lost cluster lock vg

[alexmartin]

Hello,

I have two hp-ux rp7410 servers and one HP VA. I was having 3 packages on my cluster. All packages were configured in active/passive mode with service guard. Few months ago two packages (databases) were intentionally removed from the service guard cluster to implement oracle RAC, and the RAC was successfully implemented. So, now only one package (SAMBA Shared folder) remains in the service guard cluster. The removed packages were used two volume groups (vg01 and vg02) with small size, and these two volume groups were built by several small LUNS in array. Few weeks ago I removed those small LUNS from the HP virtual array and also removed the unused volume groups (vg01 and vg02). Afterwards I created two big LUNS and also created a new volume group by using these two big LUNS and it is mounted locally on the primary node. Then I realized that the cluster lock was placed in vg01 during the cluster installation and now I removed that volume by mistake. Due to that I am getting error messages in both server syslogs. So, the cluster is now working without cluster_lock_vg and cluster_lock_ pv which is very risky.
As I have mentioned I still have one package on the service guard cluster and that package is using a volume group called VG03. So, my question is that can I change in the cluster configuration ascci file the parameters as follows and use cmapplyconf to generage and distribute cluster configuration files into both nodes. If I do that what will happen to the existing files in vg03? Will it be deleted or not?. Because the vg03 samba share consist lot of user files. Also what will happen to the existing samba package, do I need to reconfigure the package too.

From cluster.ascii file

FIRST_CLUSTER_LOCK_VG /dev/vg03 (earlier it was vg01)
FIRST_CLUSTER_LOCK_PV /dev/dsk/c5t5d3 (This is the device file which is used for VG03)
VOLUME_GROUP /dev/vg03 (earlier it was vg01)

Error on primary node

Mar 18 08:43:43 noor_pri vmunix: Unexpected error from cluster lock query (lock_id=0): Invalid argument
Mar 18 08:43:43 noor_pri cmcld: Unexpected error from cluster lock query (lock_id=0): Invalid argument
Mar 18 09:44:58 noor_pri vmunix: Unexpected error from cluster lock query (lock_id=0): Invalid argument
Mar 18 0

Thanks in advance
Alex

[Siju Vadakkan]

So, my question is that can I change in the cluster configuration ascci file the parameters as follows and use cmapplyconf to generage and distribute cluster configuration files into both nodes.

TO change your cluster lock disk u need to halt the cluster
and then ,Edit ur cluster configuration ascii and do a cmcheckconf and cmaaplyconf.


If I do that what will happen to the existing files in vg03? Will it be deleted or not?. Because the vg03 samba share consist lot of user files.
Nothoing will happen to ur existing VG vg03 and the data becz cluster only updates VGRA with cluster lock information. It is not touching the user data area.


Also what will happen to the existing samba package, do I need to reconfigure the package too.

You don't need to do anything on ur existing package, if it is not starting just do cmapplyconf for the package as well.

[alexmartin]

Hi Siju

Can i do it in this way?

1 execute cmhaltcl on the secondary sevrer where the samba package currently running

2 vgchange -a y /dev/vg03 on the secondary
server

3 cmcheckconf -k -v -C /etc/cmcluster/cmclconfig.ascii (modified ascci file) on the secondary server

4 cmapplyconf -k -v -C /etc/cmcluster/cmclconfig.ascii on the secondary server

5 vgchange -a n /dev/vg03 on the secondary server

6 cmruncl -f

Thanks
Alex

[Siju Vadakkan]

Yes , exactly.Go ahead

[alexmartin]

Hi guys,

I am planning to reconfigure the ServiceGuard cluster lock on this coming friday 18-Apr-2008 (Indain time 10:30am to 1:30pm).
"Edited to comply with ITRC Guidelines"

Thanks in advance.

Alex

[melvyn burnard]

You need to halt the cluster
vgchange -c n vg03
vgchange -a y vg03
edit the ascii file (make sure that you specify vg03 in the cluster ascii file at the bottom) and apply the new binary, then start up the cluster