We started by understanding each application as it ran in the wild, what it depended on, what it configured at startup, and how it behaved under load. That meant mapping external services (databases, caches, storage, search, and shared filesystems), collecting environment configuration from EC2, and translating the existing Ansible setup into container-friendly conventions.

Each application received a Docker file and a lightweight startup script that mirrored the original Ansible tasks. Within Kubernetes, PHP ran together with Nginx in the same pod to preserve the serving model teams were accustomed to, while gaining the consistency of container images across staging and production. We introduced a standard Helm chart to keep application configuration clear and repeatable and to capture things like environment variables, resource boundaries, and persistent mounts.

Releases moved to a continuous model: GitHub Actions now builds and pushes images; Argo CD deploys to EKS in a controlled, auditable way. Staging environments gave developers a straightforward place to validate end-to-end behaviour before promoting changes. Finally, DNS updates through Route 53 completed the cutover with minimal disruption.