We are observing performance data discrepancy for CPU in APM tool and OpenShift Console
Release : 10.7.0
Component : APM Agents
The platform monitors aren't suitable for monitoring CPU utilization. Platform monitor metrics are based on CPU clock ticks. These APIs are not reliable on virtual platforms like VMWare or Docker, where CPUs and the memory of an Operating system are partitioned. Most of these APIs assume that the entire Nodes available to the container, and they end up showing incorrect container utilization values. Our recommendation is to use the Container CPU utilization metrics instead. You can either use UMA to monitor the CPU and other performance Containers natively or use UMA to ingest CPU metrics from Prometheus if you are using Prometheus. APMIA Open Shift Extension can also be used.