How to change Cluster Lck Disk in HP UX

[Gunasekhar]

Hi,

We have migrated storage from DAS to EVA. But Cluster lock disk configured in old storage only . Now i want config in new EVA.

We have created new VG in eva and Prasented to nodes .

Let me know process further to proceed.

Thanks
Guna

[melvyn burnard]

You need to modify the device files as shown in your cluster ascii file, then halt the cluster, activate the cluster lock vg on a node, use cmcheckconf/cmapplyconf to aply the change, then deactivate the cluster lock vg, then restart the cluster.
Unless you are using Serviceguard 11.19 or 11.20 where you should be bale to do an oinline change

[Matti_Kurkela · Accepted solution]

First, which version of Serviceguard are you using? (Run "cmversion" if you're not sure.)

Serviceguard A.11.19 and newer versions allow changing the cluster lock methods while the cluster is running; with older versions, you must halt the entire cluster to modify cluster lock configuration.

The basic steps are:

1.) Use "cmgetconf -c /tmp/clusterconf.ascii" to get the current cluster configuration in ASCII mode to /tmp/clusterconf.ascii.

2.) Edit the lines referring to cluster lock disk in the /tmp/clusterconf.ascii to match the new cluster lock disk. (Note: each node has its own lines, so make sure you change all of them.)

3.) Use "cmcheckconf -v -C /tmp/clusterconf.ascii" to verify the new cluster configuration. If verification fails, fix any problems before continuing.

4.) If your Serviceguard version is A.11.18 or older, halt the cluster now. ("cmhaltcl")
If your Serviceguard version is A.11.19 or newer but you're using Veritas CVM, halt the SG-CFS-pkg package.

5.) Use "cmapplyconf -v -C /tmp/clusterconf.ascii" to apply the new cluster configuration. This command will automatically send the new configuration to all the cluster nodes.

6.) If you had to halt the cluster, restart it now. ("cmruncl")

MK

[Stephen Doud]

The only thing I can add to Melvyns explanation is that the FIRST_CLUSTER_LOCK_PV reference after each NODE_NAME section in the cluster configuration ASCII file refers to the cluster lock disk as seen from that node's point of view. Sometimes the /dev/dsk/c-t-d- device special files (DSF) differ between nodes.
Use 'strings /etc/lvmtab' on each node to inspect the file and determine the actual /dev/dsk/c-t-d- DSF that system identifies with the cluster lock disk.

[Bob_Vance]

We seen posts here in the forum where the lock disk in a data VG can cause headaches later in manipulating the data VG.
E.g.

http://h30499.www3.hp.com/t5/LVM-and-VxVM/Remove-Cluster-Lock-Flags-from-PV/m-p/5281020#M49278

It will keep things cleaner if you make a separate small VG for the lock disk. On the EVA the smallest LUN size will be 1G and it can be simple raid 1 (mirror). You'll waste that small 1G of space.

Do not use this VG for any data -- just the lock disk.
Don't even need to create a Logical Volume on it.
The VG doesn't even have to be part of any package and it doesn't need to be activated.

Of course, this requires a re-config as explained by other posts herein.


bv