[Linux-cluster] Sharing GFS Partition

Lon Hohberger lhh at redhat.com
Wed Feb 6 17:36:02 UTC 2008


On Tue, 2008-02-05 at 16:43 -0500, Marcos Ferreira da Silva wrote:
> The cluster start but I couldn't mount.
> 
> The process fenced is started.
> 21399 ?        Ssl    0:00 /sbin/ccsd
> 21421 ?        Ss     0:00 /sbin/groupd
> 21429 ?        Ss     0:00 /sbin/fenced
> 21435 ?        Ss     0:00 /sbin/dlm_controld
> 21441 ?        Ss     0:00 /sbin/gfs_controld
> 21831 ?        Ssl    0:00 clvmd -T20
> 
> 

So, what's happening is that the node you have is trying to fence the
other node, but no fencing is configured, so it fails and retries
(forever!).  If you're running RHEL5 or from CVS/head or some other
recent version of cman, you can try this manual override to make fencing
complete:

while ! [ -e "/var/run/cluster/fenced_override" ]; do
  sleep 1
done
echo vserver2.teste.br > /var/run/cluster/fenced_override

... but if you are using Xen, the best thing to do is look here for
information about how to set up fencing so this doesn't happen in the
future:

  http://sources.redhat.com/cluster/wiki/VMClusterCookbook

Also, the FAQ article on CMAN may be helpful for understanding what's
going on:

  http://sources.redhat.com/cluster/wiki/FAQ/CMAN

-- Lon




More information about the Linux-cluster mailing list