When trying to do some science to answer this question, I took a Windows 7 guest on a Server 2008 R2 host being managed by SCVMM 2012 and ran Prime95 on it to just generate some CPU usage.
Here's the Guest:
The Hyper-V host shows 12%, which is 1/8 cores (which is what is allocated), so that's looking correct:
But SCVMM is showing 0%:
I have left the stress test running for a long time, thinking that maybe SCVMM averages out over a long time (I thought it was 9 minutes, but I've been known to be wrong; just don't tell my wife).
Why is SCVMM showing 0% when everything else seems to disagree?
Best Answer
From the SCVMM Engineering Blog:
So SCVMM uses WinRM (which is the Windows implementation of WSMan) to gather those performance counters. Can you establish a WinRM session from your SCVMM server to the Hyper-V host? Can you then query those counters remotely? (They may be under the Virtualization namespace, not the CIMV2 namespace.)
Is there anything in the logs about SCVMM not being able to collect data for those counters?
Also, give "winrm quickconfig" or "winrm qc" a try on your Hyper-V host if it's not already enabled and listening.
Test remote WinRM functionality
At an elevated command prompt type:
This should produce output similar to below:
If instead an error such as below appears, this means WinRM is not set up correctly on the remote machine, or there is something preventing communication over the WinRM port between the two systems. This could be a firewall or antivirus/malware programs.
In this event, test local WinRM functionality on the remote system. If WinRM is configured correctly on the remote system as well the cause is most likely network communication between the two machines. Troubleshoot this as you would any network issue.