[Linux-cluster] Problem with cluster nodes communication

[Arnaud Gauthier]

Status:

I have reduced my cluster to 2 nodes,
Cleaned up all previous install of gfs,
re-compiled and installed kernel 2.6.9,
re-compiled and installed gfs

Still the same problem. Is it because I am using eth1 for cluster 
communication ? I was using broadcast before on eth1.

Here is my cluster.conf:
<?xml version="1.0"?>
<cluster name="tfsm" config_version="1">
  <cman>
  </cman>
  <nodes>
    <node name="l1_oas56_a" votes="1">
      <fence>
        <method name="MCdata">
          <device name="switch" port="6"/>
        </method>
      </fence>
    </node>
    <node name="l1_oas56_b" votes="1">
      <fence>
        <method name="MCdata">
          <device name="switch" port="7"/>
        </method>
      </fence>
    </node>
  </nodes>
  <fence_devices>
    <device name="switch" agent="fence_mcdata" ipaddr="192.168.10.208" 
login="Administrator" passwd="xxxxxx"/>
  </fence_devices>
</cluster>

Any idea ? I am lost...
-- 
Arnaud Gauthier
247 Real Media