Introduction Your Kubernetes autoscaler might be solving a problem that doesn't exist, while ignoring the one that does. One pod pinned at 90% CPU, four identical replicas idling under 5%. It's assumed to be a capacity problem, and HPA scales out. But it doesn't