cmclconfig checksum different

[Jeff_Traigle]

Have had this situation on a couple of clusters. Two node clusters running HP-UX 11.31 and SG 11.19. Everything in the cluster appears to work fine. cmviewcl shows proper output and everything up and running on both nodes. Failovers work without a hitch. Systems rebooted for patching and come back into the clusters with no problems. The timestamps and sizes match exactly, but the checksums on /etc/cmcluster/cmclconfig are different. My understanding was that these should always be identical files and that SG managed copying them. We have an audit script that runs to check various things before a planned failover, and this keep popping up as an error. We "fixed" one of the clusters by simply copying the cmclconfig file from one node to the other manually after halting cluster services on the receiving node. Just wondering if anyone else had run into this situation and if there's a better solution.

[tom_hutter]

During a project I had to set up ccmon for several clusters. Which ccmon we also checked the nodes of each cluster for identical checksums of cmclconfig and experienced  different checksums on several nodes. I was able to contact HP and got the following answer:

"Differences in checksum of cmclconfig on several nodes of the same cluster are no sign for an error. In environments with SGeRAC it is even possible that the cmclconfigs on the nodes differ in size."

You can run "cmcheckconf -k" on each node to verify your configuration (see man page: ... the cmcheckconf command will compare the configuration in the cluster_ascii_file against the previously  configuration information  stored  in the binary configuration file and validates the changes.)

If this succeeds everything is fine.

Cheers

Tom