[Pacemaker] stonith woes (ignore my other message with this title)

[Dejan Muhamedagic]

Hi,

On Fri, Feb 26, 2010 at 08:45:48PM +0100, Sander van Vugt wrote:
> Hi list,
> 
> My daily call for help with my stonith woes. This is a problem that I've
> been struggling with for a while and I just can't get rid of it. The
> basic configuration SLES 11 with HAE, configured for high availability
> of xen virtual machines, using clvm as the storage backend and sbd
> stonith to guarantee integrity. For those that are willing to have a
> look, the output of hb_report and supportconfig (with all the required
> documentation) are at http://www.sandervanvugt.nl/novellsupport 
> 
> The short situation description: I'm using sbd stonith and I've got a
> stonith request roaming around in the cluster (but not being executed).
> It gives messages like:
> 
> nd1 stonithd: ... tengine requests a STONITH operation RESET on node
> nd2.
> nd1 stonithd: info: we can't manage nd2, broadcast request to other
> nodes.

Looks like sbd when retrieving the list of hosts from the device
didn't find nd2. Though in that case the stonith resource
shouldn't have started. On monitor, it takes a list of nodes from
crm_node -l and then checks if they have slots on the device.

> At the same time, all my three nodes are unclean and no STONITH is
> happening at all.

How comes that all are unclean?

> So basically I have two questions now:
> *       Is there any way to get rid of the stonith action which
> shouldn't be there?

Normally, there's a good reason to fence a node. Why do you think
the requests are wrong?

> *	Is there any way to get my nodes back to a clean status?

Only by fencing or restarting the cluster. Alternatively, a node
may be deleted.

Thanks,

Dejan

> Thanks in advance,
> Sander van Vugt
> 
> 
> 
> 
> _______________________________________________
> Pacemaker mailing list
> Pacemaker at oss.clusterlabs.org
> http://oss.clusterlabs.org/mailman/listinfo/pacemaker