GitOps fundamentals labs

These labs are read-only. Write answers in your lab journal before checking the GitOps fundamentals solution.

Lab 1.1 — Map push and pull delivery

Goal: reason about control flow and credentials.

Draw two diagrams for the same change, replicaCount: 1 to replicaCount: 3:

1. GitLab CI runs helm upgrade against GKE.
2. Flux running in GKE observes a Git commit.

For each arrow, annotate:

• who initiates the network connection;
• where Kubernetes credentials exist;
• where failure is reported;
• what notices a later manual edit.

Validation: your Flux diagram must include Git, source-controller, kustomize-controller, helm-controller, Artifact Registry, and the Kubernetes API.

Lab 1.2 — Inspect the cluster as observed state

kubectl config current-context
kubectl get namespaces
kubectl get deployments,replicasets,pods,services -A
kubectl get events -A --sort-by='.lastTimestamp' | tail -n 20

Answer:

1. Which resources declare intent and which represent runtime realization?
2. If a Pod is deleted, which controller recreates it?
3. Is that behavior GitOps? Explain why or why not.

Lab 1.3 — Flux preflight without installation

flux check --pre
kubectl api-resources | grep -E 'deployment|service|customresourcedefinition'
kubectl auth can-i create customresourcedefinitions.apiextensions.k8s.io
kubectl auth can-i create clusterroles.rbac.authorization.k8s.io

Record each preflight result. Do not bootstrap Flux in this topic.

Lab 1.4 — Failure ownership game

Name the first controller and command you would inspect for each scenario:

Scenario	Controller	First evidence
Git cannot be cloned	?	?
OCI chart tag is missing	?	?
Helm install hook fails	?	?
Deployment creates no Ready Pods	?	?
Alert cannot reach its receiver	?	?

Completion checklist

• I drew both delivery models from memory.
• I can separate Kubernetes reconciliation from GitOps reconciliation.
• flux check --pre has no unexplained failure.
• I assigned every failure to the correct owner.