Cannot start cluster on rhel 4 with ZCS 5.0.5

[aberkram]

We're trying to install ZCS 5.0.5 on a single-node cluster with two RHEL 4 servers in an active/passive configuration and can't seem to get cman to start correctly. ccsd seems to start just fine but then kicks out a few errors in the log. This is what /var/log/messages looks like in relation to our current situation. We're definitely newbies in the realm of clustering with RedHat so any help would be appreciated.

Jun 9 17:14:59 mail1 ccsd: succeeded
Jun 9 17:15:02 mail1 cman: FATAL: Module cman not found. failed
Jun 9 17:15:28 mail1 ccsd[4446]: Unable to connect to cluster infrastructure after 30 seconds.

Also, if anybody has any docs on IPMI fencing with RHCS clustering we'd appreciate a link to those as well. Thanks!

[jhahm]

Hi, you may want to try reinstalling the RHCS rpms. What is your RHEL4 Update level? If you recently updated RHEL make sure you also update RHCS rpms to the matching Update level. If you have a custom kernel that may also explain why cman kernel module is missing.

[Klug]

Do a full "update -u" in order to get all the last updates and double check you have setup the correct RPMs (the whole RHCS suite).

The cman module should load itself automagically if the correct kernel is installed.

[aberkram]

This is a new install of RedHat and is dedicated for Zimbra. Linux kernel is 2.6.9-67.0.15.ELsmp. I believe we're running Update 6. When we run up2date it says all of our rpms are updated. Is there an easy way to reinstall RHCS rpms? Sorry for my ignorance, but I'm very new to Linux. Also, could our problem be related to ccsd or is that error a result of cman not starting? Does cman have to start before ccsd will connect to the cluster infrastructure?

[jhahm]

I'm not much familiar with up2date, but try --channel option to make sure you're refreshing the cluster suite software too. Your kernel may be newer than the kernel modules of RHCS. You can get the latest cluster suite rpms from Red Hat Network site. My test box has kernel 2.6.9-67.ELsmp, which is slightly older than yours. My box also has these cman-related rpms:

cman-1.0.17-0
cman-kernel-smp-2.6.9-53.5
cman-kernel-2.6.9-53.5

As you can see the kernel versions don't match up exactly. The smp kernel module is installed but not used because I'm not running that kernel.

As for order of ccsd/cman execution, ccsd runs first then cman. You don't have a problem there. The error message on your machine implies you either don't have the cman-kernel rpm installed or its version it out of sync with kernel version.

I suggest you go to the Red Hat Network site to check on your rpm versions.

[2myhre]

We got clustering working. Thanks for the help. Updating to match the kernel level was the key. Subscribing to the cluster channel allowed the up2date to update the cluster modules. I am attempting IPMI fencing now. When one goes down they try to fence each other and the operation hangs. Any suggestions?