Loading auth...
Medium15 minContainer Orchestration
UpdatedAug 2, 2026
Edit

Horizontal Pod Autoscaler (HPA)

CONCEPTS:Horizontal Pod Autoscaling (HPA)Resource Management (Requests vs. Limits)

Why This Is Asked

Dynamic scaling is one of the primary reasons companies move to Kubernetes. This question tests whether a candidate knows how to configure automatic scaling, the requirements for it to function (Metrics Server and Resource Requests), and how to prevent common issues like flapping.

Key Concepts

  • Control Loop: How often the HPA checks metrics.
  • Resource Requests: Why they are mandatory for HPA.
  • Metrics Server: The source of truth for resource usage.
  • Flapping: What it is and how to prevent it.

Answers by Technology

+ Add Variant
KubernetesImprove this answer ✏️

Expected Answer

The Horizontal Pod Autoscaler (HPA) automatically adjusts the number of replicas in a Deployment based on CPU or Memory usage.

To make it work:

  1. Metrics Server: You must have the Metrics Server installed in your cluster.
  2. Resource Requests: You must define CPU/Memory requests in your Pod template. The HPA calculates percentage based on these requests (e.g., “scale up if CPU is > 80% of what I requested”).
  3. HPA Object: You create an HPA resource that points to your Deployment and sets the min/max replicas and the target metric.

Why It Matters

HPA allows your application to handle sudden spikes in traffic (like a Black Friday sale) without manual intervention, and then scale back down to save costs when traffic is low. It is a fundamental part of building cloud-native, cost-effective systems.

Example Code

An HPA targeting a Deployment

apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata:
  name: my-app-hpa
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: my-app
  minReplicas: 2
  maxReplicas: 10
  metrics:
  - type: Resource
    resource:
      name: cpu
      target:
        type: Utilization
        averageUtilization: 80

Common Mistakes

  • No Resource Requests: The HPA will never scale because it doesn’t have a baseline to compare against.
  • Flapping: Setting the target utilization too low, causing the cluster to constantly add and remove Pods. (Solution: Use behavior settings to slow down scale-down operations).
  • Metric Server Missing: The HPA will stay in a <unknown> state.
  • Scaling by Memory only: Memory usage often grows but rarely shrinks (due to GC behavior), which can lead to the HPA scaling up but never scaling back down.

Follow-up Questions

  • What is the difference between HPA and VPA? (Answer: HPA (Horizontal) adds more Pods. VPA (Vertical) makes the existing Pods bigger by increasing their CPU/Memory requests).
  • Can you scale based on custom metrics (like HTTP requests per second)? (Answer: Yes, but it requires a custom metrics adapter, like the Prometheus Adapter or KEDA).
  • What is the Cluster Autoscaler? (Answer: It scales the number of Nodes in your cluster when Pods cannot be scheduled due to lack of resources).

References