On Wed, Dec 10, 2008 at 12:16, Michael Weiser
<[email protected]> wrote:
> Hello,
>
> I'm seeing seemingly unneccessary resource restarts with heartbeat-2.1.3
> (Centos RPM 2.1.3-3.el5.centos) on RHEL5 and a quite current Debian
> testing (heartbeat package 2.1.3-3).
>
> I have a location constraint like this:
>
> <rsc_location id="ms-drbd-srv:connected" rsc="ms-drbd-srv">
>  <rule id="ms-drbd-srv:connected:rule" role="master" score_attribute="pingd">
>  <expression id="ms-drbd-srv:connected:expr:defined" attribute="pingd" 
> operation="defined"/>
>  </rule>
> </rsc_location>
>
> It's supposed to force the master of the drbd resource onto the machine
> which has the best connectivity to its ping nodes. This works nicely: The
> resources will fail over to the node with the highest pingd attribute
> value.
>
> But: Whenever a ping node changes status on all cluster members
> simultaneously (for example because of a switch reboot) all resources
> are restarted once.
>
> Is this expected behaviour?

Not really.

> Can it be prevented?

I'd have to see it in action.  Can you create a bug and attach a
hb_report archive covering the period when this happens?
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

Reply via email to