device file changed for disk

[mogiro]

Hi,
I have changed my storage cabinet on my mc cluster and now in one of the nodes (the secondary one) the HW path has changed from /dev/dsk/c33txdx to /dev/dsk/c23txdx
I have a the map files created with vgexport on the main node. Should a vgimport on the secondary node solve the problem?
Thanks in advance!!

[Torsten.]

" have changed my storage cabinet " - how exactly? Yes, importing the VG will update the LVM configuration on that node.

[mogiro]

Physically, We have replaced our storage array, we have mounted a brand new one with more capacity but keeping the current LUN's structure, we have sinchronized data on both arrays (old and new one). As I have said, main node on cluster hace seen the array without doing any extra operation, as the Hw path has not changed. But on the secondary one, it has changed and now the node is claiming for the lock disk and so on.
I hope the issue is clearer now for you.

[Torsten.]

Not clear enough, but maybe you changed some port or switches too. I assume you did not remove the old device files before you changed the environment. Anyway, the vgimport should help.

[Emil Velez · Accepted solution]

you need to export the volume group from the secondary node and import it back in. That is unless the volume group is the one containing the lock disk. This means that you are redefining the lock disk which cannot usually be done while the cluster is up.

[mogiro]

I'm having this message on the syslog:
Jul 25 14:54:15 seplet6 cmcld: WARNING: Cluster lock disk /dev/dsk/c33t0d0 has failed: I/O error
Jul 25 14:54:15 seplet6 cmcld: Until it is fixed, a single failure could
Jul 25 14:54:15 seplet6 cmcld: cause all nodes in the cluster to crash

Because now the lock disk is c23t0d0. I can't just change it on /etc/cmcluster/cluster.conf with the cluster up?

[Stephen Doud]

If /dev/dsk/c23txdx is the new path to the cluster lock disk, you must update this in the cluster configuration ASCII file and run cmapplyconf on the file to update the cluster binary file with this new information. Doing so will also install the cluster lock structure, if it does not exist on that disk.

As for vgimporting the VG, it depends on how to created the new VG on the new array... if you did a bit-for-bit copy of the old VG disks to the new ones, then the new disks will have the same VGID as the old disks. If both the old and new disks are visible to the second server, you won't want to do a vgimport with a map file that contains the VGID at the top... doing so will make LVM discover both old and new disks.
In such a case, you have to use a map file that does not contain a VGID at the top, and specifically identify each disk path in the vgimport command.

[Deepak Kr]

Hi Moqiro,

Please confirm if you are still connecting your old storage box to your nodes or not. If yes, then running vgimport using previously created map file will not work as map file will be having old VGIDs information.

I would suggest you to take complete backup of your existing cluster configuration before you make any changes using new disks comming from new box.

[mogiro]

Hi,
My old cabinet is far away from here. But as this is a cluter, I've got the main node working with the new cabinet with no problems. What I did is vgexport on the main node and vgimport on the secondary one. However it is still remaining to solve the lock disk issue. I wonder if I can just change it on the /etc/cmcluster/cluster.conf file and cmapplyconf. Is it correct?
Thanks!!

[Deepak Kr]

Thanks for the confirmation.

What I understood that one node is able to recognise all the disk devices including lock disk. If this is the case then updating conf file and running cmcheck followed by cmapplyconf should work.



update cluster lock lun for second node e.g.

CLUSTER_LOCK_LUN /dev/dsk/c23txdx

After cmapplyconf

you can run cmrunnode -n second_node_name -v to join cluster ..

This should update cluster lock.

[Deepak Kr]

Sorry forgot to mention

You need to halt cluster to update cluster lock disk in conf file.

[mogiro]

Ok.
Now it is clear for me. I will wait for a planned shutdown of the system to update the lock disk on the cluster.
Thanks a lot to all you.

[mogiro]

.