Serviceguard
cancel
Showing results for 
Search instead for 
Did you mean: 

Migrating Serviceguard cluster to RHCS Cluster

 
Eli_pok
Frequent Advisor

Migrating Serviceguard cluster to RHCS Cluster

Hello,

I don't know if it's the right forum to ask about problems with RHCS but i'll give at a try.

I follow a documnet by HP (can be found on this site) named :
"Migrating an HP Serviceguard for Linux Cluster to Red Hat Cluster Suite in Red Hat Enterprise Linux 5 Advanced Platform"

I also read the documnet from Red-Hat site:
"Using SCSI Persistent Reservation (SCSI Fencing) on Cluster Suite in Red Hat Enterprise Linux 4/5"

I've got 2 blade servers(IBM) with RHEL 5.4,
each one of them connected to the same 2 LUNS (SAN - NetApp machine) one of them is the quorum disk, the other is the shared disk for application (with a clustered vg - managed by clvmd).

My cluster is based on the configuration described in pages 15-19 , using scsi fencing.
(One service for mounting a fs on lvol on the clustred vg)

My probelm is that when cluster is online, I push the restart button of one server, and the service does not relocate to the other server, although (accroding the the log) the server was detected as offline and was fenced successfully (with no errors as i can see)

If I restart the server throught the OS (shutdown -ry now) it works fine.

Any help will be appriciated..

3 REPLIES
Matti_Kurkela
Honored Contributor

Re: Migrating Serviceguard cluster to RHCS Cluster

> If I restart the server throught the OS (shutdown -ry now) it works fine.

When you run "shutdown -ry now", the startup/shutdown scripts will cause the server to exit the cluster in a controlled manner.

> ... when cluster is online, I push the restart button of one server, and the service does not relocate to the other server, although (accroding the the log) the server was detected as offline and was fenced successfully (with no errors as i can see)

The important question is: after you've pressed the reset button on one cluster node, does the "clustat" command report "Member Status: Quorate" on the other node?

If it says "Member Status: Inquorate", then something is wrong, probably with your quorum disk set-up. If the quorum disk won't supply the vote(s) required to maintain quorum when one of the nodes is gone, the remaining cluster member will be inquorate: it cannot activate any services, so service relocation won't happen.

Could you please post your /etc/cluster/cluster.conf file (perhaps with the service and resource definitions omitted)?

MK
MK
Eli_pok
Frequent Advisor

Re: Migrating Serviceguard cluster to RHCS Cluster

>When you run "shutdown -ry now", the >startup/shutdown scripts will cause the >server to exit the cluster in a controlled >manner.
Yes, I know. I only tried to mention it
basically works by wititing that.

And for the important qustion: Yes, the cluster remains in "Quorate" status after I restart one of thr servers (thanks to the qourum disk I think).

But anyway, It seems the problem was solved!

I updadted these 3 RPMs to their latest releases from a RHEL5.5 installation cd: cman,rgmanager and lvm2-cluster.
(The difference was only by few releases, the version was the same. Somehow, It made the difference).

I've changed nothing in cluster/server configuration, and it worked!

I've even managed to pull a blade out of its place while the service was working on it, and the service relocated to the other node successfully.
Eli_pok
Frequent Advisor

Re: Migrating Serviceguard cluster to RHCS Cluster

"Solution" above^