Restore from NetApp Snapshot (Multi-Volume)

[tkeith]

Hi All,
Wanted to put it out there if others have tried to leverage Veeam to do NetApp level snapshots on top of their regular streaming backups. We have a large organization and struggle to backup our environment in a regular backup window. We decided we could use Veeam to trigger storage snapshot backups directly using the NetApp. We understand the limitation is all files for a VM need to be on the same volume in order to restore without issues. This looks to be a large percentage of our environment and we could significantly reduce the backup times by leveraging this through Veeam and providing the ease of restoration.

Now some of our more important VMs like DB servers for example have multiple VMDK's spread across different volumes (say SSD, SATA, SAS, etc). Trying to leverage the restore capability within veeam is not possible due to them being across these different volumes.

In our environment, trying to backup 4000-8000 VMs in an 8 hour backup window, using veeam for this piece would make meeting these requirements possible and significantly reduce the virtual infrastructure footprint we required to stream this amount of VMs in our backup window.

Does anyone else use Veeam to backup a large environment and try to leverage a portion, or all, of it using NetApp snapshot jobs and see this as a valuable feature to request?

The ability to restore a VM thats across multiple volumes in NetApp snapshots could be a huge plus for us.

[foggy]

Keith, you're correct regarding the limitation and we've already had several similar requests. So thank you for the feedback. I would just note that VMs that span multiple LUNs/datastores can be successfully restored without any issues from Veeam B&R backup files, it is just restore from storage snapshots that is not supported for such VMs. Thanks.

[dwoerz]

Not to drag up an older topic but I have the following comments...

Using the Veeam backup files to do a vm restore ties up the Veeam backup files so no new backups can be taken.
We like to take backups of our main sql database hourly throughout the day but since Veeam cannot restore from the san snapshots because of different volumes I have to backup to Veeam storage But with doing this, I cannot do an instant vm restore for testing or the sandbox functionality because I need the backup files free for the next hourly backup.

While I am a small shop with 20-25 vm's, and can backup all of my vm's in 2-3 hours, I also do not have the superfast storage for backups. So functionality like the instant vm restore or running a vm in a sandbox has to be done from the Veeam storage because of the different volumes. This functionality would be a lot faster if it were run from the same storage that the production vm's were run from.

I am not entirely sure why this functionality is missing, Veeam already has the san perform the NetApp snapshots on the other volumes and Veeam is already aware of the other volumes... Hopefully this will be something that makes it into a release soon.

Thanks.

[Henk Ruis]

Hi,

Anyone experienced with Instant VM Recovery on a NetApp snapshot(BFSS), where the vmdk's reside on different datastores/volumes, that the VM is is not completely restored?
My experience is that a Instant recovery only restore the vmdk's that resides with the VM and in that snapshot on the NetApp volume.

Is there a way to overcome this issue? I like to have the complete VM on a restore (also a FULL recovery does not solve this issue)

There is some kind of logic in this because the vmdk in the other datastore is saved in another snapshot/volume.
This snapshot is not mounted to the ESXi server when the VM is build up from the other snapshot.

Note:
A VM must be completely placed within the surrounding of a datastore in that way.
A Instant VM restore from the repository works fine so thats not the issue (also part of the JOB) but here the VM is presented throughout the PowerNFS share to a ESXi host.
And all the disk are in the repository.

Henk

[foggy]

Henk, this is currently a known limitation.

[mattisonward]

Almost all of our VMs have disks on multiple Netapp backed datastores, so I agree that this would be really useful and the feature is almost useless because of this limitation at this point.

[csinetops]

Agreed, we would use it but can't due to this limitation.

[Setareh]

I agree too. We are using Nimble and we were waiting a long time for this feature and now that it out, it is useless for use because of this limitation. We really need restoring from storage snapshots as hourly backups in veeam are slow due to the merge process. Hopefully Veeam address this soon and consider finding a solution for VMs with disks on multiple datastores.

[foggy]

Yes, this is on our table. Thanks for the feedback.

[csinetops]

+1 as a request from us. This limitation keeps us from using the NetApp integration with Veeam.

[Gostev]

We are using Nimble

Pardon my potential ignorance, but I have always been curious what is the point of spreading VM across different LUNs when it comes to the modern storage?

Sure, when I was young it made sense to deploy Exchange and SQL with database and logs on different spindles of a physical server with internal storage - but what is the point of doing so with the modern SAN, when I/O from all LUNs gets blended and lands on the same bunch of spindles (or SSD cache in your case) anyway?

