clvm + GFS troubles

[andremachado <andremachado@techforce.com.br>]

Hello,
I wiped everything and restarted from scratch. (hey, fence deb package is a pain to remove!!!)
This time I was knowing a bit more.
I spotted the problems to fence and clvm daemons init scripts.
Already posted my proposed solutions at the end of:
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=336259
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=400205
Now I have GFS working at my elementary "single node" cluster (with an elementary starting concept and much-work-needing yet /etc/cluster/cluster.conf below)
Tomorrow will grow the GFS cluster to 2 nodes and improve the fencing.
Also, found a *very* useful howto here:
http://xenamo.sourceforge.net/index.html
I will try with gnbd instead of drbd.
Please, note that the clvm init script needed some modifications as I posted at 
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=400205

Regards.
Andre Felipe Machado
http://www.techforce.com.br


elementary "single node" /etc/cluster/cluster.conf for GFS testing purposes only:

<?xml version="1.0" ?>

<cluster alias="sepae_cluster_1" config_version="3" name="sepae_cluster_1">
	<cman>
	</cman>
	<fence_daemon post_fail_delay="0" post_join_delay="3"/>
	<fence/>
	<clusternodes>
		<clusternode name="serpro-1440128" votes="1">
			<fence>
				<method name="single">
					<device name="manual_fence1" nodename="serpro-1440128"/>
				</method>
			</fence>
		</clusternode>
	</clusternodes>
	
	<fencedevices>
		<fencedevice agent="fence_manual" name="manual_fence1"/>
	</fencedevices>
	<rm>
		<failoverdomains/>
		<resources>
		</resources>
	</rm>
</cluster>