[Pacemaker] postgresql failover

Takatoshi MATSUO matsuo.tak at gmail.com
Tue Aug 20 02:17:52 UTC 2013


Hi Gregg Jaskiewicz

2013/8/16 Gregg Jaskiewicz <gryzman at gmail.com>:
> Running rsync -avzPc -e 'ssh -o UserKnownHostsFile=/dev/null'
> --delete-during 10.0.1.100:/var/lib/pgsql/9.2/data/pg_archive
> /var/lib/pgsql/9.2/data/
> on each slave fixes it - but question then is - why cannot this be done
> automatically by RA ?

I think it's over the top to use rsync by RA.
In addition, It may cause timed-out of monitor.

>
> Andrew on irc suggested I use restart_on_promote, but I have a feeling this
> can be done without restarting anything - however the RA itself would have
> to be fixed, and I can't do it myself to propose a fix, or submit a patch.

I recommend using restart_on_promote parameter too.
because timeline-ID is incremented when promote is called.

If you use restart_on_promote="true",  Saves may be able to connect
new Master without rsync.

Thanks,
Takatoshi MATSUO




More information about the Pacemaker mailing list