VEEAM with NUTANIX + Windows Server 2016

[taurus1978]

Hello

we have a setup where we use NUTANIX with Hyper-V 2016 as the hypervisor.

The VEEAM Infrastructure is configured as follows:

-1 NUTANIX Cluster with 5 nodes
- 1 virtual backup server
- 5 virtual repository servers on the nutanix nodes (windows) with 1x 7 TB iSCSI Volume. Backed by a LENOVO V3700 iSCSI Storage
- Cluster is on win 2016
- Backup runs in "on-host" mode.


Now we have some performance issues during backup. The overall performance of a job with only one VM is running at about 50-70MB/s. When we look into a job with 2 or more VMs we see that every .vhdx of a VM is backup with a speed of around 35-45 MB/s. So cascading the overall performance of the job to a multiple of that with a single VM.

But 35-45 MB/s for every .vhdx seems really slow at all.

Backup job shows the source (i.m.o nutanix node or VM) as the bottleneck with 99%. So we think the problem is somewhere around the nutanix nodes or VMs.
But when we look to the nodes or the vms with a windows perfmon we dont see any performance spikes or overcommitment of resources.

so my questions are:
Is it possible to tune the performance per .vhdx in one job. ?
Any user on the forum with experience in such environments ?

Thank you for any inputs.


best regards,
Patrick

[Mike Resseler]

Hey Patrick,
When it is the source it mostly means that the host cannot do its job very well as the proxy. Do these hosts have enough resources available to do the work? Or are they very well loaded?

[taurus1978]

Hello Mike,

thats the weird thing. They dont have that much utilization. And ressources are not over provisioned at the host.

[Mike Resseler]

Patrick,

It is going to be very difficult to troubleshoot this without seeing what is happening in the logs. From the first sight, I don't see anything specifically wrong from your explanation. Can you log a support case? Post the case ID here (and the outcome after investigation of the engineers). They will be able to deep-dive in the logs and see if there is a bottleneck somewhere
Thanks
Mike

[taurus1978]

Hello Mike,

thank you. Will log a case. And post the case ID.

[taurus1978]

Hey Mike
case opened, Case ID 03422442

[taurus1978]

Hello Mike,
just for your information. The solution for us was to upgrade the VEEAM Management Server to 9.5 U4 on Windows Server 2016 and SQL Express 2017.
Now we get around 100-115 MB/s per .vhdx which, in my opinion, is sufficient. There seems to be some optimizations within win 2016 and/or veeam 9.5 U4 from which the performance in hyper-v environments benefit. Maybe you have an idea to that.

Thank you.
very best regards,
Patrick

[Mike Resseler]

Patrick,

Thanks for coming back to us and giving the resolution for your problem. I am sure that many others will benefit from it also!

Thanks
Mike