[Linux-cluster] Strange CMAN error

[Jeff Sturm]

What might cause a message like:
 
Jan 12 08:41:24 t0core-mqc02 openais[1716]: [CMAN ] Node 8 conflict,
remote cluster name='t0core-inner-rhcxvm', local='t0core-inner-rhc'

I've double- and triple-checked that /etc/cluster/cluster.conf is
identical on every node.  It starts with:

<?xml version="1.0"?>
<cluster name="t0core-inner-rhc" config_version="3">
  <fencedevices>
    <fencedevice name="xvm" agent="fence_xvm"/>
  </fencedevices>

It's as though cman is concatenating the two attributes somehow.
Coincident with this, "cman_tool nodes" says the node 8 is not a member
(though node 8 thinks it is).

We're having trouble with our cluster when fencing/rejoining a node.  It
operates fine when starting the whole cluster together.  This is on
CentOS 5.2.

Thanks for any insight...
 
Jeff