halting cluster without halting package

[Quin Hammes]

We need to update the timing variables for the cluster but we do not want to bring down the package. Is there a way to bring down a cluster without bringing down the package (or just the applications which are part of the package) so you can update timing variables? I could just edit the stop and start scripts so it does not do anything and then bring down the cluster but I don't want the volume group deactivated or anything. Any suggestions? We are not running a monitorig script nor are we using using any services. Any help would be appreciated.

Quin

[Carsten Krege · Accepted solution]

This won't work. You could replace the package control script with a simple script that does nothing else but "exit 0", leaving the package up and running when running cmhaltpkg.
But you still wouldn't be able to halt the cluster while the VG is activated in exclusive mode (vgchange -a e), what is the normal case.

Halting the node, with the package VG still activated, would give you:

"Clustered Volume Groups must be deactivated before the cluster can be halted."

Since the VG can only be deactivated when all filesystems are umounted, you need to halt your application first, before you can do this.

BTW, what timing parameters do you want to change and why?

Carsten

[melvyn burnard]

What you are looking to do cannot be done.
If you try to force the packages to stay up, i.e. VG's are activated, then the cluster will not halt. SG is not designed to work in this way. You will need to halt the cluster, make your changes and then restart the cluster and packages.

[Rita C Workman]

The experts have given you the answer...
The package is part of the cluster. You can not stop the whole, without the parts of it being stopped as well.

Rita

[Vincenzo Restuccia]

#cmhaltpkg
#vgchange -c n /dev/vgcluster
#vgchange -a y /dev/vgcluster
#mount /dev/vgcluster/lvolx /mount_point

[Vincenzo Restuccia]

sorry #cmhaltpkg -> cmhaltcl

[melvyn burnard]

And by the time you have done the last two suggestions, you could have reapplied th ebinary and restarted the cluster !