SAN disk paths to be modified in a MC/SG Cluster

[Josep Arques T.]

Hi,
soon we will have to modify the current SAN disks paths in order to improve I/O performance.

We're running a 2 node MC/SG cluster, each node running 1 different package.

The target is to update the disks paths with any service loss.

Here's the process to be executed:

1. Find the new disks paths and create the devices associated, in each node: ioscan + infs

2. In each node, for each active VG to be updated, vgextend with the new paths and vgreduce the old paths.

3. vgexport a mapfile of the updated VGs, where active, and vgimport the mapfiles in the other node.

4. As soon as possible, test the cluster by moving the packages from one node to the other one and finally to the original one.

[Steven E. Protter · Accepted solution]

Shalom,

The only possible problem is if your lock disk is on the SAN and is going to disappear.

That going away would bring the cluster down hard.

Plan some downtime for the transition.

High Availability does not mean 100% availability.

SEP

[Josep Arques T.]

Thanks Steven

It's possible that the lock disk is to be included in the upgrade process. At this moment I'm waiting for more information about the VGs involved in the process.

I will look for more information.

Thanks

[Rita C Workman]

It's been awhile since I did a data migration, but....sounds like your just maybe looking at switching HBA's (i.e. no data migration).

Maybe I need more coffee or something, but I'm thinking creating fresh mapfiles, do your I/O improvement thing, follow up with running ioscan and insf -e (step 1) and then do vgexport/vgimport.
No vgextend with new paths - in fact, you would probably need to reboot and maybe you might do some old hardware path cleanup.
Don't know what disk software you have, but if it's EMC and PowerPath you'd need to clean up there to get rid of what would now show as dead links (old paths) so it could see the active links (new paths).

Like Stephen says....take down the cluster.

Rita

[Kenan Erdey]

Hi,

if i didn't understand wrong, with 11.19 cluster lock disk changes are done online.

http://docs.hp.com/en/B3936-90140/ch07s04.html#d0e32434

[Josep Arques T.]

Hi Kenan

I've queried the SG version, swlist shows:
ServiceGuard A.11.15.00

Thanks to all, I'm closing the thread, once I get all the information I'll make a more detailed action plan.

[Josep Arques T.]

I've found the solution in the colleagues replies.