Metrics Server and the Resource Metrics Pipeline

kubectl top nodes returns <unknown> for two of three nodes. Your HPA reports FailedGetResourceMetric and refuses to scale. Both symptoms point to the same place: the resource metrics pipeline, a chain of four components where a single broken link takes down real-time visibility and autoscaling simultaneously.

Troubleshooting this pipeline is a core CKA competency (30% Troubleshooting domain). The chain runs from container runtime through kubelet to Metrics Server to the Metrics API. What follows traces each link, covers installation and verification, shows how HPA depends on the same data path, and provides a systematic checklist for when scrapes fail.