VSS Backup on Hyper-V mixed with Nutanix Native Data Protection


Badge +9
Good day,

We are currently looking at an Nutanix implementation where the customer has a policy of using an VSS Backup for all their workloads, and now want's to cover the Nutanix platform with it where they are running Hyper-V.

As they also are looking at using an off-site cluster for DR we were thinking about using the built-in Async DR to replicate the data across the wire but we ran into a slight "Huh?" when gathering information that we havn't figured out yet.

Under the Prism Web Console Guide - Data Protection the following section is seen
"Warning: Nutanix native snapshots and replication is not supported for VMs configured with Hyper-V host snapshots."

Now, since we are using a VSS solution for Hyper-V backups we sort of always Snapshot/Checkpoint the VM during the Window. As Server 2012R2 do not yet have any Change-Block-Tracking for SMB3 Shares this Window could be rather large. Will we run into a problem if the Nutanix Snapshots happends to be scheduled around the same time?

Basically we are looking into the information if it's doable to mix the VSS Backup with the In-Built DR of Nutanix without running into any issues.

Kind Regards

16 replies

Userlevel 7
Badge +30
Are you specifically talking about using VSS and a "VM level" backup solution? Or an "in-guest" backup solution?

The in-built DR for Hyper-V does not yet support VSS in itself (this is on the radar near term), and you are right that hyper-v in general doesn't support a CBT-like API for SMB3 on 2012 R2, so there could be some challenges in this setup. Tread carefully.
Badge +9
We are talking about VSS with VM Level Backup.

The support for Hyper-V VSS for the in-built DR is not an issue right now (As SQL and other workloads are in-guest), our main concern is that we will run into problem with the Native Snapshots and the Snapshots/Checkpoints that get created during the backup windows (As these Windows are a bit large do to no CBT). The larger backup Windows is also not an concern of us as long as we still can do an in-built DR while the VSS Backup is fetching blocks from NDFS.

It is a bit unclear if the meaning with "Hyper-V Snapshots are not supported" also covers the actual process of Snapshot/Checkpoint during the backup. If I recall correct, this period of time when the .avhdx file is there is rather small and only applicable until the Nutanix Storage has done a NDFS Snapshot, if we compare it to an regular Snapshot/Checkpoint which is persistent until we remove it.
Userlevel 4
Badge +18
I'm double checking on this now and will report back. Would the backup application being used happen to be using VSS for SMB shares? This would take a container wide Nutanix snapshot and therefore make any VM snapshots short lived.
Badge +9
Hi,

And huge thanks for looking into this!

The backup suite in question is actually Commvault, but do the immense amount that they backup right now in the whole company, and reports of issues with upgrading to V11 they have been adviced to hold of the upgrade until the new SP comes out which means no IntelliSnap support for Nutanix at the moment.

As far as I know they do invoke the Hyper-V VSS which in return should go through the normal Remote SMB3 provider and in turn take an Nutanix SnapShot. We can see I/O in PRISM during the backup windows without seeing any .avhdx or any "Backup" state which should point towards that they indeed do integrate correctly to the Remote VSS stack.
Badge +9
Good morning,

Any update regarding this question?

Regards,
Userlevel 4
Badge +18
I confirmed that if we detect a Hyper-V checkpoint we will mark the VM as CBR incapable meaning we won't let you add it to a protection domain. In this case however the checkpoint would be short lived so you shouldn't be blocked. This leads to a different question of interop between the container snapshot taken with commvault (VSS for SMB shares) and async protection domains. Waiting to hear back on this second question but will reply back ASAP.
Badge +9
Good day,

We just had the second cluster setup and are about to configure the Protection Domains, any new information regarding this matter?

Kind Regards
Badge +9
and  And new information on this, or if you are able to point me towards me a contact?
Userlevel 7
Badge +30
I'll defer to Mike since he's closer to this issue.
Userlevel 4
Badge +18
Sorry for the delay. The official response is that the combination of VSS for SMB share level snapshots and native protection domain snapshots is not supported. Reason being is that Hyper-V checkpoints are not supported in combination with native protection domains. We cannot guarantee that the schedules will not overlap and lead to an issue where a checkoint exists due to the use of VSS for SMB.
Badge +9
Hi,

A follow up question on this that have been raised.

