Cluster lock problem

[Luis Amoros]

When trying to build cluster, I get following error: "Error: First cluster lock physical volume "/dev/dsk/c1t3d0" on node "nmssrv2" does not correspond to the same device as the First physical volume specified for node "nmssrv1". It cannot be used for cluster lock.
Maximum configured packages parameter is 4.
Configuring 3 package(s).
1 package(s) can be added to this cluster.
cmapplyconf : Unable to verify cluster file: /etc/cmcluster/cmclconf.ascii.
Invalid argument.".

I have two nodes and have upgraded MC/Serviceguard from 11.3 to 11.9 .

[Clemens van Everdingen]

Hi,

Looks like your cluster lockdisk is changed in one of the configuration files on one of the systems.

They should be configured as the same disk.

C.

[melvyn burnard]

So you need to verify the two PV's specified in the cluster ascii file are the same disc.
check the entry with a strings /etc/lvmtab on the relevant node
If they are not the same disc, then this will be the problem.
One way to get a valid disc is to redo a cmquerycl -C asciifile -n node1 -n node2 and see what disc that identifies as useable for cluster lock

[G. Vrijhoeven]

Hi,

I must agree with melvyn. Just remember that the c name does noet have to be the same on both nodes, so the device files do not have to be identical.

Gideon

[Luis Amoros]

Unfortunatly now I have other error as I read a similar case in the BBS. It said to do a vgcfgrestore -n /dev/vgdb01 /dev/dsk/c1t3d0, after that I think I should have to recover VG from tape as I get following error when I try to activate the VG: "Quorum not present or some phisical volume are missing".

What can I do?.

[melvyn burnard]

I think you are digging yourself into a deeper hole.
You need to verify:
which PV's on which node belong to which VG.
Rectify any of these issues.
Then try to sort out the ServiceGuard issue.
Of course, it may already be too late and you could be reduced to recreating your VG's from the beginning.
Best of luck

[Ashwani Kashyap]

Melvyn said it right . Before trying to solve your cluster issues , first try to sort out which PV, belong to what VG's on both the nodes .

Remeber as far as shared disks goes , both nodes should have the same PV's in their respective VG's .

Once you have this information , it will be easier to sort your cluster issues .

[Luis Amoros]

Thanks for the hints guys!.

I recreated lvmtab file by doing a vgscan, and then after doing vgcfgrestore, mirror was rebuild and now I can activate VG without errors, so what was wrong?.
I think I will export VG on other node and import VG, just to be sure that lvmtab is OK on other node.
After this I will continue with my problem of cmapplyconf.

Br,

[Hamdy Al-Sebaey]

Hi Luis,
May be I'm little bit late, but I 'd the same problem in the past,with vgscan, en vgrestore, I was able to solve this problem.
Regards,
Hamdy

[Clemens van Everdingen]

Luis,

Assign some points to the answers given.

C.