vg remove from cluster

[Jeeshan]

I have two rp7420 servers with MC Serviceguard and OS are 11.11. I’m using raw file for cluster file system named vgraw, vgraw1 and vgraw2.

Previously it has Oracle 9i RAC and it’s using raw file vgraw and vgraw1. In vgraw there are 3 LUN’s and one LUN is assigned as cluster Lock Disk. Later on 9i RAC was upgraded to Oracle 10g RAC with CRS using vgraw2. Now RAC is running fine in 10g.

I now want to remove all LUN’s from vgraw1 and vgraw leaving Cluster Lock Disk LUN.

My action plan is to remove LV, PV & VG of vgraw completely and reduce vgraw leaving on LUN. And then modify the cluster configuration file removing vgraw1 and rebuild the cluster.

Is it enough or I need to take other concern in my action plan?

[Suraj K Sankari · Accepted solution]

Hi,

Your plan is absolutely correct.

Suraj

[Duncan Edmonstone]

well which VG is the cluster lock disk in?

you can't remove the VG with the cluster lock disk (at least not whilst the cluster is running anyway - you could stop the cluster and re-assign the cluster lock into vgraw2)

HTH

Duncan

[Jeeshan]

well, Lock Disk in vgraw.

Sorry as i mentioned.

My action plan is to remove LV, PV & VG of vgraw1 completely and reduce vgraw leaving the Lock Disk LUN. And then modify the cluster configuration file removing vgraw1 and rebuild the cluster.

[Stephen Doud]

In the man page of most of the LVM commands is this statement:
cannot be performed if the volume group is activated in shared mode.

This statement is referring to volume groups marked for use in SGeRAC environments. vgdisplay will show the status of such VGs as "active, shared".
Before you can do LVM changes to a VG or LV, you must deactivate vgraw1 and decluster it (vgchange -c n -S n vgraw1). The same goes for your lock VG (vgraw).

If vgraw1 will be entirely removed from cluster use, remove it's OPS_VOLUME_GROUP reference from the cluster ASCII file.

If the lock VG will no longer be used for raw access, you can change it's designation from OPS_VOLUME_GROUP to VOLUME_GROUP at the bottom of the cluster ASCII file, then activate the VG and then do the cmapplyconf on the file.

To insure the cluster lock VG maintains the lock structure, halt the cluster, activate the lock VG (you may have to decluster it and activate it (vgchange -c n -S n vgraw, vgchange -a y vgraw), before cmapplyconf'ing the cluster ASCII file.

[Jeeshan]

Thanks buddies. I have done the activity and everything was fine.


But anyway, I have found much more considerations and findings whilst working. I'll share with you later on.