topic Re: cluster fencing in Operating System - Linux

cluster fencing

nasmel — Mon, 19 Mar 2007 01:54:04 GMT

Dear Experts

I have 2 DL 580 nodes cluster,each has one service and fencing is configured through ILO.

when i disconnect the node2 serice is switching over to node2 and when the node1 is available its coming back.

but when i disconnect the node 1, node 1 is showing that node 2 is down and node2 shows that node1 is down and i could nt ping to the service.
after reconnecting node1 its not joining to the cluster also

what will be the problem

#uname -r
2.6.9-42.ELsmp

when i tries to stop the cluster service its taking long time and i had to rebbot it mannually(rgmanager and cman).and i tried to start the fenced its also taking long time

please advice
Thanks

Re: cluster fencing

Steven E. Protter — Mon, 19 Mar 2007 04:00:26 GMT

Shalom,

A new release of clustering and the 2.6.9-42.0.3. may alleviate this issue.

Having the iLo firmware up to date is another good precaution. There are known bugs with previous versions of the kernel in concert with rhcs.

SEP

Re: cluster fencing

nasmel — Mon, 19 Mar 2007 05:11:32 GMT

dear.

I was using the 42.0.3 kernel and i was unable to install the cluster packages and i downgraded the kernel to 2.6.9-42.EL
then it allowed me to install the cluster packages

thnaks

Re: cluster fencing

nasmel — Mon, 19 Mar 2007 07:33:36 GMT

Reconfigured the ILO ,its working now

Thanks

Re: cluster fencing

Steven E. Protter — Mon, 19 Mar 2007 07:40:16 GMT

Note:

You have to update clustering at the same time as you update clustering.

The latest RHCS release DOES support the .03 kernel. I'm running it in clusters all around the world.

Note also that I did advise you to check the ilo configuration which apparently SOLVED the problem.

Good Luck,

SEP