Looking at the kubevirt_vm_resource_requests metric, we see that most VMs do not have a value for memory and CPU values.

We must report a value for every VM. If its not reported as requests, then there should be a different way to get this value.

This is highly important to be able to calculate utilization (used out of total allocated).
and it is also very important for capacity planning and right sizing calculations.

Looks like there is also an issue with the actual values.
For VM "ubuntu-bpfi88jjc64bshyw" in namespace "jwang" I see that the allocated CPU is 2GBi , but the metric report 2.15GBi. See screenshots. 
I think the data for this metric is not taken from the proper place.

1.
2.
3.