARTITION_FOUND) &&
(thd->lex->alter_info == NULL ||
!(thd->lex->alter_info->flags & Alter_info::ALTER_TRUNCATE_PARTITION))) {
m_part_info->print_no_partition_found(thd, m_table);
// print_no_partition_found() reports an error, so we can just return here.
DBUG_RETURN(false);
} else if (error == HA_ERR_ROW_IN_WRONG_PARTITION) {
/*
Should only happen on DELETE or UPDATE!
Or in ALTER TABLE REBUILD/REORGANIZE where there are a misplaced
row that needed to move to an old partition (not in the given set).
*/
DBUG_ASSERT(thd_sql_command(thd) == SQLCOM_DELETE ||
thd_sql_command(thd) == SQLCOM_DELETE_MULTI ||
thd_sql_command(thd) == SQLCOM_UPDATE ||
thd_sql_command(thd) == SQLCOM_UPDATE_MULTI ||
thd_sql_command(thd) == SQLCOM_ALTER_TABLE);
DBUG_ASSERT(m_err_rec);
if (m_err_rec) {
size_t max_length;
char buf[MAX_KEY_LENGTH];
String str(buf, sizeof(buf), system_charset_info);
uint32 part_id;
DBUG_ASSERT(m_last_part < m_tot_parts);
str.length(0);
if (thd_sql_command(thd) == SQLCOM_ALTER_TABLE) {
str.append("from REBUILD/REORGANIZED partition: ");
str.append_ulonglong(m_last_part);
str.append(" to non included partition (new definition): ");
} else {
str.append_ulonglong(m_last_part);
str.append(". Correct is ");
}
if (get_part_for_delete(m_err_rec, m_table->record[0], m_part_info,
&part_id)) {
str.append("?");
} else {
str.append_ulonglong(part_id);
}
append_row_to_str(str, m_err_rec, m_table);
/* Log this error, so the DBA can notice it and fix it! */
LogErr(ERROR_LEVEL, ER_ROW_IN_WRONG_PARTITION_PLEASE_REPAIR,
m_table->s->table_name.str, str.c_ptr_safe());
max_length =
(MYSQL_ERRMSG_SIZE - strlen(ER_THD(thd, ER_ROW_IN_WRONG_PARTITION)));
if (str.length() >= max_length) {
str.length(max_length - 4);
str.append(STRING_WITH_LEN("..."));
}