Horizontal Pod Autoscaler: Metrics-Based Scaling

You sized your Deployment for Friday's traffic spike, and now you're paying for those extra replicas the other six days. The HorizontalPodAutoscaler fixes that: it watches metrics, computes the replica count your workload actually needs, and adjusts the Deployment's scale in real time. No cron jobs, no manual intervention, no wasted capacity.

HPA is a core competency in the CKA Workloads & Scheduling domain (15% of the exam). You need both the imperative shortcut (kubectl autoscale) and the declarative autoscaling/v2 manifest. Beyond the exam, understanding the scaling algorithm, behavior configuration, and custom metrics pipeline is what separates "it autoscales" from "it autoscales predictably under load."