Warning message "...average CPU load..." since V6 Upgrade

[renatorichina]

We had to upgrade to VBR 12.2 and therefore to Veeam AHV V 6 because of the security flaws. Now, I have several Nutanix backup jobs showing the following warning in the job:

"The average CPU load of worker Embedded has reached 95% during the past 10min. Task performance may be affected".

With V 5.1 I never had this warning.

We use "only" an embedded worker, however he has already much more vCPU/RAM assigned than by default (which is 6 + 6):

8 vCPU
32GB RAM

Max concurrent task is also at the default of 4.

Any way, to get rid off this warning message from the backup jobs?

[ybarrap2003]

@renatorichina have you opened up a trouble ticket for this message? You can also try using a worker and disabling the embedded worker to see if you get the same error. I would still highly recommend opening up a ticket since this forum is not designed for troubleshooting issues

Workers doc link
https://helpcenter.veeam.com/docs/vbahv ... html?ver=6

[Gostev]

No need to bother support you will find a checkbox for this warning in the Configuration > Settings > Notifications tab.

[ybarrap2003]

Thanks @goster @renatorichina here is the link for the notifications tab - https://xxx.xxx.xxx.xxx/#/configuration ... ifications (replace xxx's with your appliance IP address

[renatorichina]

Thanks Pete, found it. However the warning strangefully no longer appears now on any job after one day. But in case it will again, I know where to adjust those settings. Seems to be new for Version 6 of Veeam for Nutanix I think. Never saw those settings before tbh.

[Gostev]

These settings existed in v5 as well actually in any case, now you know!

[Didi7]

Same here, we used v5 of the Proxa Appliance on 9 different NTNX-cluster-environments and never got this warning message. Since the upgrade to 12.2.0.334 (due to security issue), it is now common, that a warning message is presented in each backup job

No warning message is the exception. Well of course, you could disable the warning but does that really make sense?

[Gostev]

I believe it's meant to prompt you to consider increasing proxy appliance compute resources to reduce your backup window. Perhaps it did not trigger much in V5 simply because the code was not as optimized and could not fully consume all available CPU?

[Didi7]

Probably yes!

[Didi7]

The new Proxy Appliance for Nutanix defaults to 6 CPUs with 6GB RAM, but the Warning did not vanish, so we decided to upgrade the Proxy Appliance to 8GB RAM with 8 vCPUs and this might have fixed it now for the time being.

[lgamait]

I have the same message, my configuration has 8 simultaneous tasks, 16 CPU and 10 GB RAM, and the CPU consumption is 97%.

I have an open case #07429283, and support asked me to update the Nutanix plug-in to 12.6.0.636, and I still have high CPU consumption.

My environment has about 150 VMs.

If anyone has found a solution, please share it here.

[Gostev]

Normally it's not a problem you need to solve if you get good backup performance and fit your backup window.

High CPU usage on proxy simply means it's the current data processing bottleneck. Other times bottkeneck can be speed of production storage, backup repository, or available network bandwidth.

See also VMware backup FAQ

[ronnmartin61]

After diligently researching the apparent CPU load statistics discrepancy from v5 to v6 it does appear that we are accounting for cycles that do not represent load from active backup processing. This will be corrected in the next release.

[V.ingenc]

When I assign additional vCPUs to the AHV Proxies and Worker Nodes in Veeam, Nutanix only adds another vCPU but does not increase the number of CPU cores. In Nutanix, however, increasing the number of CPU cores rather than just adding vCPUs is more effective in enhancing performance. Simply adding vCPUs does not provide the same performance benefits, as many workloads, including those on Worker Nodes, rely on additional cores for improved processing power.

For example, it would be more beneficial to assign 1 vCPU with 6 cores rather than 6 vCPUs with 1 core each to maximize efficiency.

Could you please confirm if this is indeed the best practice for optimizing performance in Nutanix with AHV Proxies and Worker Nodes?

[ronnmartin61]

@V.ingenc adding cores may improve performance for other workloads however in our testing we do not see any material difference between the two approaches and have opted for simply adding single core sockets as it also provides the added benefit of CPU hotadd so as to reduce required reboots. We have spoken with Nutanix about this in the past and they have not expressed any explicit preference.

[V.ingenc]

Thank you for the detailed response and insights on core sockets and performance. I appreciate the clarification!

[ronnmartin61]

You're very welcome!