[Pacemaker] Multi-State Resources and failback ?

[Andrew Beekhof]

On Wed, Feb 17, 2010 at 3:00 PM,  <rvm at free.fr> wrote:
> Hi all,
>
> Today, I'm trying a MS set on my two nodes cluster :
>
> primitive TEST ocf:me:mysql \
>        op monitor interval="10s" role="Master" timeout="10s" \
>        op monitor interval="11s" role="Slave" timeout="10s" \
>        meta migration-threshold="2"
> ms MS TEST \
>        meta clone-max="2" notify="false" globally-unique="false" target-role="Started"
>
> Mysql ocf script is based on heartbeat:Stateful script.
> Node A is master, Node B is slave.
> After two failures of mysql on A (killall mysqld), B becomes master and promotes  MySQL. --> ok.
> Then, I clean my resource with : resource cleanup MS. A becomes slave and start MySQL. No change on B. --> ok.
> Now if I kill my mysqld daemon on B, my monitor shows this error and A become master, B slave. migration-threshold is not used ?

Not if the failure was of a start action (though there is an option to
enable this).

> Any ideas, comments, suggestions ?

I'd need to see logs

> For this test, I use Debian Lenny, pacemaker 1.0.7+hg20100203-1 and heartbeat 1:3.0.2-1~bpo50+1.
>
> Thanks ;-)
> Regards
>
> _______________________________________________
> Pacemaker mailing list
> Pacemaker at oss.clusterlabs.org
> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>