VG won't activate

[Doug Short]

Sirs,

I have a 2 node cluster that uses an EMC san for all disks. My customer updated some firmware on the EMC, which somehow changed some disk addresses. I've since exported the VG and re-imported to get the proper disk addresses in /etc/lvmtab, but when I go to activate the VG with 'vgchange -a e vg01' I get something to the effect of Quroum not present or physical path to disk invalid. So just for the sake of arguement, I do this 'vgchange -a e -q n vg01' and still I get the same arguement about the quorum not being present. Any thoughts? All disk paths appear to be valid and on line. Any help is appreciated, thanks.

[Deoncia Grayson_1]

How are you managing your disks? Are you using power path? If anything, you can check to make sure all disks are showing up claimed and not throwing any others errors in the syslog. If your using powerpath, then you can run powermt display patsh to make sure there no dead paths or powermt display dev=all to make sure you can see both paths to the disks.

[Doug Short]

Yes, I am using Powerpath to manage disks, and all links appear active and alive in the powermt display. I don't see anything that would lead me to believe that the EMC is experiencing difficulties.

[Deoncia Grayson_1]

another suggestion could be to make sure lvmtab file is current, you can move the current copy to for example lvmtab.old and run a vgscan -a to recreate the lvmtab and try to activate the volume group.

[Mark Nieuwboer]

Doug,

Probebly you can activate the volume group with vgchange -a y vg01.
If this is not working .
move the lvmtab to lvmtab.old and do a vgscan.
Futher more you use MC so you have the change your cmcluster.cnf. After you start the volume group you have to do a vgexport -p -s -m mapfile and ftp the map file to the second node and do a vgimport there.
see also this doc. http://docs.hp.com/en/B3936-90073/ch05s04.html#d0e10090
Because al MC volumegroups has to be the same.


grtz. Mark

[Arunkumar.B]

Hi Doug,

I hope its a cluster setup.Pls try to do this


1)halt the cluster
2)check the cluster lock device file is correct if its got changed , replace with new device file
3)Activate the vg which has cluster lock
4)cmcheckconf & cmapplyconf
5)Then start the cluster & Package


This might help you..

[Geoff Wild]

"I've since exported the VG and re-imported to get the proper disk addresses in /etc/lvmtab"

Do you know how many disks were in the vg prior to the change?

Is the same amount in the imported vg?

Just do a strings on /etc/lvmtab.

BTW - I usually make a backup of the /etc/lvmtab file prior to doing any vgexports/imports.

Also, for all the disks that are in the vg now, do a diskinfo on the /dev/rdsk/c....

Do you get the expected result?

Rgds...Geoff

[Marvin Strong]

if you are getting quoroum then it sounds like you are missing some physical disks.

vgscan -a -p -v might be useful.

I suggest the -p option first to preview any changes it might want to make.

[Bill Costigan]

Any messages duringin the vgimport? That should report the number of disks and if any are missing.

The /etc/lvmtab on the other system in the cluster should still have an entry for this VG if you haven't exported it there yet. I would do a strings on it to get an idea of the number of PVs in the that were in the VG. You might also want to do a diskinfo on each of the PVs. A size of zero means you can see the PV but do not have access to it.

[Doug Short]

Thanks for all of the helpful advice. It was always my gut feeling that this wasn't necessarily an HP issue, as 2 servers not able to access VG's after a SAN upgrade pointed more to the SAN then to the server. Unknown to me at the time, this customer also upgraded some Cisco switches, which turned out to be where the problem was. Once the switches were resolved, I could then activate all VG's normally. Sorry to have wasted all of your time, but I do appreciate the responses.

[Doug Short]

Cisco switches were the culprit here. These were upgraded at the same time as the SAN, but the customer failed to inform me of this. Thanks for all of your help, and sorry to have wasted your time.

[Mark Nieuwboer]

Doug,

appreciate the responses with pionts.

grtz. Mark