[Pacemaker] Storage Locking To Prevent Split Brain

[Andrew Beekhof]

On Thu, Oct 18, 2012 at 6:40 AM, David Morton <davidmorton78 at gmail.com> wrote:
> We're changing our SAN shortly and I'm putting together the procedure /
> config now for the shared storage. This will be based on XFS on top of
> clustered LVM2 via Pacemaker.
>
> I've implemented the exclusive=yes directive on the LVM resources (volume
> groups) but I am still able to mount on both cluster nodes (2 node cluster)
> in my testing environment. I would have thought that this directive would
> prevent multiple mounting / VG activation given that even when there are no
> cluster comms (this is how i create the split brain, and reboot both nodes)
> the storage is available to both nodes ?
>
> What is the recommended way of guaranteeing storage exclusiveness in this
> (albeit unlikely) situation where both nodes think they are running by
> themselves ? I've looked at sdb and sfex but have previously been steered
> away from them.

Perhaps look at the Red Hat OCF agent for LVM (yes, Pacemaker can use
those too :-)
I believe they implemented the exclusivity requirements you're after.

I believe the existing Linux-HA OCF LVM agent relies on Pacemaker to
enforce the "active on only one location" property.

>
> Dave
>
> _______________________________________________
> 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://bugs.clusterlabs.org
>