Best approach to back up a 2-node windows server cluster running storage spaces direct (S2D)

[jloube]

I'm trying to figure out the best approach to back up these systems

2x Windows Server 2019 VMware VMs in a failover cluster with a file share witness for quorum
Running only a file server role
Using 'storage spaces direct'/S2D with 5x250GB virtual disks on each of the 2 nodes

These are backed up via a 'protection group' defined on the Veeam server. This deploys the veeam agent to each machine and backs up at a volume level instead of VM level.

A recent issue required a restore of one of the 2 VMs, and the process was an absolute nightmare, which has led me to believe that it’s not really possible to properly restore from this type of backup into a virtual environment. Everything about it seems to assume that you’re backing up from and restoring to bare-metal.

I was able to restore the system volume from the veeam server but it completely skipped the clustered data disks. I guess it assumes that you will restore the clustered disks afterwards? Anyways, the issue is that the data volume is split across multiple virtual disks and those all need to be restored and recreated for S2D to work, but since it’s a volume level backup there’s no concepts of the underlying virtual disks.

I've tested doing regular VM backups instead, and this seems to completely negate all the above issues, full VM restores are trivial. But it introduces a new one - file level restores no longer work. The clustered/shared volume does not appear at all in the veeam backup file browser.

Any suggestions for a backup technique that would both allow file restoration and full VM restoration?

[tyler.jurgens]

Since this is a file server, have you thought about skipping all those shenanigans and backing up the file server itself with Veeam Unstructured Data backups? https://helpcenter.veeam.com/docs/backu ... ml?ver=120

Seems like the perfect use case for backing up your file server as currently configured. It would also allow much more flexible recovery options than you currently have configured. https://helpcenter.veeam.com/docs/backu ... ml?ver=120

You wouldn't need to do a full VM recovery at that point if your VMs are only running the file shares.

[jloube]

Interesting, I have not looked at the unstructured data backup, I will test that out. That still wouldn't have any concept of the 10 underlying virtual disks that the data lives on though right? So a full restore could still be an ordeal. Maybe a 'hybrid' approach where the VMs are backed up weekly and the data backup runs nightly or something could do it.

[tyler.jurgens]

This method would just backup the files in the file shares themselves, ignoring everything about what you use to present those file shares.

In the case of a restore, you would pick where to restore those file shares onto - maybe you have some NAS, or you want to rebuild the S2D cluster and restore onto that - it would be flexible that way. You wouldn't necessarily need to handle backing up the VMs anymore, as they become disposable.

[Andreas Neufert]

We do not support Storage Spaces Direct in VMs or Agents in general. Microsoft does not support VM backups likely why you looked into the Agent.
The reason why we do not support it, is that our file level recovery do not know how to mount the disks.

I would like to ask you as well to look into the specific requirements of S2D in VMs. It is is very important to implement the changes outlined in this article.

As you use a file server, yes, I would use our NAS backup to backup the file shares over the SMB protocoll daily. Then backup the VMware VMs in the same job at the same time. Ensure that at backup time there are enough free Proxy/Repository task slots available that both VMs are backed up nearly at same time. Maybe do this weekly, potentially excluding the data that you backup with the NAS backup. Yes ignoring the Microsoft not supported backup method.

For regular file level recover demand, you can just use our file level recovery wizard from NAS backup. As well a point of backup restore is possible.

For disaster recovery you can restore the 2 VMs (potentially instant restore it), recreate the shared disk, setup S2D to the point that it is working again, and then do a file level recovery from the NAS backup to ensure consistency of the filer data.

[jloube]

Awesome, thanks. Testing out the SMB backups now.

It is is very important to implement the changes outlined in this article

Did you mean to link to an article there?

Then backup the VMware VMs in the same job at the same time.

Meaning back up the VMs AND the shares in one single job? I didn't think that was possible?

-Joel

[Andreas Neufert]

Sorry, here is the link: https://learn.microsoft.com/en-us/windo ... rect-in-vm

The idea was to backup the share independently with NAS backup.
Then backup for disaster recovery the VMs (mainly for the OS) so that you can recover the configuration before you restore the share data in case of a major disaster.

[jloube]

Got it, yeah that's kinda what I figured, just wanted to make sure I wasn't misunderstanding. In testing, the SMB backups combined with VM backups seems to all be working well, so I think that's the way I will go. Thanks for the help and suggestions!