VMware Cloud Community
burtonboarder23
Contributor
Contributor

PSOD after upgrading VM to 6 CPU

After completing our 5.5 vCenter + Host upgrades, we wanted to do some performance testing. We added some additional CPU's to one of our vm's and as soon as the VM came back online the host it was on suffered a PSOD. Migrated the vm to the other host in the cluster and turned it on. Same thing happens. After doing some googling, i found a few other cases where others have suffered similar issues. My question is, is this something many people are having an issue with? My concern is that we have multiple production sql boxes running > 5 cpu.

I've attached a shot of the PSOD.

Tags (2)
0 Kudos
6 Replies
rbos3
Enthusiast
Enthusiast

Hi,

I would say check the firmware versions of your ESXi host and upgrade them when possible. Next, have you tried assigning 8 vCPUs/cores to a VM instead of 6 and see what it does?

Looking forward hearing from you,

Cheers,

René

--- If you found this information useful, please consider awarding points for "Correct" or "Helpful". Thanks! Visit my blog at http://snowvm.com ---
0 Kudos
beckham007fifa

possible reason could'hv been CPU overcommitment, how did you make the changes, did you touch only the sockets or the core per socket as well?

Is that issue resolved?

Regards, ABFS
0 Kudos
burtonboarder23
Contributor
Contributor

The firmware is up to date with the latest from HP. When we ran the test I just added 6 virtual sockets. This was when i got the PSOD 2 times. Last night i added 6 vCPU to another vm but this time used 2 virtual sockets and 3 cores per. The vm's have been running stable since.

What concerns me is if one of my admin configures a vm for 6 virtual sockets / 1 core and the PSOD comes back.

0 Kudos
beckham007fifa

There are some best practices to allot vcpus, which is highly depended on even and odd's numbers of core available. if you have 4 physical cores available, you can easily allocate 6 virtual ones. If you have 8 physical cores then you could allocate 12.

Do you have NUMA vms in place?

check this best practice document, which will guide you, and the reason it is getting PSOD is certainly due to cpu overcommitment.

Regards, ABFS
0 Kudos
burtonboarder23
Contributor
Contributor

We do not have vNUMA in place however I plan to look into this. Was there suppose to be a link to a best practice guide that you referred to before?

0 Kudos
beckham007fifa

0 Kudos