[Pacemaker] Location constraints when dealing with multi-site clusters

[Dejan Muhamedagic]

Hi,

On Mon, May 26, 2014 at 02:56:18PM +0200, Christian Ciach wrote:
> Hello,
> 
> I am dealing with a three-node multi-site cluster using pacemaker and
> corosync. The third node is a quorum-only node.
> 
> The three nodes form a single cluster and are connected using corosync over
> the internet (using VPN tunnels). Of cource, using corosync over the
> internet and using a third node with the full corosync/pacemaker stack
> installed doesn't seem to be the ideal solution.
> 
> 
> So I took a look at the booth daemon, which, at first glance, seems like a
> better alternative. Unfortunately, because I would have to split my single
> cluster into two local ones, I don't see a way to still use location
> constraints for my resources. For my use case it is important to always
> migrate the master resource based on a calculated score. This works great
> with single clusters.
> 
> Am I missing something or is this indeed a disadvantage when using booth?

booth has been designed with quite a different purpose in mind.
A ticket has only two states* and depending on that a resource
may be allowed to start (or promote) on the cluster or not. A
ticket fails over to another site in case the existing site is
not reachable anymore. Or the user defined external program
fails. Perhaps you could fiddle with the latter, but I doubt
that it'd be of any use in the end. So, the short answer is that
booth isn't meant for what you want to achieve here.

Thanks,

Dejan

*) Apart from granted/revoked there's also standby, but that's
not relevant here.

> 
> Best regards,
> Christian

> _______________________________________________
> 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