Horizontal Pod Autoscaler

The HPA scales the number of pods in a Deployment up or down based on CPU, memory, or custom metrics.

While the Cluster Autoscaler and Karpenter scale nodes, the Horizontal Pod Autoscaler (HPA) scales pods. It adjusts a Deployment's replica count to keep a metric — usually CPU utilization — near a target.

How it works

  1. The HPA reads pod metrics from the Metrics Server (install it as an add-on).
  2. If average CPU exceeds the target, it adds replicas; if it drops, it removes them.
  3. Node autoscaling then supplies capacity for the extra pods if needed.

HPA and Cluster Autoscaler/Karpenter work as a team: pods scale first, nodes follow.

Example

Example · yaml
apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata:
  name: web
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: web
  minReplicas: 2
  maxReplicas: 10
  metrics:
    - type: Resource
      resource:
        name: cpu
        target:
          type: Utilization
          averageUtilization: 60

When to use it

  • An API team deploys an HPA on their user-facing service so pod count scales from 3 to 50 automatically during a marketing campaign traffic spike.
  • A data processing team configures an HPA on a consumer Deployment targeting 70% CPU, keeping pods from being overwhelmed by message queue bursts.
  • A cost team uses HPA with memory metrics to scale a JVM service based on heap usage, right-sizing replicas without setting CPU thresholds that JVMs saturate unevenly.

More examples

Create an HPA for a Deployment

Creates an HPA that keeps CPU utilization around 60% by scaling the api Deployment between 2 and 20 replicas.

Example · bash
kubectl autoscale deployment api \
  --min=2 \
  --max=20 \
  --cpu-percent=60
kubectl get hpa api

HPA manifest with metrics

A declarative HPA manifest using autoscaling/v2 that targets average CPU utilization across all pods in the Deployment.

Example · yaml
apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata:
  name: api-hpa
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: api
  minReplicas: 2
  maxReplicas: 20
  metrics:
  - type: Resource
    resource:
      name: cpu
      target:
        type: Utilization
        averageUtilization: 60

Install Metrics Server for HPA

Installs the Metrics Server add-on that the HPA requires to collect CPU and memory usage data from kubelets.

Example · bash
kubectl apply -f https://github.com/kubernetes-sigs/metrics-server/releases/latest/download/components.yaml
kubectl top nodes
kubectl top pods

Related

eks cluster autoscalereks karpenter

Discussion

  • Be the first to comment on this lesson.