[Pacemaker] Quorum settings 2-node cluster

[Tim Serong]

On 2/14/2010 at 02:07 AM, Sander van Vugt <mail at sandervanvugt.nl> wrote: 
> Hi, 
>  
> I have a small test cluster, using two nodes with SLES 11 + SLES HAE, 
> completely patched up to date. I've noticed that if one of the nodes 
> goes down, the cluster looses quorum (crm_mon -i 1 on the remaining node 
> shows 
>  
> Last updated: Sat Feb 13 16:04:35 2010 
> Current DC: node1 - partition WITHOUT quorum 
> Version: 1.0.3-0080ec086ae9c20ad5c4c3562000c0ad68374f0a 
> 2 Nodes configured, 2 expected votes 
> 3 Resources configured. 
>  
> This sounds weird, as the no quorum policy is set to "stop" (which does 
> mean that resources will be non-operational and hence my nice APC 
> rackpdu that is configured as the STONITH device cannot do it's work. 
>  
> Am I missing something? 

With OpenAIS clusters, you can only have quorum if more than half the
nodes in the cluster are online, so in a two node cluster, if one node
goes down, you always lose quorum.  If no-quorum-policy is set to "stop",
resources on the surviving node will be stopped.  If you want the
surviving node to attempt to STONITH the other node and take over its
resources, you need to set no-quorum-policy to "ignore".

Regards,

Tim


-- 
Tim Serong <tserong at novell.com>
Senior Clustering Engineer, Novell Inc.