[Linux-cluster] Problem with cluster nodes communication
Lon Hohberger
lhh at redhat.com
Thu Oct 28 17:15:28 UTC 2004
On Thu, 2004-10-28 at 17:39 +0200, Arnaud Gauthier wrote:
> Le jeudi, 28 Octobre 2004 17.31, Jonathan E Brassow a écrit :
> > Note that a two node cluster is a special case, and you will need to:
> >
> > <Dave Teigland's message>
> >
> > ...following the info under "Two node clusters" in this doc?
>
> I am not really interested on still being alive with only one node running. My
> problem is that the first node can't see the second node even after a full
> re-install. I am still at the init level, not the usage :-))
Could you set your expected votes to 1, wait for a quorum to form, run
"cpt lock", and describe what happens (and any strange kernel
messages... ;) ). I've had a strange occurrence where broadcast
messaging is broken and I'm trying to help isolate it.
You could try switching to multicast as well; it worked for me:
<?xml version="1.0"?>
<cluster name="pretty" config_version="6">
<cman>
<multicast addr="225.0.0.12" int="eth0"/>
</cman>
<dlm>
</dlm>
<nodes>
<node name="node0" votes="4">
<multicast addr="225.0.0.12" interface="eth0"/>
<fence>
<method name="power">
<device name="wti" port="1"/>
</method>
</fence>
</node>
<node name="node1" votes="3">
<multicast addr="225.0.0.12" interface="eth0"/>
<fence>
<method name="power">
<device name="wti" port="2"/>
</method>
</fence>
</node>
</nodes>
</cluster>
-- Lon
More information about the Linux-cluster
mailing list