[Pacemaker] Pacemaker moving a sticky multi-state resource when another node is brought online

Andrew Beekhof andrew at beekhof.net
Thu Nov 1 01:03:54 UTC 2012


On Wed, Oct 31, 2012 at 7:39 AM, Tupja, Ravik <RTupja at broadviewnet.com> wrote:
> Hello,
>
>
>
> We have a multi-state resource called callp running in a 2 node cluster
> (node A and node B).
>
>
>
> In a customer deployment, we had node A in which callp was running as a
> master and in node B it was running as a slave.
>
>
>
> The following happened:
>
> 1 - Node B was placed in standby mode.  After this, callp (the multi-state
> resource), was continuing to run as master on node A as expected.
>
> 2 - Next, node B was placed in online mode.
>
> 3 - PROBLEM -> The cluster decided to move the master callp to node B by
> demoting callp on node A and trying to promote it on node B.  This is an
> undesirable operation.  The default stickiness level in our cluster is set
> to 99 to prevent such movement of resources.   I have confirmed that callp
> is using the default stickiness of 99.
>

Which pacemaker version?
This is a known problem in older versions but should be fixed in the
1.1.8 and 1.0.12

>
>
> Therefore, the cluster should not move callp when another node is brought
> online due to the stickiness being 99.  I am not able to reproduce the
> problem because it happens very rarely but I need to know of a way to
> prevent this from happening again for sure.
>
>
>
> I would appreciate any help.
>
>
>
>
>
> Ravik Tupja
>
> Sen. Software Developer
>
> silhouette Research and  Development
>
> Broadview Networks
>
> p: +1.613.280.2010
>
> e: rtupja at broadviewnet.com
>
>
>
>
> _______________________________________________
> Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>
> Project Home: http://www.clusterlabs.org
> Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
> Bugs: http://bugs.clusterlabs.org
>




More information about the Pacemaker mailing list