Memory overcommitment and pressure

[Sb1314]

Hi all!

Running 9.5.0.3254

Monitoring Hyper-V cluster with two hosts, each with 128GB RAM.
Geting cluster overcommitment memory error and host average memory pressure for 1st host.

1st host runs 5 VMs of which 4 uses 2GB RAM and one uses 4GB, total 26,69GB Hyper-V Services Memory Consumed per Veeam - Task manager shows 18% memory consumption.
2nd host runs 6 VMs of which 4 uses 4GB RAM, one uses 2GB and one uses 3GB, total 30,12GB Hyper-V Services Memory Consumed per Veeam - Task manager shows 21% memory consumption.

Host1:
http://pasteboard.co/GQRQ60S.png

https://pasteboard.co/GQRP3g1.png

Host2:
https://pasteboard.co/GQRPxvs.png

Cluster overal:
https://pasteboard.co/GQRQn4a.png

All VMs are in Healthy status, what am I missing here?

[Shestakov]

Hi and welcome to the community!
Veeam ONE simply triggers the alarm if cluster status displayed in the System Center Virtual Machine Manager 2008 Administrator's Console change from "OK" to "over-committed" after a cluster refresh operation completes.
Please check this Microsoft KB article.
Thanks!

[Sb1314]

Hi Shestakov,

First, thank you for comment.
We don't use SCVMM at all, Veeam is connected directly to Hyper-v servers.
We only use Failover Cluster Manager and Hyper-v Mavager and Veeam for monitoring.

Regards!

[Shestakov]

Got it, let`s check if condition of Hyper-V over-commitment is fulfilled.
Could you check if the sum of free slots in the entire cluster is greater than the sum of slots in the largest host (both free and used)?
By the way, for some reason images from your first post were not displayed. I`ve pasted them as links.
Thanks!

[Sb1314]

I don't know why forum don't displays inline images:
https://pasteboard.co/GQRQ60S.png
https://pasteboard.co/GQRP3g1.png
https://pasteboard.co/GQRPxvs.png
https://pasteboard.co/GQRQn4a.png

Are you talking about this calculation?
https://blogs.technet.microsoft.com/scv ... ager-2008/

There is one strange thing, 4.10. when I installed this cluster and added it to Veeam without running any VMs I got 100% overcommitment alarm.
I can't show it to you any more because I have removed cluster from Veeam and that purged perf counters and logs.
Right now I'm on 20% memory usage on both nodes and cluster average.

[Shestakov]

Memory over-commitment is not about memory usage, but about memory allocation. You may have 1 VM with huge memory allocation what makes slot size big enough to turn the cluster into "over-committed" state.
I would make the calculation provided in both your and my links.
Again Veeam ONE doesn`t operate with own logic, but takes the over-commitment event from Hyper-V.

[Sb1314]

One more thing...
Every now an then I get "Bad Hyper-V username logon attampt" alarm and after investigation I found this:
https://windowsserver.uservoice.com/for ... rting-bits
https://social.technet.microsoft.com/Fo ... rum=ws2016

If you manually start BITS service you get this alert in Veeam.

So you monitor event log for errors and warnings, tell me which windows event triggers over-commitment and host memory pressure alarms?

[Sb1314]

Can you please explain this:

Cluster memory overcommitment
Based on event VeeamHvClusterReserveStateOkEvent, VeeamHvClusterReserveStateErrorEvent
When placing a virtual machine in a failover cluster, the placement process calculates whether the new virtual machine will over-commit the cluster. If the action will over-commit the cluster, the corresponding alarm will be fired.

https://helpcenter.veeam.com/docs/one/a ... tml?ver=95

Who generates this events, everything else is based on MS events or perf counters.

[Shestakov]

Can you please explain this:

Cluster memory overcommitment
Based on event VeeamHvClusterReserveStateOkEvent, VeeamHvClusterReserveStateErrorEvent
When placing a virtual machine in a failover cluster, the placement process calculates whether the new virtual machine will over-commit the cluster. If the action will over-commit the cluster, the corresponding alarm will be fired.

https://helpcenter.veeam.com/docs/one/a ... tml?ver=95

Who generates this events, everything else is based on MS events or perf counters.

Periodically Veeam ONE checks HyperVClusterReserveState and once state is changed to 2(Overcommitted), Veeam ONE generates the event to fire the alarm.

So you monitor event log for errors and warnings, tell me which windows event triggers over-commitment and host memory pressure alarms?

Some alarms are event-based. some performance-based.
Memory pressure is performance-based one it fires when Memory Pressure counter is above defined threshold for defined period of rime(15 min by default).
You can check and change under "Rules" tab of the alarm.
Thanks!