UMA - CPU Request (mCore) and Memory Request (Bytes) not showing correct values

Products

• CA Application Performance Management (APM / Wily / Introscope)

Issue/Introduction

Describe node output does not match CPU Request (mCore) and Memory Request (Bytes) metrics

Resolution

Below is the root cause as to why node cpu request(mcore) count is different in describe output with UMA.

Describe node output is aggregating the cpu requests metrics from the pods. It only aggregate the values from the Running pods only. It doesn't aggregate from (complete and error) state pods.

UMA also aggregate this cpu request node metrics from the pods but it aggregate these statues ( Running+completed+error). So that is why UMA reports more values then node describe command does. 

Why we consider the completed(status) for pod while aggregating at the node level because

 kubectl summarizes the status of pods as follows:

Then the pod.status.containerStatuses are taken into account where it is set to a Waiting, Terminated or Completed state. The Completed state is mapped to Running or NotReady if at least one container in a pod is running.

If wanting  the values of the describe node output and UMA to be same, then please raise a ER request for the same.