[Setareh]

No worries. One reason is that we are using hybrid array not all flash and we don't want the logs to be cached but we want the active data be cached to have better performance.

[csinetops]

Pardon my potential ignorance, but I have always been curious what is the point of spreading VM across different LUNs when it comes to the modern storage?

Sure, when I was young it made sense to deploy Exchange and SQL with database and logs on different spindles of a physical server with internal storage - but what is the point of doing so with the modern SAN, when I/O from all LUNs gets blended and lands on the same bunch of spindles (or SSD cache in your case) anyway?

We have a hybrid NetApp array that has flash cache for some of the disk but not all. We use this in a tiered manner for our file servers. The server will have a disk or two on a LUN that has flash cache for the OS drive and fast file storage drive. It will also have a disk on a non-flash cached LUN for files that are not accessed as much.

[ARadschun]

Hi,

we need this feature, too. We have different storage tiers for our customers. Currently, however, our customers do not have the possibility to mix the storage tier within a Vm. So for us this is an important feature, because we sell different storage tiers!

Andre

[ARadschun]

Hello everyone, is there any news on this? Is it already on the list as a feature request? We are very interested in this! Is there already a timetable for the implementation of this?

Andre

[foggy]

Yes, it's on the high priority list.

[ARadschun]

Hi foggy, thanks for the quick answer! When can we expect this feature?

[foggy]

Unfortunately we cannot provide any ETA at the moment, but this feature has a high priority in our books.

[ARadschun]

Okay, great, thank you!

[Partly_Cloudy]

I discovered this limitation today during a DR. I have an exchange server that is protected by veeam snapshot from netapp storage. When trying to perform an instant VM recovery it was informed this is not possible. Though that is a disappointment, what I need to know now is what options for recovery do I have? Can I still use these snapshots to recover the server just not instantly recover it?

[foggy]

Hi Jeff, do you have a regular backup of this VM as well? There are no limitations on restore from the Veeam backup files for such VMs.

[Partly_Cloudy]

No regular backups. The entire environment runs as a single job from storage snapshots. I know we can recover individual items but what we really want is VM instant recovery which is not possible with the storage integrated snapshots. I am fine creating new backup jobs that do not use the NetApp integrated snapshots I just need to know the next most efficient way to do this since moving everything to the same volume is an option.

[foggy]

By regular backup I mean backup job that creates a backup file (VBK) on the backup repository (either using data from storage snapshot or directly from datastore), not a snapshot-only job that produces a storage snapshot as the only outcome of its operation. The discussed limitation applies to restores from the snapshots, but you can restore from the backup file just fine (even if it was created using storage integration). If you do not have regular backups and just rely on the storage snapshot, you're not covered, since storage snapshot cannot be considered a real backup.

[MajesticTJ]

Any idea when this will be implemented? We are interested in having the ability to restore from snapshots for VMs with multiple disks. Thanks!

[foggy]

We're actively working on this feature, but I cannot share the dates yet.

[giacy]

Hello there is news about that? We are really intrested to restore from netapp snapshot for VM that reside in different datastore.

Thanks!!!

[ChrisGundry]

Hi,

I think I am right in saying (from my testing) that if you have a VM which has .vmx and disks in datastore1 and then some other disks in datastore2, if you do an instant recovery it only seems to restore disks in datastore1. At least after the restore the VM doesn't have the datastore2 disks attached.

If this is correct, can it be looked at to resolve? I needed to do an instant recovery the other day and ended up with a server with most of its disks missing as a result. So the instant restore wasn't very useful as a result.

It has been a while since I needed to restore a whole VM (rather than instant recovery) so not sure if it restores the disks on both datastores on a normal recovery? Are you able to confirm?

Thanks

[Vitaliy S.]

Hi Chris,

Yes, currently it is a known limitation. Don't be confused with the topic name, this behavior can be seen regardless of the NetApp snapshot usage. Thanks!

[ChrisGundry]

Ok thanks Vitaliy. It seems this has been an issue for some time and no update on this topic. Is there an offical ETA for this? Update 4?

Would Surebackup suffer the same issue with these VMs that have disks on multiple volumes?
If asked Veeam B&R to do a full VM restore of the same VM, would it restore all the disks to all the volumes correctly?

[foggy]

Hi Chris, could you please clarify, are you talking about storage snapshots integration? Since this issue is specific to restores from storage snapshots, not regular backup files. Restore from backup files should work fine.