A pragmatic guide to when and how to adopt EKS for large-scale teams, written by a self-described serverless advocate. The post outlines a compute decision ladder: Lambda first, then Fargate, then EKS only when scale demands it. The recommended EKS blueprint combines Flux for GitOps reconciliation, Karpenter for dynamic node provisioning, Helm and Kustomize for workload templating, External Secrets Operator for credential management, and a thin Fargate layer to bootstrap system controllers. Key topics include solving the Karpenter bootstrap circular dependency with Fargate profiles, using Spot instances for 60-80% compute savings, and honest caveats about version upgrade cadence, blast radius of GitOps, and the real need for Kubernetes-literate engineers.
