The closest one-to-one match to a real production setup: minimal but correct.
Goal
Enable any engineer to run their app the same way it will run in a production cluster, using GitOps, containerd workflows, and local Kubernetes.
Install these tools (laptop)
| Tool | Purpose |
|---|---|
| Rancher Desktop (containerd mode) | Local Kubernetes runtime (k3s) |
| kubectl | Manage the cluster |
| Helm | Deployment packaging / templating |
| Git | Version control |
| Visual Studio Code (or JetBrains) | Editor, YAML plugins recommended |
Cluster config
- Enable containerd
- Enable Traefik ingress
- Create a local-path storage class if it is not the default
Cluster services to deploy
| Component | Why |
|---|---|
| cert-manager | TLS automation |
| Traefik | Ingress + local DNS routing |
| Gitea | Local Git for the GitOps flow |
| Harbor (no scanner) | Local image registry |
| Flux | Sync the cluster from Git |
| Kaniko Job template | In-cluster builds (dev parity) |
Dev workflow (hands-on)
- Clone a service repo locally.
- Write a Deployment, Service, and Ingress YAML.
- Push the code to Gitea.
- Flux syncs the configs and the app appears in the cluster.
- Trigger Kaniko: the image builds, is pushed, and the deployment updates.
- Test the app through the Traefik ingress URL.
- Debug with
kubectl logsandkubectl describe.
You are now working in the same execution model as a real production environment.
What this teaches
| Skill | Why it matters |
|---|---|
| Kubernetes basics | Most modern services run here |
| GitOps discipline | No manual drift |
| OCI-native builds | No Docker daemon assumptions |
| TLS + ingress | Every production service uses it |
| Registry workflows | Promotion + traceability |
| YAML deployments | You must speak the platform |
Completion criteria
You can:
- Deploy your app through GitOps.
- Route traffic through Ingress + TLS.
- Build images inside Kubernetes and push them to Harbor.
- Understand namespaces, PVCs, events, and rollout status.
- Debug failed deployments without searching for an hour.
At that point you are DevOps compatible and can work confidently on a production platform.
Final philosophy
Build on your laptop, run in the cluster, no special casing. If it works here, it will work in production.
