Adding vdisk in Metrocluster Env

[china_siney_tan]

Hi All.

We want to add the extra vdisk to enlarge the filesystem size which service as our database in MetroCluster Env.I need to know the steps for the MetroCluster portion and is it necessary to take the cluster down for it , thanks


OS version HP-UX B11.11
SAN storage EVA4K & EVA8K

[Matti_Kurkela]

It's been a while since I operated a MetroCluster, and we used EMC storage instead of EVA, which means some things may be different.

But in general terms:

1.) Verify that the new vdisk is replicated from the primary storage to the failover storage. Make sure you know how to identify the vdisks visible to your nodes: you and your storage administrator should have some common way to uniquely identify each vdisk.

You should also know the replication mapping between the primary and failover storage, so that you can verify that the failover node sees the correct disks. As the failover node uses a different storage system than the primary one, your storage admin must give you this documentation.

2.) Update the cluster documentation to include the new vdisk.
This is important: do not skip it.
When you're on vacation and there is a problem, your fellow sysadmins will hate you if your documentation is out of date.

3.) Verify the actions needed to "flip" the direction of replication when a failover happens, and make sure MetroCluster configuration is updated to include the necessary storage commands (if any). This part is specific to the storage system type, and I don't know what needs to be done with EVAs.
(This is assuming that EVA vdisk replication makes the replica on the failover storage be read-only, and the storage needs to mark the primary side as read-only when the failover side becomes active. EVA might be different here.)

4.) The primary node should now be able to access the new vdisk in the primary storage, and the failover node should be able to access the corresponding new mirrored vdisk in the failover storage.
The failover node should NOT see the primary storage, and the primary node should NOT see the failover storage.

5.) Add the new vdisk to the VG and enlarge the filesystem on the primary node, just like on a regular ServiceGuard. No need to shutdown the cluster for this.

6.) If the EVA allows read-only access to the failover side when the primary side is active, you should be able to export the Metrocluster VG on the failover node, copy the VG map file from the primary node to failover node and re-import the VG to make it include the new vdisk (in this case, the failover replica of the new vdisk) just like with regular ServiceGuard.

If EVA does not allow read-only access to the failover side, you'll need to do something different. Refer to the "MetroCluster for EVA" installation instructions: see how it was done at the installation time. A similar procedure is probably applicable here too.

7.) Because MetroCluster is more complex than a regular ServiceGuard, schedule a failover test at the earliest convenient time, especially if you haven't done this before.
This is important: do not skip it.

MK

[Stephen Doud]

Page 168 in the document titled
"Designing Disaster Tolerant HA Clusters
Using Metrocluster and Continentalclusters"

at http://docs.hp.com/en/B7660-90026/B7660-90026.pdf

contains instructions on vgimport'ing volume groups on EVAs on local or remote sites.