[Linux-cluster] Fencing node automatically...

[Digimer]

On 12/30/2011 01:24 AM, SATHYA - IT wrote:
> Hi,
> 
> On my two node cluster which got configured inRHEL 6.2 (with ctdb + gfs2
> + rgmanager + cman), primary node fences the secondary node
> automatically ( in random). During this fence the message appears in the
> primary server is as follows:

That syslog snippet should have had some important bits before it.
Please re-paste the log files from that time period from both the victim
and the surviving host, starting from the first log message in the same
time period as the fence event, clear through to the surviving node has
restored services.

Also, please past your full cluster.conf file, and please only obfuscate
passwords only. Likewise, please paste your full network configuration.

With that said; You're probably facing transient network issues. Are
there any logs in your switches around the same time period? Is
multicast working properly? Is STP enabled and perhaps causing a
temporary block in network traffic?

The cluster certainly doesn't break randomly, there has to be a cause. :)

-- 
Digimer
E-Mail:              digimer at alteeve.com
Freenode handle:     digimer
Papers and Projects: http://alteeve.com
Node Assassin:       http://nodeassassin.org
"omg my singularity battery is dead again.
stupid hawking radiation." - epitron