ReFS repository inside VHDX on single Hyper-V host

[burak]

Hello,

I would like to hear some feedback about a design question for a backup storage in a smaller environment. I have a dedicated backup Hyper-V host, which runs a Server 2019 VM for VBR. The backup storage is a RAID10 of 4x10TB NL-SAS attached to a dedicated RAID controller. No storage spaces or SAN/iSCSI techniques is used here. VBR is installed inside a VM to have an abstraction layer from the physical machine and to have the ability to test backups by creating VMs on the Hyper-V host.

My goal is to use a 20TB ReFS backup repository, where the file system is formatted with ReFS 64k cluster size.

However, since storage can not be pass through, the question arise where the backup repository should be attached. From my point of view, only three options are possible:

1. Attach the Hyper-v host itself as backup repository. The backup files are then just located on the Hyper-V host.
2. Create a ~20TB VHDX file on a ReFS 64k volume and attach it to the VBR VM and format this VHDX again with ReFS 64k.
3. Create a ~20TB VHDX file on a ReFS 64k volume and attach it to the separate storage only VM and format this VHDX again with ReFS 64k.

I'm sure that all three options should technically work, but are they supported and/or acceptable? Or are there any disadvantages?

Thank you very much!

Best regards
Burak

[HannesK]

Hello,
I would keep it simple and go for b. It's a consistent concept.

a and c split the backup infrastructure into multiple parts. yes, they will also work and I see no issues with support.

Best regards,
Hannes

[burak]

Hello Hannes,

thanks for your answer. Do you see any performance issue with the cluster size of parent and guest disks? As I remember, Veeam repository with ReFS should be formatted with 64k. Microsoft recommendation for Hyper-V VM storage is ReFS 4k. But I guess, using 4k in the parent disk will give performance penalty, shouldn't it?

Thanks!

[HannesK]

Hello,
not relevant from my point of view

Best regards,
Hannes

[nmdange]

I prefer to keep the ReFS repository separate from the VBR VM just because of past issues with memory usage or unresponsiveness during fast clone operations. I'd rather that not impact the VBR software itself.

[lpezda]

Hi,
I'm in the same situation as OP. I've the new server as Hyper-V host with VBR VM and additional 2-3 VMs (ex. AD 2nd DC). The server is a backup server, no production. I'm in small organisation and this purchase was big win for me and cannot afford to make it bare metal for VBR only.

In last few days I's doing research and cannot find a good, approved solution. Many questions but zero feedback after a while if it works or not. Three most often mentioned across the web are:
1) repository as a vhdx disk attached to VBR VM (easy to deploy and move but slow to access in DR),
2) Hyper-v host added itself as Windows server repository (kind of strange for me as VBR in VM will be contacting with host to access the it's storage as it could be another physical server),
3) create partition on host and use pass-through disk to VM (never use it but there are big no-no across the web about this)

What is the best way to make repository on Hyper-V host for VBR VM in 2023?

[HannesK]

Hello,
the best way is to avoid chicken egg issues and have an independent backup system.

As you say that this is not an option, you could go for 3 (less things that can break because no virtualization layer) or 1 (easier to manage). 2 is worst from my point of view and I would not do it.

Best regards,
Hannes

[lpezda]

Thank You for response.
Must say that I was leaning to option #2 before Your post. I don't want to use #3 as pass-through is no longer recommended by community in favor of vhdx, so I I'll stick with #1...

but I'll actively lobby for a new server only for VBR.