/etc/lvmtab entries for two node cluster

[Adam Noble]

Hi,

Can somebody help. I have a two node VCS cluster with a very simple configuration. There is one external volume group which is vg01 and in the event of a failure it obviously is exported onto the second node. The concern I have is that at present the second nodes /etc/lvmtab file only has disk entries for vg00. There are device file entries for vg01 on the 2nd node. The volume had been previously exported to the second node and we have failed over many times without issue. My concern is that other clusters we have seem to have entries for the de-activated volume groups on the failover node in /etc/lvmtab whereas this cluster does not. Also when you do a vgdisplay on these other clusters you get information for the de-activated volume groups simply saying they are not presently active on that system. This again is not the case on this cluster. Is there a problem here?

[Fabio Ettore]

Hi,

two possibilities I see here:
1) vgexport of vg01 on the second node was launched;
2) vgimport of vg01 on the second node was not launched.

In any cases just

'vgexport -p -m ...' from the first node
copy mapfile on the second node
'vgimport ...' on the second node

After that you should see /etc/lvmtab populated with the vg01.

If you can then take a test to switch the package on the second node.

HTH.

Best regards,
Fabio

[Adam Noble]

Its ok I've worked out what has caused this. I had to back out some OS patches recently on the server and restored from ignite. I assume the lvmtab is actually created rather than restored and therefore the entries are no longer there. Obviously in a rather precarious position now so will need to schedule a change to export and import group onto server. Thanks anyway.

[Fabio Ettore]

Hi Adam,

thanks for the update.
Anyway the "vgexport -p ..." on first node and vgimport on the second node doesn't require any stops for Serviceguard packages, you can do it just now. Of course you should schedule a switch of the involved packages on the second node to check if all is ok.

Best regards,
Fabio