Incorrect CPU utilization reported

[rwazar]

Greetings all.....

Here's my issue: I P2V'd an old dell desktop running Win2k Pro. The migration went well other than failure to install VMware Tools (guest os = other). I got past that by sharing the iso. Anyhow, VC is reporting this VM's CPU utilization as 100% with timely dips down to 93% every 6 minutes. The web interface for the ESX host shows the same maxed out utilization as well. This particular server is not heavily utilized so I logged into the console to see what was up. Here's the kicker...The windows performance shows nearly idle cpu usage.

Someone is lying. Any idea as to what may cause this?

So far I removed the CD and floppy from the VM (no change). Can't use lsilogic (boot failure). The parallel port is connected and I cant disconnect it unless I shut down the guest.

Version info: ESX 3.0.2 update 1, VC 2.02, ESX hosts 4xDL385 G2's

[mstahl75]

Number of CPUs and the HAL on the server?

My initial guess is that you are running with 1 vCPU and a multi-processor HAL. If that is the case downgrade the HAL to uni-processor and see if the VM reports more accurately.

[rwazar]

I just checked that, and you are correct. 1vcpu assigned, ACPI Multiprocessor PC.

It wont hurt anything if I change the driver to uniprocessor will it? Will it require a reboot?

[rwazar]

Also...Uniprocessor is NOT in the compatible hardware list. Is it ok to select it from "Show all hardware of this device class"?

Forgive my ignorance with windoze, it's not my forte.

[mstahl75]

I haven't had to mess around with this for some time but you should have the option for ACPI Uniprocessor PC. At least that is available on 2000 Server and I assume it would be there for Professional also. You might need to go through the update wizard and select to show all hardware for the device class.

One other thing that I would recommend would be to disconnect any devices that are not being used and since the machine is one that has been P2V'ed to clean out stale hardware that isn't on the machine anymore.

Bring up the command prompt and do the following:

set devmgr_show_nonpresent_devices=1

devmgmt.msc

In the device manager, be sure to show hidden devices and get rid of things that aren't on the VM. Also, changing the HAL requires a reboot as far as I recall.

[rwazar]

Either way i'm gonna snapshot it first just in case I end up hosing it.

I did clean up the hidden nic's but thats all I've been cleaning up. Is it safe to run through the list and delete all the greyed out stuff or just the obvious things like video adapters, drives, and nic's? Of course I could probably just revert to the snapshot if I do delete too much and screw it up.

Now if only these people would go to lunch so I can jack with it.

Thanks for your help, it's greatly appreciated, I bow before your infinite awsomeness.

[mstahl75 · Accepted solution]

I've done a bit of both. Early on I erred on the side of caution and only did the most obvious stuff. I got to were I delete most things that showed up in the list. It should[/i] be fine to delete almost everything -- it is only supposed to be devices that aren't present anymore but if in doubt it probably won't hurt to leave things.

But a snapshot is probably a smart idea just in case :smileygrin:

[rwazar]

You are my new hero.

The change to ACPI uniprocessor has things back to normal. :smileygrin:

.....wonder what else I can break today....

I do have another issue pretaining to VM files on datastores that I can delete......I'll post another topic on that when I get a chance.