EKS Cost Optimization: Scaling, Sizing, and Savings

Code samples and manifests for the AWS Prescriptive Guidance guide on Amazon EKS cost optimization.

Overview

This repository contains Kubernetes YAML manifests, shell scripts, and configuration samples referenced in the APG guide. Each folder corresponds to a section of the guide:

Folder	Section	Contents
01-idle-cluster-cleanup/	Idle Cluster Decommissioning	Scripts to detect idle clusters via kubectl and CloudWatch
02-application-cleanup/	Application Cleanup	Scripts to find stale namespaces and remove redundant workloads
03-eks-control-plane-upgrade/	EKS Control Plane Upgrade	Version audit, cost estimation, pre-upgrade health checks
04-karpenter-cost-optimization/	Karpenter Cost Optimization	(see 07 and 09 for Karpenter NodePool manifests)
05-cluster-autoscaler-optimization/	Cluster Autoscaler Tuning	Scripts to verify and monitor autoscaler behavior
06-cpu-memory-right-sizing/	CPU and Memory Right-Sizing	VPA and HPA manifests for recommendation and scaling
07-graviton-instance-migration/	Graviton Instance Migration	Karpenter NodePool for ARM64, readiness detection
08-vendor-custom-metrics-reduction/	Metrics Cost Reduction	ServiceMonitor with relabeling, Container Insights audit
09-spark-jobs-optimization/	Spark Jobs Optimization	Spark NodePool, cost-optimized Spark config
10-additional-strategies/	Additional Cost Savings	gp3 StorageClass, namespace quotas, off-hours scaling, topology routing

Prerequisites

• Amazon EKS clusters running Kubernetes 1.23+
• kubectl configured for your EKS cluster
• AWS CLI v2.x configured with appropriate permissions
• Karpenter v0.33+ (for NodePool manifests)
• Helm 3.x (for VPA and monitoring stack installation)

Usage

Shell Scripts

All .sh scripts accept cluster name as the first argument:

# Detect idle clusters
./01-idle-cluster-cleanup/detect-idle-clusters.sh

# CloudWatch-based detection
./01-idle-cluster-cleanup/detect-idle-clusters-cloudwatch.sh my-cluster 7

# Audit cluster versions
./03-eks-control-plane-upgrade/audit-cluster-versions.sh "us-east-1 us-west-2"

YAML Manifests

Apply manifests with kubectl:

# Deploy VPA in recommendation mode
kubectl apply -f 06-cpu-memory-right-sizing/vpa-recommendation-mode.yaml

# Deploy cost-efficient HPA
kubectl apply -f 06-cpu-memory-right-sizing/hpa-efficient-scaling.yaml

# Create Graviton-preferred NodePool
kubectl apply -f 07-graviton-instance-migration/graviton-nodepool.yaml

# Create dedicated Spark NodePool
kubectl apply -f 09-spark-jobs-optimization/spark-nodepool.yaml

Security

See CONTRIBUTING for reporting security issues.

License

This library is licensed under the MIT-0 License. See the LICENSE file.