Karpenter Autoscaling

When to Use

• Replacing Cluster Autoscaler with faster, smarter autoscaling
• Optimizing EKS cluster costs (target: 20%+ savings)
• Implementing Spot instance strategies (30-70% Spot mix)
• Need sub-minute node provisioning (seconds vs minutes)
• Workloads with variable resource requirements
• Multi-instance-type flexibility without node group management
• GPU or specialized instance provisioning
• Consolidating underutilized nodes automatically

Prerequisites

• EKS cluster running Kubernetes 1.23+
• Terraform or Helm for installation
• IRSA or EKS Pod Identity enabled
• Small node group for Karpenter controller (2-3 nodes)
• VPC subnets and security groups tagged for Karpenter discovery

Quick Start

1. Install Karpenter (Helm)

# Add Karpenter Helm repo
helm repo add karpenter https://charts.karpenter.sh
helm repo update

# Install Karpenter v1.0+
helm upgrade --install karpenter karpenter/karpenter \
  --namespace kube-system \
  --set settings.clusterName=my-cluster \
  --set settings.interruptionQueue=my-cluster \
  --set controller.resources.requests.cpu=1 \
  --set controller.resources.requests.memory=1Gi \
  --set controller.resources.limits.cpu=1 \
  --set controller.resources.limits.memory=1Gi \
  --wait

See: references/installation.md for complete setup including IRSA/Pod Identity

2. Create NodePool and EC2NodeClass

NodePool (defines scheduling requirements and limits):

apiVersion: karpenter.sh/v1