I was wondering if anyone has had experience with the following:
We have numerous VMs in remote locations but two of them in particular seem to randomly task the CPU at 100%. They are Oracle database servers and I'm not sure if that has anyting to do with it, or if the problem may simply be that they are heavy resource hogs...
When the CPU is tasked at 100%, it is the Oracle.exe process that is using it up...
Doing one of two things seems to bring the CPU down to a flatline...
1. connecting to the ESX server that is hosting the troublesome virtual machine via putty and restarting the mgmt-vmware service... when doing this, the ESX server and it's managed virtual machines all go grey (into a disconnected state) in the Virtual Center for a moment... as soon as they come back to a regular state, the CPU drops down to 0%.
2. disconnecting and then connecting the ESX host in question from the Virtual Center
Doing both of these seem to remedy the situation.
I think I'm going to try simply right clicking on the ESX host next time and selecting "Reconfigure for VMWare HA"
We are using ESX 3.5.0 Rel 64607 and Virtual Center 2.5.0.64192
Andrew
Try this to adjust the vmoverheadgrowth limit and more info below.
http://communities.vmware.com/message/952326
-KjB
I agree. We had to set this even though U1 fixed the issue. Some patches in June failed a regression test and the issue reared it's head again. Not sure about U2 or U2v2.
Ben
Hi Guys
Many thanks to both of you... I will research and try it out in test...
All the best,
Andrew
Don't forget to leave posts for helpful/correct posts.
-KjB
Will do KjB
Just need some time to test as the issue is quite sparatic and random.
Thanks a lot for the suggestion and will get back soon.
Andrew
Are these VMs running Windows 2000? We have an issue with Windows 2000 guests with multiple VCPU's that show 100% CPU utilization. They were also P2V VMs. Our solution was to rebuild to 2003 instead of trying to troubleshoot this for hours on end.
Philip Arnason
Hi All
I just wanted to post an update here on this...
After changing one of the data center's OverHead settings to 5, I watched the virtual systems in it for a few days.
I would still see them go into a Red alarm indicating that they were using 90%+ of the CPU at times... sometimes they were guest memory usage alarms...
In the case I just witnessed, I looked at the server to see if it was actually doing anything intensive and it didn't appear that way... however the CPU was up around 80 to 90% usage in the task manager...
After checking on that, I went to the data center and disabled DRS.
As soon as I did this, the CPU usage on that particular server dropped to a flat line.
Doesn't look like the overhead setting helped in this case unfortunately... however I will keep an eye on it a while longer before switching it back.
Andrew