Quick Recovery Guide?

[SnIphe]

Hi all,

I`m looking for a Quick Recovery Guide for Service Guard... something that tell you how to make a good backup of your SG cluster.

So if you lose the cluster, for some reason, you can recover it quickly.

Another question. If you lose the LockLun?? What is the response of the cluster? How can you recovery this cluster?

I have installed on SG cluster yesterday, and all is working OK. My customer wants to make a full backup, for security reasons.

Thanks a lot.

[Serviceguard for Linux]

On the question of back-up, are you looking to back up the entire server or just the Serviceguard info. For the Serviceguard info, refer to the Release Notes and the section on rolling upgrades should tell you what files to back up.

You should be able to use your normal backup software to backup the entire server. When backing up shared storage, you need to be aware of which VG is mounted on which server.

If you lose the Lock LUN the cluster will keep on running. You will get some messages in the syslog. In a two node cluster, if a server fails and the Lock LUN cannot be accessed by the other node, then the cluster will go down. In this dual failure case this is done to cover the case when heartbeats are lost but both servers are still running. This is a potential "split brain" scenario and tkaing down the cluster prevents data corruption.

[Carsten Krege]

To backup the SG configuration, you create a tar archive of the cmcluster (Suse /opt/cmcluster, RHEL /usr/local/cmcluster) directory. If you have a proper backup of your nodes, you save this anyway.

If your lock LUN fails, nothing happens but SG complaining. The lock LUN is only used if there is a failure of exactly 50% of the cluster nodes. It is not used during cmrunnode/cmhaltnode/cmruncl/cmhaltcl. If it fails, you should repair it as soon as possible. To reinitialize it you require a cmapplyconf to be run.


With SG/LX A.11.18 (probably July) a new command will be available to reinitialize the lock LUN online.

Carsten

[skt_skt]

If you Lose lock LUN, nothing will happne to the running cluster.Lock disk is used only during cluster reform operation.You can replace the disk if possible with the same ctd number.