Zimbra not playing nice with Pacemaker

[BCamp]

I'm setting up a new installation of Zimbra that is going to have 2 nodes for high availability. I'm using Heartbeat and Pacemaker as my cluster monitoring stack, and DRBD as my network replication. I've got a happy Zimbra installation, and a happy cluster configuration. When kept separate, they both work fine. I'm having a bit of trouble integrating Zimbra to the cluster. I'm using the regular init script for Zimbra, since there is no OCF script that I can find anywhere. As far as I can tell, the Zimbra init script is LSB compliant, as described here.

Appendix*G.*Is This init Script LSB Compatible?

I *think* the problem is Pacemaker is timing out before Zimbra can finish starting, which causes a failure. The timeout is somewhere in the neighborhood of 20 seconds. LSB resources can't be given any parameters in Pacemaker like OCF can, they're just run "as is". I've tried adding "op timeout=360s" to the Zimbra resource, but that didn't change anything.

It works fine if I let it time out and fail and then do a zmcontrol startup. The crm_mon still sees it as stopped, so I'm thinking it times out and then stops monitoring it.

Is anybody here familiar with this sort of setup and can help me out? I'll post configs and logs if so. If not anybody on the forums, does tech support help with this sort of stuff, or will it be brushed away as a Pacemaker problem? We've got a support contract and I don't mind using a ticket for this, but I don't want to waste one if it's something they're not going to help me with.

[bdial]

zimbra on drbd is not supported, so you will most likely not get any help from zimbra support. theres only one thread i've seen that seemed to have any value for this and thats

[SOLVED] Zimbra on DRBD

[Searcher]

BCamp or anyone else, have you been able to get Zimbra working with Corosync and Pacemaker?

[adrian.gibanel.btactic]

I am also trying to making it work, not sure if I will post the final result because our setup is too specific, doc is being written in Spanish and other things. If I ever do it I will use the Solved DRBD post.

But I can tell you telling you what I am going to do to try to solve this same problem.

According to OCF Resource Agents you have to write a proper OCF agent. Reusing an LSB script as an OCF is not right. I mean old heartbeat used LSB scripts but pacemaker no longer uses them.

It would seem that you are lacking the monitor parametre.

So what I am going to do is to copy the:

/usr/lib/ocf/resource.d/heartbeat/postfix

file into this new file:

/usr/lib/ocf/resource.d/btactic/zimbra

.

Then using a find postfix string and replace it with zimbra string might help a bit. But be sure to read the Debugging part of the link I have given you.

And the other part of the setup should be this one. When setting up pacemaker with crm instead of using:

ocf:hearbeat:postfix

you should use:

ocf:btactic:zimbra

.

As I said this is how I understand all this stuff but I haven't tested it yet.

Just hope that it helps someone else.