Run Kubernetes Clusters for Less with Amazon EC2 Spot and Karpenter

Running Kubernetes workloads on Amazon EKS (Elastic Kubernetes Service) is powerful but can become costly, especially for high-scale environments. Fortunately, AWS offers EC2 Spot Instances to cut compute costs by up to 90%, and Karpenter, an open-source autoscaler, to optimize instance provisioning in real-time.

In this guide, I will walk you through:

• What Spot Instances and Karpenter are

• Why they are cost-effective together

• A step-by-step guide to deploy Karpenter with EKS

• Best practices

What Are EC2 Spot Instances?

Spot Instances let you use AWS’s unused EC2 capacity at reduced prices. However, they can be interrupted with just a 2-minute warning if AWS reclaims capacity.

Pros:

• Up to 90% cheaper than On-Demand

• Access to high-performance instances

Cons:

• Not suitable for stateful, long-running jobs unless fault-tolerant

• Can be terminated any time

---

🚀 What is Karpenter?

Karpenter is an open-source autoscaler for Kubernetes, purpose-built for AWS. Unlike the default Cluster Autoscaler:

• It launches instances dynamically based on pod needs.

• It selects the best instance types and capacity pools.

• Supports Spot, On-Demand, and Graviton instances.

• Supports zones, architectures, taints, labels, affinities, etc.

---

📊 Why Use EC2 Spot + Karpenter?

Combining both gives you:

• Cost-efficient workloads (Spot)

• Smart and dynamic scaling (Karpenter)

• Minimized waste through pod-to-instance right-sizing

---

🛠️ Prerequisites

---

🧰 Step-by-Step Guide

---

Step 1: Create an EKS Cluster

eksctl create cluster \
  --name karpenter-demo \
  --region us-east-1 \
  --version 1.29 \
  --nodegroup-name default-ng \
  --nodes 2 \
  --nodes-min 1 \
  --nodes-max 5 \
  --node-type t3.medium \
  --managed

---

Step 2: Install Karpenter Controller

2.1 Add Helm Repo

helm repo add karpenter https://charts.karpenter.sh
helm repo update

---

Step 3: Create Karpenter IAM Role & Instance Profile

3.1 Download AWS CloudFormation Template

curl -fsSL https://karpenter.sh/docs/getting-started/cloudformation.yaml -o karpenter-iam.yaml
aws cloudformation deploy \
  --stack-name "KarpenterIRSA" \
  --template-file karpenter-iam.yaml \
  --capabilities CAPABILITY_NAMED_IAM \
  --parameter-overrides \
    ClusterName=karpenter-demo \
    ServiceAccountNamespace=karpenter \
    ServiceAccountName=karpenter \
    ClusterEndpoint=$(aws eks describe-cluster --name karpenter-demo --region us-east-1 --query "cluster.endpoint" --output text) \
    OIDCProvider=$(aws eks describe-cluster --name karpenter-demo --region us-east-1 --query "cluster.identity.oidc.issuer" --output text | sed -e "s/^https:\/\///")

---

Step 4: Install Karpenter via Helm

helm upgrade --install karpenter karpenter/karpenter \
  --namespace karpenter --create-namespace \
  --set serviceAccount.create=false \
  --set serviceAccount.name=karpenter \
  --set settings.clusterName=karpenter-demo \
  --set settings.clusterEndpoint=$(aws eks describe-cluster --name karpenter-demo --region us-east-1 --query "cluster.endpoint" --output text) \
  --set settings.aws.defaultInstanceProfile=KarpenterNodeInstanceProfile-karpenter-demo

---

Step 5: Create a Provisioner

# karpenter-provisioner.yaml
apiVersion: karpenter.sh/v1alpha5
kind: Provisioner
metadata:
  name: default
spec:
  requirements:
    - key: "karpenter.sh/capacity-type"
      operator: In
      values: ["spot"]   # Only use spot instances
    - key: "node.kubernetes.io/instance-type"
      operator: In
      values: ["m5.large", "m5.xlarge", "t3.medium"]
  limits:
    resources:
      cpu: 1000
  provider:
    subnetSelector:
      kubernetes.io/cluster/karpenter-demo: owned
    securityGroupSelector:
      kubernetes.io/cluster/karpenter-demo: owned
  ttlSecondsAfterEmpty: 30

Apply it:

kubectl apply -f karpenter-provisioner.yaml

---

Step 6: Deploy Sample Workload

# sample-deployment.yaml
apiVersion: apps/v1
kind: Deployment
metadata:
  name: inflate
spec:
  replicas: 5
  selector:
    matchLabels:
      app: inflate
  template:
    metadata:
      labels:
        app: inflate
    spec:
      containers:
        - name: inflate
          image: public.ecr.aws/eks-distro/kubernetes/pause:3.5
          resources:
            requests:
              cpu: "1"

Deploy it:

kubectl apply -f sample-deployment.yaml

Karpenter will observe that the default nodes don’t have capacity and will launch Spot Instances that fit.

---

📉 Cost Comparison Illustration

Provisioning 100 pods needing 1 vCPU each on Spot vs On-Demand could reduce your bill from $230/month to ~$70/month.

---

📊 Monitoring Karpenter

You can view logs and activity:

kubectl logs -n karpenter -l app.kubernetes.io/name=karpenter

To check provisioned nodes:

kubectl get nodes -l karpenter.sh/provisioner-name=default

---

🛡️ Spot Best Practices with Karpenter

1. Diversify instance types in the Provisioner.

2. Set taints and affinity for critical workloads to run on On-Demand.

3. Use Node Termination Handler (NTH) to drain pods gracefully.

4. Enable Pod Disruption Budgets (PDBs) to protect critical workloads.

5. Track Spot Interruption metrics with Prometheus.

---

📌 Final Thoughts

Using Karpenter with EC2 Spot Instances offers a powerful way to cut infrastructure costs while still delivering resilient, scalable Kubernetes workloads.