Scaling

Horizontal Pod Autoscaler (HPA)

Scale based on CPU/memory or custom metrics.

apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata:
  name: api-server-hpa
  namespace: production
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: api-server
  minReplicas: 2
  maxReplicas: 20
  metrics:
    - type: Resource
      resource:
        name: cpu
        target:
          type: Utilization
          averageUtilization: 70
    - type: Resource
      resource:
        name: memory
        target:
          type: AverageValue
          averageValue: 200Mi
  behavior:
    scaleDown:
      stabilizationWindowSeconds: 300   # wait 5m before scaling down
      policies:
        - type: Percent
          value: 20
          periodSeconds: 60
    scaleUp:
      stabilizationWindowSeconds: 0
      policies:
        - type: Percent
          value: 100
          periodSeconds: 15

KEDA (event-driven autoscaling)

Scale on external signals: queue depth, HTTP request rate, Prometheus metrics.

apiVersion: keda.sh/v1alpha1
kind: ScaledObject
metadata:
  name: worker-scaler
  namespace: production
spec:
  scaleTargetRef:
    name: worker-deployment
  minReplicaCount: 0      # scale to zero when idle
  maxReplicaCount: 50
  cooldownPeriod: 120
  triggers:
    - type: rabbitmq
      metadata:
        queueName: tasks
        queueLength: "10"   # one replica per 10 messages
      authenticationRef:
        name: rabbitmq-trigger-auth

Cluster Autoscaler (AWS)

# Annotate the node group
eksctl scale nodegroup \
  --cluster my-cluster \
  --name workers \
  --nodes-min 2 \
  --nodes-max 20

Key Cluster Autoscaler flags:

Flag	Value	Notes
--scale-down-delay-after-add	5m	Wait after scale-up before evaluating scale-down
--scale-down-unneeded-time	10m	How long a node must be unneeded before removal
--skip-nodes-with-system-pods	true	Protect nodes with kube-system pods
--balance-similar-node-groups	true	Keep AZ balance

Vertical Pod Autoscaler (VPA)

Use in Off mode to get recommendations without auto-applying in production:

apiVersion: autoscaling.k8s.io/v1
kind: VerticalPodAutoscaler
metadata:
  name: api-server-vpa
spec:
  targetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: api-server
  updatePolicy:
    updateMode: Off    # recommendations only; set to Auto with caution

kubectl get vpa api-server-vpa -o jsonpath='{.status.recommendation}' | jq .