Update 4 Issue?: CPU Usage Graph Show 60% CPU but inside VM Task Manager Shows 3-10%

[s1xth]

Hey guys..

Hoping someone else has noticed this. I have a host running with a freshly updated U4 release running three VM's. One of the three is showing abnormal CPU Usage on the VIM Client graphs, but when I sit inside the VM the CPU usage is normal abotu 3-10% with a little high spikes every now and than. Has anyone else noticed anything weird like this? I restarted the VM and the host and its still showing this. VMware Tools are updated etc. Its a W2K host with 2vCPU's and 3GB of ram.

Thanks!

[runclear]

1. Whats the Hardware

2. What does "esxtop" output?

[s1xth]

Thanks for the reply. I just noticed I posted this in esx instead of esxi, but it shouldn't matter to much. I will run that command from the console, and see what I get.

And btw this is running on a Dell pe r200, I know its not on the hcl but I haven't had a problem with it, all the way from U2 and U3. plus it seems like it is only this single VM that is displaying different cpu graphs than what is really going on inside the VM...

Sent from my Nokia E71

[runclear]

was this VM a P2V?

I would try to change the VM to 1vCPU and see what that does.....

[s1xth]

Yes it was a P2V.... why do you ask? I will change it to one cpu tonight.

Sent from my Nokia E71

[runclear]

I've p2v'ed a few 2k multi proc servers, I recall one that had similar

issues that you describe. I believe it has something to do with the

multi processor hal.

Sent from my mobile device

On Apr 8, 2009, at 5:41 AM, "s1xth" <communities-emailer@vmware.com

[s1xth]

Interesting. Do you happen to remember what you did to fix it? (change to 1vCPU etc)? thanks!!

[runclear]

It's been a while, so im really not sure if that was it or not - I tried searching the forum for others, it looks like others had similar issues as well, but no clear resolution path....

[ian4563 · Accepted solution]

Every single Windows 2000 multi proc machine that I have P2V'd has had this issue. No exceptions. The odd thing is that Windows 2000 VM's built from scratch are fine. Changing the HAL to single proc will fix the problem, but if you really need multiple procs (as we do) then the only option I've found is to play with the advanced VM option "monitor.idleLoopSpinUS". A value of 2000 is a good compromise for us.

[s1xth]

Very good information. Thanks so much, it is odd, because another W2K server from scratch is fine, exactly like you said. This is the only W2K vm that I have and is showing this. I will try what you said and see if that helps. I will post back my results.

Edit- I did some more searching and found some similiar posts like you were talking about. Seems like alot of people have reported this issue, I wonder why it is back now in U4, its been fine in U2 and U3. I am going to change the VMX file to monitor.idleLoopSpinUS = 1000 after hours one night.

I dont see this line anywhere in the VMX file, I am going to add the line at the end of the vmx file, see if that helps.

[s1xth]

Ian,

I just want to thank you, that worked perfectly. I added that line to the VMX file, set it to 2000 for now, and now the client is showing the correct CPU usage. Cant believe this isnt more of an issue that Vmware isnt fixing. Thanks again for the GREAT information!

Well setting it to 2000 helped, but it wasn't enough, it was still showing the wrong usage after a while, I changed to 1000 and that helped more and is closing the gap more and more, but its still showing more than it should by about 10%. I am thinking of dropping it to 500 and see if that does it completely.