[Pacemaker] 2-Node Reconciliation...?

[Andrew Beekhof]

On Tue, Jun 22, 2010 at 8:06 PM, Eric Cho <echoecho at gmail.com> wrote:
> Hey everyone
>
> I've been attempting to find an answer to my question, but I've been coming
> up with nothing. If this has been asked before, I apologize. I would
> appreciate any guidance that you can provide.
>
> I have a 2-node cluster running pacemaker 1.0.8/corosync 1.2.5 on SLES11. I
> have 3 resources configured: IP, filesystem & oracle. The filesystem is
> coming from a fibre SAN with an ext3 filesystem. Everything works
> wonderfully and I've been very satisified with the setup.  Migrating
> resources works correctly and turning off the corosync service starts the
> service on the other node. However, my question involves the behavior of the
> cluster when the active node goes down.
>
> When I disconnect the network cable on the active system (AS), the AS does
> not disable the configured resources that it is responsible for. The
> secondary system (SS) sees that the active node is down, and starts to take
> responsibility for the resources. Nothing is 'wrong' in that i don't get any
> error messages, but i am a little concerned about this behavior, as i
> effectively have 2 active systems (even though only 1 is technically
> accessible).
>
> My basic question is this: is this normal?

Yes.  You've created a split-brain condition.
This is why you need fencing enabled.

> Secondary question: when both systems are 'active', both are mounting the
> filesystem (an ext3 partition), but ext3 is not a cluster filesystem. Should
> i change this and use something like ocfs?

Without fencing, this wouldn't help much.

   http://theclusterguy.clusterlabs.org/post/113230399/highly-available-data-corruption

> I appreciate any information that you can provide. Thanks!
>
> ---
> Grace and peace...
>
> Eric Cho
>
> _______________________________________________
> Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>
> Project Home: http://www.clusterlabs.org
> Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
> Bugs:
> http://developerbugs.linux-foundation.org/enter_bug.cgi?product=Pacemaker
>
>