[Pacemaker] Odd behaviour produced by resource migration

Andrew Beekhof andrew at beekhof.net
Thu Feb 3 07:36:34 UTC 2011


On Wed, Feb 2, 2011 at 3:37 PM, Dominik Dingel
<dingel at hep.physik.uni-siegen.de> wrote:
> Hello there,
>
> I'm currently setting up a two node cluster with:
> - dual primary drbd
> - dlm
> - o2cb
> - ocfs2 filesystem
> - openvz container
> to be a active-passiv cluster solution.
>
> Stonith is configured to use IPMI and meatware as fallback, so even in
> case where a server would be removed, the remaining node could be powered
> on with little effort.
> So now I'm testing all scenarios of the cluster. The migration of the
> container is one thing which does not work as I would like it to work.
>
> After a fresh power-on of both nodes and so the cluster came online,
> resulting in this output:
> http://pastebin.com/4Hdz5bjy
>
> After issuing then the line:
>> crm resource migrate resVZ
> to move the container from node2 to node1 all services on node2 will be
> stopped as msDRBD on node2 is demoted.
> http://pastebin.com/yxgBkYTf
> I don't understand why the resource (drbd) is demoted in the first place
> and if there is a way to prevent this.

It was a bug that was fixed by:
   http://hg.clusterlabs.org/pacemaker/1.1/rev/7a4fd2dd2bdd
and appears to have been backported as:
   http://hg.clusterlabs.org/pacemaker/stable-1.0/rev/5fe02f48c47b

which should have meant it was fixed in 1.0.10 - but this does not
appear to be the case.




More information about the Pacemaker mailing list