Cluster Failure

[Robert DJ]

Hi,

We hav a 2 node cluster and failed due to disk problem but the 2nd node didnt automatically switch over. We had to manually switch to the 2nd node.

1. Need to know why the switch over didnt happen ?
2. Need to know whether there is any other thing which was missed in the config file?

Thank you in advance.

Regards,
Robs.

[Steven E. Protter]

Shalom Robs.

1. Look at the logs in /etc/cmcluster . Additional data may be in /var/adm/syslog/syslog.log
2. Check the package scripts and see if they were configured to correctly fail over the packages.

Answer to your #2. The only way to be certain is to configure and test.

SEP

[Geoff Wild]

package_switching might not have been enabled.....

do a cmviewcl -v

Check AUTO_RUN - should be enabled...

Rgds...Geoff

[Patrick Wallek]

What kind of disk failure? Why would you need to switch to a 2nd node because of a failed disk? If you have MC/SG installed, you should have everything protected via Mirroring at the least. You should be able to withstand a disk failure without switching.

If it was a data disk on the shared VG, what good would switching to the other node do? The disk is still failed.

I think more details of the failure and more explanation of what you expected to happen are in order.

[Robert DJ]

HI All,

What are the actually things that could allow the switch over to fail ?

Everything seems to be ok.

This is a SCSI Device failure to be precise, but that has been resolved.

The only worry for me is that when all things are looking good and the package switching is also enabled, it didnt switch over automatically but happened to do it manually.

Is there any other way of going about this.

Thank you.

Regards,
Robs.

[Cem Tugrul]

We may need the output of the command;
#cmviewcl -v

[Duncan Edmonstone]

If you take your car to a mechanic and say 'it broke down', but won't let him look under the hood, then you'll never know why!

Post some of the diagnostics and logs others have already asked for here. The most importnat logs we need to see are the syslogs from both servers at the time of the failure.

HTH

Duncan