[Linux-HA] Monitored resource[s] always re-started first (instead
of migrated) regardless of the stickiness (2.0.8) - SOLVED
Andrew Beekhof
beekhof at gmail.com
Wed Feb 21 08:02:05 MST 2007
On 2/21/07, Andrey Brozhko <andrey at brozhko.com> wrote:
> This is just for the record (for somebody else searching for the answer
> in the archive):
>
> To instruct the HB to migrate the resources immediately after the
> monitored resource failure set the default_resource_failure_stickiness
> to -INFINITY.
the only caveat being that they'll never go back until you clear the
failure_count for that resource on that node (with crm_failcount)
> Andrey Brozhko wrote:
> > Hello,
> >
> > I would like my resource group to immediately migrate to the backup node
> > after failure of one or more resources (as quick as possible). I've
> > written the monitor action for my resource agents and I can see it is
> > called every second for every resource - just the way I configured it in
> > the cib.xml file. I've read all about the forced failover on:
> > http://www.linux-ha.org/v2/faq/forced_failover?highlight=%28faq%29
> > and setup my resources stickiness's as follows:
> >
> > - default_resource_failure_stickiness is -501
> > - default_resource_stickiness is 500
> >
> > Nodes have the same (default) score for this resource group (marked as X
> > in the formula below). So I assumed this should make the resources
> > migrate immediately to the other node if one of them fails sinse:
> >
> > (X - X + 500) / abs(-501) = 0,09.
> >
> > But that does not happen - instead heartbeat attempts to start the
> > failed resource first and only when that fails it migrates the whole
> > group to the other node. Is there a way of altering this behaviour, i.e.
> > if one of the resources fails the group migrates to the backup node
> > immediately?
> >
> >
> > Thanks you,
> > Andrey
> > _______________________________________________
> > Linux-HA mailing list
> > Linux-HA at lists.linux-ha.org
> > http://lists.linux-ha.org/mailman/listinfo/linux-ha
> > See also: http://linux-ha.org/ReportingProblems
> >
>
> _______________________________________________
> Linux-HA mailing list
> Linux-HA at lists.linux-ha.org
> http://lists.linux-ha.org/mailman/listinfo/linux-ha
> See also: http://linux-ha.org/ReportingProblems
>
More information about the Linux-HA
mailing list