V11 - CDP & vm-snapshot question

[guy_hd]

Hi guys,

I'm sure this is a simple question, however I'm a bit green on vmware I/O Filtering so thought I'd double-check.

A client is interested in CDP. They already have cross-site (cluster) Veeam replication at the moment which is generally working well. Even whilst leveraging storage snapshots, they have however experienced 1 or 2 issues with an application crashing due to vm-snapshots. Coupled with the RPO offered by CDP, the feature appears to look like a winner. Just checking the guide though and the obvious gotcha is "VMs that you plan to protect must not have snapshots.".

Coupled with the replication, we have a well established Veeam backup routine, taking backups to independent backup repositories etc - naturally using vm-snapshots.

Can I just check my assumption that any vm's within a CDP can not ever have vm-snapshots taken for the filter I/O based CDP to work? Meaning agent based backups only for these vm's?

Cheers,
Guy

[Gostev]

Hi, Guy

No, there are no limitations like that.

The only limitation is that the I/O filter cannot be attached to VMs with snapshots. So you need to remove all snapshots from VMs to be protected before creating a CDP policy. But once that is in place, you can use VM snapshots as you please.

Thanks!

[guy_hd]

Amazing! Thanks for clearing that up Gostev.

[RyanJ]

I'm tagging onto this topic because I've witnessed something completely different. I'm pasting an email I sent to our SE's describing the behavior I'm seeing.

The documentation lists one of the requirements of CDP is that no VMs can have snapshots on them, which I understand. However, what happens when you’re backing up one of those same VMs with Veeam and it takes a snapshot? I tested this, and the behavior I observed was quite interesting. CDP appeared to be ok with backup job creating the snapshot, however once it went to remove the snapshot and perform the disk consolidation, CDP essentially broke. The data transfer was still continuing, but the task and I/O I was seeing was indicative of the VMDKs being almost fully reseeded back to the DR site. The RPO of my test VM climbed up to almost 20 minutes while I was waiting for the sync to happen again. After about a day, the journal caught back up and everything was fine, but I would imagine for large data sets, that won’t be true.

How are we supposed to backup VMs with Veeam, while simultaneously protecting them with CDP?

[Gostev]

There are no known issues with this scenario at this time, so you should open a support case to investigate what is going on there.