tx conflicts

39 views
Skip to first unread message

Daniel Black

unread,
Jul 14, 2014, 9:46:14 PM7/14/14
to codersh...@googlegroups.com

With log messages like the following is there a sane way to parse this back into something useful like a database, table and/or row? If so can it be done by galera instead of me?

9:40:57 [Note] WSREP: trx conflict for key (1,FLAT8)65ff9f39 e8c2fa20: source: f734dbd4-03ad-11e4-adb0-67819bed06e8 version: 3 local: 0 state: CERTIFYING flags: 1 conn_id: 317902 trx_id: 60435815 seqnos (l: 1682417, g: 21956902, s: 21956885, d: 21956734, ts: 794238915267495) <--X--> source: 535a24cb-03ae-11e4-a429-4723cd38bd9a version: 3 local: 1 state: COMMITTED flags: 1 conn_id: 475341 trx_id: 65728464 seqnos (l: 1682409, g: 21956894, s: 21956886, d: 21956734, ts: 794117838622465)

If it can't be made sense of can I ask for a feature request that galera write the binlog data from this to its own log.

--
Daniel Black, Engineer @ Open Query (http://openquery.com.au)
Remote expertise & maintenance for MySQL/MariaDB server environments.

alexey.y...@galeracluster.com

unread,
Jul 18, 2014, 9:08:18 AM7/18/14
to codersh...@googlegroups.com
Daniel, setting repl.key_format=FLAT8A or FLAT16A you should see key
annotations.

And yes, such feature request seems reasonable, however it is not as
straightforward as it may seem. E.g. dumping all of them into a single
file will require some extraction utility so that you could use
mysqlbinlog to parse individual writesets. Perhaps dumping them in pairs
would make more sense.

Daniel Black

unread,
Jul 20, 2014, 2:01:00 AM7/20/14
to codersh...@googlegroups.com


----- Original Message -----
> Daniel, setting repl.key_format=FLAT8A or FLAT16A you should see key
> annotations.

Thank you, I'll start to use these by default.

> And yes, such feature request seems reasonable, however it is not as
> straightforward as it may seem. E.g. dumping all of them into a single
> file will require some extraction utility so that you could use
> mysqlbinlog to parse individual writesets.

So an annotation bin log entry indicating start of writeset making the output file a normal binlog file?

> Perhaps dumping them in pairs would make more sense.

Not sure which pairs you're talking about here. Either way I'll take any improvement you're offering :-)

--
Reply all
Reply to author
Forward
0 new messages