Backup VM with S2D and SQL Server

[mkevenaar]

Support case: #03683204

Hi,

We have the following two VMs setup:

• Windows Server 201 6 Datacenter, running a Microsoft Failover Cluster.

• The VMs have one OS disk and five "normal" virtual disks attached that are configured in a Storage Spaces Direct (S2D) setup.

• Inside the S2D pool, three volumes are created for storing SQL Server Data, log and backups

• I also added a HA File Share on a forth disk.

These VMs are running on a Windows Server 2016 Datacenter Failover Cluster with Hyper-V role enabled, using an SMBv3 file share.

With Support we figured out that a snapshot inside the VM is currently not possible.

Does anybody has experience with this setup and creating a backup from Hyper-V (without an agent)

Thanks,

Maurice

[Mike Resseler]

Hey Maurice,

I quickly checked your support case and I have a few ideas. It looks like (for some reason) the VM cannot take the production checkpoint. As a simple test, can you do a production checkpoint outside of Veeam (note: by default it will failover to a crash-consistent snapshot in Hyper-V if production doesn't work... so you might want to change that setting first to "force" it to take a snapshot.). I think there might be an issue with the in-guest SQL VSS writer

[mkevenaar]

Hoi Mike,

thank you for your reply.

The VSS production checkpoint is failing with the following event error messages:

Volume Shadow Copy Service error: Unexpected error calling routine IVssAsrWriterBackup::GetAsrMetadata. hr = 0x8007048f, The device is not connected.
Volume Shadow Copy Service warning: ASR writer Error 0x8007048f. hr = 0x00000000, The operation completed successfully.
Volume Shadow Copy Service error: Unexpected error while calling GetStorageDependencyInformation. hr = 0x8007048f, The device is not connected.

The only "device" that is not connected that I can think of, would be the DVD drive or one of the CSV disks that other node in the cluster is the owner of.

[Mike Resseler]

Check if there is an ISO in the DVD-drive connected. That could be the issue.
Second thing to check: On the VM itself, run vssadmin list writers. I'm guessing one or more writers will be in a fault state

[mkevenaar]

Hi Mike,

Support and I already checked the VSS writers.
The writer with an error:

Code: Select all

Writer name: 'ASR Writer'
   State: [1] Stable
   Last error: Non-retryable error

[nmdange]

Have you considered using SQL AlwaysOn Availability Groups instead of trying to set up a failover cluster with S2D inside VMs (which I don't think is supported by Microsoft)

[mkevenaar]

Hi nmdange,

I would love to, however the software using the DB does not support AlwaysOn AG.
I think running S2D inside VMs is supported: https://docs.microsoft.com/en-us/window ... rect-in-vm

However, reading this document showed me this:

Not supported: Host level virtual disk snapshot/restore

Instead use traditional guest level backup solutions to backup and restore the data on the Storage Spaces Direct volumes.

So, a (production) checkpoint is not supported. I think we need to switch to Agent backups.