Open GoogleCodeExporter opened 9 years ago
Original comment by Yoshinor...@gmail.com
on 8 May 2014 at 5:59
Any updates on this?
Original comment by ege...@indeed.com
on 11 Sep 2014 at 6:01
Hi All,
I suspect that's not entirely MasterHA fault - it seems you just hit errant
transactions -
http://www.percona.com/blog/2014/05/19/errant-transactions-major-hurdle-for-gtid
-based-failover-in-mysql-5-6/
If you are using GTID you do not need to write to slave - period. Any writes
will create new transaction on slave which is absent on master and after log
purging - boom! - your failover failed.
We faced with same problem because of pt-heartbeat with SUPER privileges. You
need to be extremely cautious with SUPER users on slaves or apply some patch
for MySQL for "super-read-only" mode when even SUPER user not able to write to
read-only slave.
Or you can check for errant transactions before failover. Or both.
Original comment by denis.zhdanov
on 13 Nov 2014 at 9:16
PS: it will be great to have errant transaction check for masterha_check_repl
(will try to make PR)
Original comment by denis.zhdanov
on 13 Nov 2014 at 9:21
Original issue reported on code.google.com by
stephane...@percona.com
on 8 May 2014 at 9:29