Red Hat

Hi all
I am working on a scenario to safely mount a SAN volume from a RHEL 5 host (HOSTA) to another RHEL 5 host (HOSTB) in case if HOSTA crashes.I have used LVM vgexport and vgimport utilities which are not helpful in case of system crash.
Everytime you have to deactivate volume groume on HOSTA and manually or by using some automated scripts export the volume group to HOSTB.Following link can explain it more well:
http://kbase.redhat.com/faq/docs/DOC-4124
Any information on this scenario wil be highly appreciated.

I think what you really want to do is a cluster with shared file system. You can use VERITAS or RedHat cluster to do this. I think there is free version of Veritas SF that allows you to mount one file system, not sure if cluster is included in this. And you realize that KB link is talking about physically moving a local disk, not a SAN disk. The disk on the SAN would still need to be zoned to both hosts.

Thanks for the reply. We don't want to use any cluster software. Just wants to mount the SAN volume to the new failover node. Yes, although KB link talks about local disk, but I was successfull in testing it for SAN LUN.

You may not think you "want to" but the inconsistencies you may experience with this, at the most inopportune times, will be a possibility. This is one of the very reasons that a clustered file system exists.


Have you thought, perhaps, instead of just using drbd?

Mark

running these commands on hostb after shutting down hosta solved my issue.

Pvscan
vgacan

vmimport -f $vggroupname

thank you guys for your time.

The rule to do a "poor man's cluster" is to not use the same names of vg..
Since the LVM comes from HP, if it works the same way, you need knowing how to use vgimport vgexport vgchange, there is no need to vgexport each time when changing server (which means you need to vgimport again etc...) once both boxes are set, you only use vgexport/vgimport if you have changed disks configuration... otherwise its useless:
How can you vgexport on a crashed system?
You use vgchange! to activate/deactivate a volume group.
One thing to remember a true cluster has a special mode E (exclusive), you have nothing of the sort and in case or error (mounted both side) you loose everything (trust me -I tried for you long ago...)

Vbe

I am not able to understand you.My appologies.
My method is right or wrong.
I specified earlier as I don't need to setup any cluster.
My scenario just talks about using a failover server if primary crashes. I won't be using same volume on primary and failover server same time.