# Scale orders on in-flight requests per pod, not CPU. Target: 5 in flight on average.
apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata: {name: orders, namespace: demo}
spec:
  scaleTargetRef: {apiVersion: apps/v1, kind: Deployment, name: orders}
  minReplicas: 1
  maxReplicas: 4
  metrics:
    - type: Pods
      pods:
        metric: {name: app_inflight_requests}
        target: {type: AverageValue, averageValue: "5"}
  behavior:
    scaleUp:
      stabilizationWindowSeconds: 0
    scaleDown:
      # The default is 300 s. Shortened so the demo shows a scale-down inside a few minutes;
      # keep the default (or longer) in production.
      stabilizationWindowSeconds: 30