Change FC storage backend

Latest response

I'm planning on migrating our RHEV storage from IBM DS4800 to IBM Storwize V7000U next week. The storwize supports virtualizing other storage, so this migration has been trivial for all our standalone servers. All I've needed to do on the servers is to update the multipath.conf to have a device section for this new storage system, and also update any "multipath { wwid=*" } where we had configured that.

 

Is there anything we need to do on the thin rhev-hypervisors to have it use the new storage system, or will it detect any FC-device and not care about wwid's ?

Responses

If you're only adding new storage domains, all you need to do is connect the cables and rescan the bus (or reboot the hosts). If you're actually moving data over, this can be a bit trickier

 

I'm actually moving the data.

 

The plan is to:

    1 - take down all hypervisors

    2 - Change LUN masks on old storage system so that only the storwize system can see them.

    3 - Virtualize old storage behind new Storwize system, and map the LUNs back to the rhev hypervisors.

    4 - take up hypervisors

 

Alle the LUNs will then be the same, except they come from a new storage controller, and have new wwid's. This has been transparent for Oracle ASM volumes, IBM GPFS, plain LVM2, etc.. on several systems, so is there any reason this should be more tricky for RHEV ?

Should work with 3.0 (will not work with 2.2 without additional metadata and DB manipulations)

 

Just in case, keep the old LUNs intact until the migration is verified to be successful

Dan

 

Correct me if I'm wrong but my understanding is that all FC type storage is placed under LVM control in RHEV-H.

Since WWID's are irrelevant to LVM I would expect this to be transparent to the hypervisor and the migration would be successful.

However, if you boot your hypervisors from the SAN, then that's a whole different kettle of fish !

This is why I said the procedure should work for 3.0, but not for 2.2. In 2.2 the WWNs were actually hard written into the storage domain's and master SD metadata and into the database. This is one of the major changes in 3.0, and one of the reasons we have a V2 storage domain format

Unfortunately we have both a RHEV2.2 and RHEV3 environment we need to migrate to this new storage. Is there any way we can change the WWNs in the v2.2 SD metadata/database ?

It is possible to do, but you'll need to do that through a support ticket - we would need to take a database dump and logs from the most recent SPM and devise a set of scripts for you. It is a lot easier if you simply add a new SD from the new SAN to the existing DC, and move VMs over from within the admin GUI (shut VM down, right click, "move")

Unfortunately I'm out of free FC ports, so I can't have the hypervisors connected to both storage systems :-(

 

I think I'll have to postpone this migration until we've moved all RHEV 2.2 VMs over to our RHEV3 environment.

Unfortunately I'm out of free FC ports, so I can't have the hypervisors connected to both storage systems :-(

 

I think I'll have to postpone this migration until we've moved all RHEV 2.2 VMs over to our RHEV3 environment.