We keep hearing that it doesn’t matter if your machine is physical or virtual. Your software will still work just fine. That’s true most of the time but there are some exceptions. Monitoring is one of those exceptions. In truth, the monitoring tool will work and will give you accurate information but it will be meaningless.
Remember that in a hypervisor based virtual environment, the guest OS is typically unaware that the hardware has been abstracted and that resource scheduling is taking place to provide shared computing resources based on some preset business rules (some guests may be configured to get more resources than others).
In a scenario like this, a legacy monitoring tool that is targeted at the guest VM, may get false positives concerning resource availability. Typically you might see near 100% CPU consumed. This will be based on the telemetry coming back from the guest VM indicating that it is nearly out of resources, when in reality, there may be more resources available just not committed or allocated to that VM at that point in time.
To get a more realistic and complete view of what is actually happening, the monitoring tool would need to monitor the hypervisor and all of the guests correlating the telemetry form all of them and providing a more holistic view of the availability of resources.
Products like this are starting to emerge. Make sure that when you plan to migrate a server or application to a virtual environment you also plan for the monitoring requirements.