cmhaltnode fails ***URGENT!!***

[James Lanaway]

Hi,

I'm trying to shut down a ServiceGuard node and I get the following message in syslog...

Apr 2 03:21:24 erp03 cmlvmd: Volume group vgktnats is still active.
Apr 2 03:21:24 erp03 cmlvmd: Shared LVM busy: cannot shutdown

vgktnats was removed a while ago - but obviuosly not correctly, and is not in the lvmtab so I can't do a 'vgchange -a n' on it.

Help - I need to shut this node down and reboot the server!

Thanks,

James

[G. Vrijhoeven]

Hi James,

Seems like MC/SG is not able to vgchange -a n vgktnats but you figured out that part.
Normally you get this message when there is one lvol still active. Could it be the cluster lock vg?
If you have the time you could wait for the timeout, if not i would make sure all applications/dbases are down and do a hard reset ( TOC ).
Mind not to start the cluster on this node before fixing this problem. ( /etc/rc.config.d/cmcluster AUTOSTART_CMCLD=0)

HTH,

Gideon

[Victor Fridyev]

Hi,

Check with bdf what file systems from this vg are still mounted. Halt appropriate packages or/and umount the fs manually.

Good luck

[James Lanaway]

Thanks Gideon

There's no cluster lock disk in this package... This was a package that was removed a while back.

I think I'm heading towards restarting the server with the node up. BUT... how would I fix the problem when I get the box back up?

Thanks for your help,

James

[James Lanaway]

Also thanks Victor...

But... there are no packages running on this server and the filesystem is not active - and won't activate/de-activate anyway because it's not in the lvmtab.

Any other suggestions??

[Karthik S S]

James,

I think you deleted the vgktnats directory under /dev and removed the lvmtab entries by running vgscan. I would suggest you to restore the VG by creating a directory under /dev, create a group file and import the VG configuration back to the system. Then deactivate the VG after which you should be able to halt the node.

-Karthik S S

[G. Vrijhoeven]

Hi,

MS/CG checks VGs in the ascii file,

Default /etc/cmcluster/cmclconfig.ascii The volume groups listed at the bottum are the vg's MC/SG will try a vgchange -c y on ( at start).
I am not sure if you can alter ( remove line,#cd /etc/cmcluster && cmcheckconf -v -C *.ascii&& cmapplconf -v -C *.ascii) this online. The check will give an error if it can not be done online.)

HTH,

Gideon

[Shaikh Imran]

Hi,
Not a Solution But...
Try using fuser -k on that VG.
Have you tried using cmhaltnode -f command.

Reg

[Brian Crabtree]

Have you tried issuing a vgexport of the volume group you want to remove?

Brian

[Stephen Doud]

Since you are going to reboot the box, try this:

cmsetsafety off

Continue (y/[n])? y

This will allow you to kill cmcld and cmlvmd on this server without forcing a TOC (panic/reboot). After that, you can "shutdown" gracefully.

.SD.