Horizontal Pod Autoscaler

It increases/ decreases no. of replicas in deployment ( control’s no. of replicas ) based on cpu utilisation, RAM..

Horizontal Scaling : Adding more pods to cluster

Vertical Scaling : Increase size of machines ( 4 core → 8 core ). Pods will have access to more resources.

Architecture

An infinite control loop constantly checks cpu utilisation in pods.

• cAdvisor : uses by K8s to find out how much resource is used up.
• metrics : node cpu utilisation & pod cpu utilisation are metrics that are tracked by K8s Cluster
• metricsServer : Its a server that runs on every node. That lets u calculate metrics.

Tracking metrics in K8s Cluster

View resource usage with the Kubernetes Metrics Server - Amazon EKS

Step 1 : Create a K8s Cluster

eksctl create cluster --name hpa-cluster --region ap-south-1 --nodes 3 --node-type t2.medium

Step 2 : Add metricsServer to your K8s Cluster

kubectl apply -f <https://github.com/kubernetes-sigs/metrics-server/releases/latest/download/components.yaml>

Step 3 : Clean re-install

Your cluster already had a metrics-server Deployment installed earlier (maybe via EKS addons or Helm), and kubectl apply is trying to patch it, which is failing because:

1. Duplicate port name https → The Deployment manifest you applied has two container ports both named https.
2. Selector is immutable → Kubernetes doesn’t allow you to change .spec.selector after a Deployment is created. Since the one you applied differs from the existing one, it errors out.