Details
-
Bug
-
Status: Open (View Workflow)
-
Major
-
Resolution: Unresolved
-
10.4.13
-
None
-
binlog_format = ROW
gtid_strict_mode = 1
multi-master setup
slaves:
* Parallel_Mode: conservative (default)
* Using_Gtid: Current_Pos
Description
I have a table with a counter that is updated relatively by multiple masters. One of the slaves (that has 2 masters) stopped replicating after an unrelated failure (one of the masters deleted a row that the other master inserted previously, apparently there was some lag).
After restarting the replication from both master (using start all slaves), I ended up with a mismatch of data between the two masters, and the slave. Both masters updated the same counter 6 times, but the slave ended up with a final value of 11 instead of 12:
redacted binlog from master1, with irrelevant fields from Account removed and redacted ID:
#200722 8:28:32 server id 1 end_log_pos 12219708 CRC32 0xf87669df Annotate_rows:
|
#Q> INSERT INTO `Account` (`id`, `testHits`) VALUES (9, 0) |
#200722 8:50:18 server id 1 end_log_pos 12992418 CRC32 0x239cb8e2 Annotate_rows:
|
#Q> UPDATE `Account` SET `testHits` = `testHits` + 1 WHERE `id` = 9 |
|
|
#200727 8:26:22 server id 1 end_log_pos 12421103 CRC32 0x712982e4 Annotate_rows:
|
#Q> UPDATE `Account` SET `testHits` = `testHits` + 1 WHERE `id` = 9 |
#200727 8:26:36 server id 1 end_log_pos 12431646 CRC32 0x3e2f3505 Annotate_rows:
|
#Q> UPDATE `Account` SET `testHits` = `testHits` + 1 WHERE `id` = 9 |
#200727 8:26:43 server id 1 end_log_pos 12436404 CRC32 0x8ab1931a Annotate_rows:
|
#Q> UPDATE `Account` SET `testHits` = `testHits` + 1 WHERE `id` = 9 |
|
|
#200728 7:02:02 server id 1 end_log_pos 340028656 CRC32 0xcf9797a5 Annotate_rows:
|
#Q> UPDATE `Account` SET `testHits` = `testHits` + 1 WHERE `id` = 9 |
#200728 7:02:19 server id 1 end_log_pos 340032448 CRC32 0x71109d4b Annotate_rows:
|
#Q> UPDATE `Account` SET `testHits` = `testHits` + 1 WHERE `id` = 9 |
binlog from master1:
#200722 8:31:20 server id 3 end_log_pos 4420014 CRC32 0xb6bc7350 Annotate_rows:
|
#Q> UPDATE `Account` SET `testHits` = `testHits` + 1 WHERE `id` = 9 |
#200722 9:49:32 server id 3 end_log_pos 5237738 CRC32 0x9627872a Annotate_rows:
|
#Q> UPDATE `Account` SET `testHits` = `testHits` + 1 WHERE `id` = 9 |
|
|
#200727 14:04:32 server id 3 end_log_pos 8793329 CRC32 0x2651f5ea Annotate_rows:
|
#Q> UPDATE `Account` SET `testHits` = `testHits` + 1 WHERE `id` = 9 |
#200727 14:04:45 server id 3 end_log_pos 8794212 CRC32 0x619663f3 Annotate_rows:
|
#Q> UPDATE `Account` SET `testHits` = `testHits` + 1 WHERE `id` = 9 |
#200727 14:12:01 server id 3 end_log_pos 8821614 CRC32 0xef4ded02 Annotate_rows:
|
#Q> UPDATE `Account` SET `testHits` = `testHits` + 1 WHERE `id` = 9 |
|
|
#200728 7:02:11 server id 3 end_log_pos 3108883 CRC32 0x6c2bf4fd Annotate_rows:
|
#Q> UPDATE `Account` SET `testHits` = `testHits` + 1 WHERE `id` = 9 |
Replication was halted at 200728 04:03:11. Both masters had testHits=12, but this particular slave ended up with testHits=11
Is this expected behaviour?