The deployable approach uses GitLab CI/CD to build Docker images, push to ECR, and deploy to Kubernetes with OIDC-based AWS authentication to avoid long-lived keys.

The project demonstrates how Terraform, GitLab, Kubernetes, AWS services, and monitoring tools work together, emphasizing the importance of verifying recovery paths and understanding trade-offs at each layer.

High availability is built across multiple AZs with an EKS cluster, an Application Load Balancer for traffic, Kubernetes pod replacement, a Horizontal Pod Autoscaler scaling from 2 to 6 pods, and RDS configured Multi-AZ for resilience.

Environment separation ensures development and production each have their own infrastructure to prevent cross-environment issues.

Observability relies on Prometheus for metrics and Grafana for visualization, with alerts configured to detect issues without constant dashboard monitoring.

The author transitions a local Dockerized app (LibraryCorner with frontend, FastAPI backend, and PostgreSQL) into a production-ready service on AWS using EKS and RDS, emphasizing deployability, high availability, observability, and recoverability.

Recoverability is demonstrated by a real restoration test of the RDS database, achieving an approximate 14-minute recovery time and underscoring practical disaster recovery planning.

Reliability comes from the ability to verify behavior under failure, not merely assuming it will work.

Traffic flows from Route 53 to an Application Load Balancer, into EKS, and finally to RDS PostgreSQL, with EKS nodes and database in private subnets and the ALB serving as the public entry point.

Infrastructure as Code and tooling hinge on Terraform for AWS, GitLab CI/CD for automation, Docker images in ECR, Secrets Manager for sensitive data, Prometheus and Grafana for monitoring, CloudWatch for AWS metrics, and S3 for Terraform state.