In the case of utilizing CMVLT IntelliSnap with Nutanix on Hyper-V my understanding is that we are only able to take crash-consistent backups, are this correct?

Also, utilizing IntelliSnap with the above setup, are we able to take advantage of the CBR functionality from NDFS? I have seen occasions where the VSS Streaming Backups take a long time and by looking on the metrics is somewhat causing havoc on the Content Cache Hit Ratio as we need to Read everything once.

Another question in regards to the current way of performing VSS Streaming Backups on Hyper-V:
- Utilizing VSS Streaming Backups have a Best-Practice of maximum 50 VMs per Container. Is this still applicable and what is the reasoning behind this in terms of problems that might be seen? Also, is this also applicable when utilizing IntelliSnap on Hyper-V?
Userlevel 4
Badge +18
to help with some of the answers.

I think there is a recommended limit for intellisnap because a snapshot is being created which encompasses the entire container. This limit I believe has been raised to 200 with recent AOS releases, although 50 is a safe number.

There isn't CBR for Hyper-V today, so the observation regarding full reads is still true. Server 2016 will have native CBR and Nutanix is also developing native CBR capabilities, but that is not available today.
Userlevel 4
Badge +21
hi,

Mike is right about the CBT comment but Commvault will use brute-force CDP via checksums. 200 number is right but I would plan around 100 because when you turn your back you know more VM's are going to land there.
Badge +9
Hi,

And thank you both for the input.
If I recall correctly, the "Data Protection and Disaster Recovery" Solutions brief, that seems to be written for AOS 4.6 still recommends a maximum of 50 VMs per Container where VSS Backups are taken. But the Protection Groups in the Native Protection has been increased to 200.
Is this information still valid, or have it actually been raised to 200 VMs per container where VSS Backups are taken?

If I understand you both correctly and put all information together we have the following situations:

In case of VSS Streaming Backups (CMVLT, Veeam), we are limited to 50 VMs per Container (Still going by the latest Solution Brief and Veeam Best Practice until is has been confirmed) with no CBT at all. VirtualMachines are Application-Consistent as long as Integration Services on the VM is installed.

In case of IntelliSnap on Hyper-V, we are limited to 200 VMs per Container, but instead of CBT we can use CRC. VirtualMachines are Crash-Consistent.

In case of Native Nutanix Data Protection, we are not limited to any amount of VMs per Container, but instead limits us to 200 VMs per Protection Group with CBT support. VirtualMachines are Crash-Consistent.

Is this understanding correct when putting it all together?
Userlevel 4
Badge +21
If I recall correctly, the "Data Protection and Disaster Recovery" Solutions brief, that seems to be written for AOS 4.6 still recommends a maximum of 50 VMs per Container where VSS Backups are taken. But the Protection Groups in the Native Protection has been increased to 200.
Is this information still valid, or have it actually been raised to 200 VMs per container where VSS Backups are taken?

It's moved to 200 for VSS backups. That is what has been tested. With SMB3 it takes a snapshot of the whole container. It would still be wise to plan for under this number.


In case of VSS Streaming Backups (CMVLT, Veeam), we are limited to 50 VMs per Container (Still going by the latest Solution Brief and Veeam Best Practice until is has been confirmed) with no CBT at all. VirtualMachines are Application-Consistent as long as Integration Services on the VM is installed.
The Veeam guide is long in the tooth, I can't comment on the Intergreation services but I think that is only a veeam specfic thing.

In case of IntelliSnap on Hyper-V, we are limited to 200 VMs per Container, but instead of CBT we can use CRC. VirtualMachines are Crash-Consistent. - still confirming with IntelliSnap

In case of Native Nutanix Data Protection, we are not limited to any amount of VMs per Container, but instead limits us to 200 VMs per Protection Group with CBT support. VirtualMachines are Crash-Consistent. CBT is only VMware at this time. Hyper-v will get it in the next release and we are working on something with Nutanix Guest Tools to be hypevisor agnostic.
Userlevel 4
Badge +21
Commvault crash consistent with SMB intellisnap. App consistent without intellisnap

Reply