Mixing Backup with Compute Storage with HyperV Cluster

[stevenm]

Hi All

I have a customer who is looking to purchase a HPE MSA 2052 iscsi to replace their slow current solution and are being cheap about it. They want to use 1 raid set (SAS10k) for VM storage and the other raidset (NLSAS) for backup storage.
They have 2 sites so will duplicate this and replicate on the storage layer.

I'm not keen on the idea as ive always had backup storage separate and consider it best practice and would at least see them target a NAS or windows server full of disk, but because they are mirroring the storage at both sites they feel that will be the fail-over scenario and not too concerned bout mixing it.

Has anyone had experience with this setup, how they've set it up in HyperV? I know ill do CSVs over iscsi for the hyperv cluster but has anyone also targetted CSVs with veeam for a backup repository?
The Veeam server will also be a VM on the same hyperv cluster/SAN as the only physical servers are to be hyperv hosts.

Thanks

[vClintWyckoff]

Hey Steven,
I'd recommend to first off start by making sure that the client understands that this is really not the best setup and advise that they purchase either another device or another Windows Server with a good RAID controller, this will give the best performance for backups and restores for their money. You would also want to be sure to configure Backup Copy Jobs to copy the backups from Site A to B and Site B to A. This way the customer has optimal protection in both directions and is compliance with the 3 (Copies of their data) - 2 (Different types of media) - 1 (offsite) rule.

has anyone also targetted CSVs with veeam for a backup repository?

Here's a pretty good thread on this topic as it relates to Storage Spaces and S2D

[stevenm]

Thanks Clint.

Im not keen on the idea either and at this point am struggling if its even feasible to work if Veeam cant use CSV as a backup repository.
They currently have backups going to dedicated Dell Storage servers and backup copies are used to do cross site backups. Trying to push to keep this in place at least but all their hardware is over 3yrs old so they are downsizing on costs etc.

The hosts run 2012 r2 so wont be upgrading to 2016 anytime yet and ive looked into previously using them for storage spaces and need hardware upgrades. Currently using SOFS platform for HyperV Storage.

[vClintWyckoff]

Just as a follow-up I'm doing some internal research with support to see what their take is on this configuration, I will report back shortly.

[vClintWyckoff]

I checked with support on this configuration and I have personally tested this on my ws2016 cluster (no 2k12r2 cluster in my lab ) and it works. I tested migrating the disk from node1 to node2 with an active backup and hard rebooting the host during the backup too. It would be pretty quick and easy for you to test this out to see how it works with your 2k12r2 existing cluster.

In the job (C:\Program Data\Veeam\Veeam\Backup\JOBNAME\Job.log) log you'll notice for repository connection we've bound to the cluster instance name (in the image below this is Hyper-V.aperaturelabs.biz) and not the owning node of the disk. In the past with in the add a repo wizard there were issues with the populate not working correctly which seem to have been resolved, at least for my ws2016 cluster.



Support also confirmed that they have seen customers configure their repository's in a WSFC w/ CSV without issue. But from a QA perspective this is not a repository configuration that has been officially tested and thus is not officially supported.

[stevenm]

Thanks Clint good to see it works. Not being supported is a good turn off.

I cant test if it works on 2012 R2 as they use SMB storage, no CSVs to test expanding.

Reviewing their quote the NL-SAS tier costs similar to buying a decent windows backup repository rack server so ill push that idea instead.