Hello,

Try running pvscan && vgscan && vgmknodes.

This should re-create the vg-dir under /dev, as you’re forcing it to create the nodes (i.e /dev/myvg/lvol0, etc) without the need to stop clvmd.

I haven’t tested this, but it should work and only be required the first time you make changes to the lvm.

Cheers,

I'll probably be opening a support ticket with RedHat on this today. I can't restart clvmd every time I need a new LV, as it requires umounting all the active LVs in each clustered VG.

Thanks for confirming!

Kevin

From: rhelv5-list-bounces@redhat.com [mailto:rhelv5-list-bounces@redhat.com] On Behalf Of Zavodsky, Daniel (GE Money)
Sent: Sunday, April 27, 2008 11:58 PM
To: Red Hat Enterprise Linux 5 (Tikanga) discussion mailing-list
Subject: RE: [rhelv5-list] Clustered LVM

Hello,

    I have been experiencing exactly the same issues... and I found the trick with restarting clvmd myself too. :-) It looks like the problem happens when you are using both non-clustered and clustered LVM.

    I don't think you are doing anything wrong.

Regards,

    Daniel

From: rhelv5-list-bounces@redhat.com [mailto:rhelv5-list-bounces@redhat.com] On Behalf Of Collins, Kevin [Beeline]
Sent: Friday, April 25, 2008 9:08 PM
To: rhelv5-list@redhat.com
Subject: [rhelv5-list] Clustered LVM

Hi,

        I am working through the process of building my first GFS installation and cluster. So far, things are going well. I have shared SAN disk that is visible to both nodes, I've got the basic cluster up and I am now starting to create my LVM structure.

I spent a lot of time trying to determine why I was seeing errors similar to the following when trying to create LVs in a new VG:

root# lvcreate -L 50M /dev/vgtest
  Rounding up size to full physical extent 52.00 MB
  Error locking on node cpafisxb: Volume group for uuid not found: IC070PzNGG68uEesi33dH902E4GeGEvwcAccZY4AdnJRkPHUL7EYJzL0Xxkg3eqV

  Error locking on node cpafisxa: Volume group for uuid not found: IC070PzNGG68uEesi33dH902E4GeGEvwcAccZY4AdnJRkPHUL7EYJzL0Xxkg3eqV

  Failed to activate new LV.

I noticed that I did not have a /dev/vgtest on the node where I created the VG. I discovered (by googling the error message above) that another person had seen a similar problem and resolved it by restarting the clvmd service. I did the same thing and the problem was resolved - I don't see that error when creating LVs in that VG, and I now have a /dev/vgtest and associated files (only on the node I created it on).

Now, I created a new VG, try to create a new LV within it and see the same error! I see in the clvmd man page that there is a "-R" option, which sounds like maybe what should be done instead of restarting clvmd. Tried it, but no luck - same error and no /dev/ directory created. Restart clvmd, instant fix.

So - what am I doing wrong here? I can't imagine I am supposed to restart clvmd every time I create a new VG, am I?

Additionally, I can see the clustered VGs and LVs on my other node, but there are not any files in /dev for the clustered VGs - is this normal? I have also restarted clvmd there with no effect.

Thanks,

Kevin