MD3000 snapshot LVM PV to same server

Hi

 

yes I mapped it to another server, pvchanged the uuid then mapped it back to the original server.

It seems like quite a hassle though.

 

I take it snapshot/disk copy is not designed to be used with LVM ?

 

We have an 800GB Oracle DB that we want to create a copy of every night, then the users can access that copy the following day. If we have to present it to another server first to change the uuid then I guess were going to have to look at doing this a different way.

 

Thanks for your help.

 

Mount the snapshot to another server. This way the matching UUIDs isn't an issue.

Mounting a snap back to the original system has a lot of quirks and caveats.

On a Windows server, it usually works ok as the 2nd 'disk' with the same disk signature gets it's signature rewritten. Without LVM Linux can rewrite the signature as well I believe. Mounting a snap back to a Windows cluster that owns the original disk can be very dangerous; if you give the snap to the passive node (the node that doesn't own the original resource/disk) the snap gets the drive letter of the original disk, but if then the cluster fails over to this node the 'real' disk can get it's signature rewritten as the snap already has the signature and drive letter causing your cluster to start using the 'stale' data in the snapshot instead of the real data, and you end up with downtime to fix the disk signature on the original disk.

Thanks for the info. I take it the same applies to disk copy ?

 

I reckon we could get round this by not using LVM but that will present it's own problems further down the line.

 

Oh well, what would life be like without these little niggles ;)

 

Yes, the same would happen with diskcopy.

One other thing about diskcopy on the MD3000(i) is that while it's performing the diskcopy, the source disk is set to read-only mode, which can be very intrusive.

One way around this I believe is to snap the original virtual disk and then diskcopy the snap.