Migrating an MC/SG cluster from one EMC Symmetrix to another

[Jerry Jordak]

Hi all,

I have an MC/ServiceGuard of seven L2000s and an N4000 all running HP-UX 11.00 and MC/SG A.11.09. All run packages with Oracle databases. The N4000 is a failover node for the L2000s, and other than running a couple minor admin databases, it is used strictly as insurance against a major hardware failure on one of the L2000s.

All nodes are dual-attached via a SAN to an EMC Symmetrix 8530. The L2000s all run PowerPath, and the N4000 uses PV links.

We need to migrate these boxes to a new EMC DMX 1000 and a new SAN fabric. The data migration should be pretty straightforward--we plan to disconnect one fibre connection from each server, attach it to the new SAN, add disks from the DMX 1000 to the existing volume groups, mirror the LVs to the new disks, and then delete the LVs from the old disks and reduce out the old PVs. (The size of the new PVs will be the same as the existing ones.) Finally, we'll move the second fibre connection over to the new SAN and zone the disks again to create the alternate paths via PowerPath or PV links. The last step would be to reimport the VGs from the L2000 nodes onto the N4000 to preserve failover.

Theoretically, I should be able to do this without stopping the package or cluster, since I'm not modifying any VG definitions--just the specific disks the VGs reside on. I see this the same as if all of the disks were given mirrors and then the mirrors were removed, which is really all I'm doing. Most importantly, since I can create and break the mirrors online, I can should be able to avoid any downtime during the data migration. (I'll have to take an outage on each node to upgrade PowerPath, HBA drivers, and other items that EMC recommends, but I can deal with that.)

Does anyone see any problems with this data migration approach that could cause cluster issues later? I've thought through it and can't come up with any myself. However, since this is one of our higher-profile environments, I wanted to get some other opinions from anyone who might know more about how to do this or who has done this before and hopefully save me some potential problems later.

Thanks in advance,
Jerry

[Geoff Wild]

Everything looks good except for one thing - cluster lock disk. You will have to reconfigure your cluster as soon as possible after removing second fibre connection....


Rgds...Geoff

[Sridhar Bhaskarla · Accepted solution]

Hi Jerry,

That's eactly what I have been doing here except for cluster lock disk where I ensure I do not take out the path corresponding to it first. But in your case I suspect you don't use cluster lock as yours is an 8 node cluster. You should be alright.

I don't know how many paths you have to your servers. If this is a high-profile production environment, then I suspect that you may have more than two paths in which you are safe. If you have only two paths, then depending on the time it takes to mirror, you will have to cross your fingers as any path failures can impact your production. Just ensure you have the map files taken and the configuration saved before you attempt to do the work.

-Sri

[Geoff Wild]

Sorry - just realized it's a 8 node cluster - no points please.

Your plan should work fine.

Rgds....Geoff

[Jerry Jordak]

Thanks, guys. That's what I needed to know!

Jerry