Veeam Proxy vSAN high read latency

[unsichtbarre]

Case #02734805
I am dealing with a situation where we have 2 Veeam Proxies configured in Virtual Appliance (hotadd) mode protecting a vxRail vSAN environment. The proxy servers run on the same vSAN/Cluster as all of the VMs they are protecting and replication - however we are seeing read latencies of 80ms. - 150ms. on the Veeam Proxy servers and backups/replication are extremely slow and large VMs (1TB+) are simply not finishing.

I am presently trying NBD to see if transitioning data flow to the individual VMs and ESXi will improve anything, although this strategy is clearly not best-practice, I acknowledge.

Can anyone suggest strategies or diagnostic measures we can use to localize the latency issues?

THX,
-JB

[rennerstefan]

Hi,

what does the vSAN disk backend look like?
Is it all flash or hybrid?

Thanks

[PetrM]

Hello,

This article describes the most common vSAN performance troubleshooting practices with different examples. It's worth looking through it.

Thanks!

[unsichtbarre]

Thanks for the responses rennerstefan and PetrM

Very useful article!

I will ask owner about disk type, however this problem is recent. Something changed.

[gswan]

Interestingly enough I am seeing the same thing on my organizations VMware cluster. We are currently running:
8 vSAN Ready Nodes with a Hybrid disk configuration.

We started to see this issues after upgrading to vSphere/ESXi/vSAN to the most recent versions in late august. We are currently running Veeam version 11.0.1.12621P20220302

I have a case open with VMware as well as with Dell, both entities are not finding any configuration issues in my environment that would cause these issues. I do see the high latency spikes during our Veeam B&R backup windows, which we are currently using hotadd mode as well.

[PetrM]

Hi Gregory,

Instead of looking for configuration issues that VMware and Dell support teams could find in your environment, I'd suggest asking them to investigate deeper the specific issue with high latency spikes during the backup job. Also, you may contact our support team to run some performance tests to isolate the problem at the level of your infrastructure.

Thanks!