HPA creates demand; cluster autoscaling creates capacity

Each scaler changes a different object. Horizontal Pod Autoscaling changes replicas, Vertical Pod Autoscaling recommends or changes requests, and infrastructure scaling reacts to unschedulable Pods.

1Metric risesHPA compares observed metric with target.

2Replicas increaseNew Pods carry resource requests.

3Scheduler tests fitPods bind immediately or remain Pending.

4Node scaler reactsAutopilot or cluster autoscaler provisions capacity.

5VPA improves requestsBetter sizing makes future placement and cost more accurate.

The chain works only when requests, quotas and pool maxima leave a feasible path.

HPA creates demand; cluster autoscaling creates capacity

GKE autoscaling is easiest to understand by separating the Kubernetes contract from the Google Cloud implementation. HPA, VPA and cluster autoscaling solve different dimensions. HPA is driven by workload metrics, while cluster autoscaling reacts to Pods that cannot schedule because of capacity. The Kubernetes objects stay familiar, but GKE supplies controllers, infrastructure and safe defaults around them. This is why a team can move from an on-premises cluster without rewriting every workload, while still needing to redesign networking, identity and operational ownership for the cloud environment.

A useful inspection step is `kubectl describe hpa checkout -n checkout`. Read the output as evidence, not as a ritual: first confirm the desired object exists, then look at status conditions, events and the Google Cloud resource it represents. In production, capture the expected result in a runbook or automated check so an operator can distinguish slow reconciliation from a configuration error.

Production gotcha: Running VPA in automatic mode alongside HPA on CPU or memory requires careful coordination because both can react to the same signal. The safe habit is to verify quotas, regional availability and feature support against current Google Cloud documentation before rollout. Release channels control the version stream.