Compute — Kubernetes
Kubernetes: orchestrate many containerized services — scheduling, rollout, service discovery, when ops capacity and service count make it earned.
- Multiple services, independent scale, real platform team (or strong desire + time to learn)
- You already have Docker discipline per unit
Skip when
Section titled “Skip when”- Single app / single site — this is usually Path 01 smoke. Run
kiss. - No one owns the cluster — prefer Clouds managed runtimes or Docker on Fly/Cloud Run/ECS
Starter DAG
Section titled “Starter DAG”flowchart TD A[kiss: is k8s earned?] --> B[Deploy units + resource boundaries] B --> C[Images from Docker path] C --> D[Manifests/Helm/Kustomize - least mechanism] D --> E[CI: affected charts + kubeconform/policy] E --> F[ESC for cluster credentials - short-lived] F --> G[stage then prod namespaces] G --> H[observe-it: SLOs, rollouts, events]Skills
Section titled “Skills”kiss · repos architecture · repos ci · repos secrets · stage-it · ship-it · observe-it · agents slap if GitOps agents thrash · tidy-up
Watchouts
Section titled “Watchouts”- Cluster sprawl without CODEOWNERS
- Long-lived kubeconfigs in GitHub Secrets
- Deploying the world on every package bump