Anyone change the snapshot location partition?

[danm190]

Hello,

I have a High Priority VMWARE datastore which is an NFS volume presented by an EMC VNX 5300. The EMC provides the NFS replication to a target VNX box.

Since the NFS volume is replicated I was trying to keep changes to the partition to a minimum.

Veeam creates snapshots in the VM folders by default? It seems to be the smart play would be moving the snapshot location to another datastore?

Is it as simple as editing a configuration file? Does anything need to be done to the VM's? Power off and then on?

[tsightler]

Veeam does not really keep the snapshot anywhere, Veeam is simply taking normal vSphere snapshots. Changing the snapshot location was a common best practice for VMware admins with ESX(i) 4.1 and older, but ESXi 5.x has changed the way snapshots are stored and made this somewhat more difficult. Information on changing the location of snapshots for 5.0 is available in VMware KB article 2007563 and there are also links to the procedure for older versions. You should be sure you understand the impact of these changes prior to making them.

[danm190]

Yes, as I understand it.. Veeam is telling VMWARE to take the snapshots.. and Veeam also tells vmware to remove the snapshots.

From my reading it looks like ESX 4.x best practice to to relocate the snapshot workingdir folder...

http://kb.vmware.com/selfservice/micros ... Id=1002929

[dellock6]

Also remember, from a design perspactive, high IO virtual machines can grow their snapshot to high value while Veeam is saving them, be sure to relocate the snapshot to a datastore with enough space, I already saw this kind of design error, maybe placing snapshots in faster but smaller disks/datastores.

Also, from the VNX perspective, nothing really changes for the I/O: writes to disks are to be recorded anyway, in the primary vmdk or in its snapshot, but in both cases every IO is to be written to the datastore. Same story for the separated snapshot: you can save 1 write IO operation while snapshot is active, but when committing a write operation per bit has to be written to the primary datastore.

Luca.

[Berniebgf]

Also remember, VMware snapshots are NOT Copy On Write (COW) they are "Redirected Write".

So if you change your snapshot location, then your DR site RPO will be effected the moment you create a snapshot (if the snapshot location is not replicated).
Example, you have a snapshot open for 24-48hours for backup...your DR VM recovery state is now 1-2 days old. Worth thinking about...

Bernie