Monitoring Cluster and Application Resources

A Pod is stuck Pending. kubectl describe shows FailedScheduling: insufficient cpu, but kubectl top nodes says every node is under 50% utilization. The numbers do not add up until you realize kubectl top shows usage, not requests, and the scheduler only cares about requests.

Monitoring resource usage falls within the CKA Troubleshooting domain (30%). The exam expects you to quickly identify resource-constrained nodes and pods, read quota status, and trace why a workload cannot schedule. This lesson covers the monitoring fundamentals: how the metrics pipeline works, how to use kubectl top effectively, and how to identify resource-constrained workloads. The dedicated deep dives on Metrics Server and the Resource Metrics Pipeline and Troubleshooting Resource Quotas and LimitRanges follow in this module.