Red Hat Cluster fencing documentation
 

Hi.


I'm setting up clustering using Red Hat Cluster Suite, and are in need of good documentation of configuration strategies. In particular, I'm interested in documentation on (fencing) strategies in this scenario:

• Two nodes running application (DB2)
• No shared resources/storage (the nodes access separate disks etc)
• No network power switch

It seems like most of the information I find on clustering assumes that one uses network power switches, and that the nodes access shared storage. But this is not the case in my set up.


I hope someone knows of any good reading on the subject, or maybe have done this herself/himself and can advice me on how to proceed.


Regards,
kenneho

The rgmanager heavily relies on the fencing concept.

fence_manual is the most minimalistic form of all fencing agents. However it is not useable in productive environments.

You may write a bogus fencing agent that always returns true or introduces a selfkill via sysrq-trigger kernel panic.

But there are several questions to ask that also prevent productive use:

e.g.

- how do you determine if it is safe to failback a service?
- how do you make sure that the failed node does not corrupt the cluster communication?