EKS Retail Platform at Scale

(01)Overview

A highly available, multi-AZ AWS EKS platform provisioned with Terraform, running five production microservices — UI, Cart, Catalog, Orders, and Checkout — written in Java Spring Boot, Node.js, and Go. Every layer is automated: GitOps delivery from commit to cluster, multi-layer autoscaling on Spot capacity, a full OpenTelemetry observability pipeline, and a polyglot AWS data layer for event-driven communication.

(02)What we built

1. 1. Multi-AZ EKS cluster provisioned with Terraform — remote state, custom VPC with public/private subnets, least-privilege IAM, Route53 DNS, and ACM-managed TLS.
2. 2. 5 microservices (Java Spring Boot, Node.js, Go) deployed with standardized, environment-specific Helm charts.
3. 3. GitOps CI/CD with GitHub Actions and Argo CD: commit → Docker build → ECR push → Helm values update → EKS deploy, with OIDC authentication and zero hardcoded AWS credentials.
4. 4. Argo CD auto-sync, self-healing, drift detection/correction, and automated rollback across environments.
5. 5. Observability with OpenTelemetry (ADOT): distributed tracing to AWS X-Ray, logs to CloudWatch, metrics to Amazon Managed Prometheus and Grafana.
6. 6. Health-check trace filtering across the observability stack to cut distributed-tracing cost by 85%.
7. 7. Karpenter node autoscaling on Spot instance pools with automated consolidation to cut compute cost by 70%.
8. 8. Multi-layer high availability: HPA (pods), Karpenter (nodes), Pod Disruption Budgets, and topology spread constraints.
9. 9. EKS Pod Identity, fine-grained RBAC, Secrets Store CSI Driver for secret injection, and EBS CSI for persistent volumes.
10. 10. Polyglot data layer — DynamoDB, RDS (MySQL/PostgreSQL), ElastiCache Redis, and SQS for asynchronous, event-driven communication.