Autoscaling prerequisites

Opening your lesson

A HorizontalPodAutoscaler changes a workload's desired replica count from observed metrics. It doesn't create nodes or repair an application that can't start. CPU utilization targets depend on suitable CPU requests and a metrics source; a percentage without that context doesn't describe usable capacity. Set the workload reference and replica bounds deliberately, then inspect the metric target. This local workspace validates the configuration rather than running a metrics server or generating changing production demand. These tasks retain a minimum service count and a bounded maximum, while the surrounding notes explain what an operator would still need to verify on a real cluster.

printf '%s' 'apiVersion: "autoscaling/v2"
kind: "HorizontalPodAutoscaler"
metadata:
  name: "cobalt-320"
spec:
  scaleTargetRef:
    apiVersion: "apps/v1"
    kind: "Deployment"
    name: "cobalt-320"
  minReplicas: 2
  maxReplicas: 5
  metrics:
    - type: "Resource"
      resource:
        name: "cpu"
        target:
          type: "Utilization"
          averageUtilization: 60
' > 'workload.yaml'
kubectl apply -f workload.yaml

An HPA targets a workload and a metric. It doesn't add nodes, fix missing resource requests or manufacture metrics.

Choose your character

Your Work Gym account