[Linux-cluster] Heartbeat tolerance on busy networks (RHCS3)

[Karl Podesta]

Hi there, 

I'm looking to increase the tolerance level for heartbeats in a RHEL 4 U4
cluster. The network gets unavoidably busy at a single point each day. 
Short of seperating heartbeats onto a seperate network, is there any way 
I can increase the tolerance to prevent fencing due to missed hearbeats
during this busy period?

I've tried increasing deadnode_timeout, using the following in cluster.conf:
<cman deadnode_timeout="21">

(as per instructions in the FAQ for including it in the cman section, and 
for using cman_tool to update versions & transferring it between nodes)
... but unfortunately this has given me configuration errors when I try
to restart the cluster services so I'm forced to take the option out. 

Any ideas on what I'm doing wrong, or alternative recommendations?

Thanks!

Karl

--
Karl Podesta
Systems Engineer, Securelinx Ltd., Ireland 
http://www.securelinx.ie/