Labs

One playbook per phase. Every YAML file is complete and copy-pasteable, every step shows the expected output, and every lab ends with a verify block. Where the syllabus asks you to write something by hand, the solution is in a collapsible block - try first, then peek.

Ground rules

1. Type the kubectl commands yourself. Copy-paste YAML, but type commands - that's where muscle memory comes from.
2. Read every error. When output doesn't match the expected output shown, stop and figure out why before continuing. That debugging is the actual learning.
3. Do the labs in order. Lab 4 assumes the cluster from Lab 1, Lab 6 reuses manifests from Lab 2.

The demo stack

All labs use the same small set of public, multi-arch images (they work on Apple Silicon):

Image	Role in the labs
traefik/whoami	The main app. Tiny HTTP server that echoes its hostname - perfect for seeing load balancing and rolling updates. Tags v1.10 and v1.11 give us real version upgrades.
postgres:16-alpine	The stateful workload for config, secrets, and persistent volumes.
adminer	A web UI for Postgres - our "app that depends on a database" for probe exercises.
nginx	Static frontend for the ingress lab.
curlimages/curl	In-cluster test client.
busybox:1.36	CronJob payloads.
polinux/stress	Deliberately eats memory so we can trigger OOMKilled on purpose.

Prerequisites

brew install kind kubectl k9s helm
docker --version   # any recent version, must be running

Workspace

Keep lab files in a directory per lab:

mkdir -p ~/project/k8dev-labs/{lab-1,lab-2,lab-3,lab-4,lab-5,lab-6}
cd ~/project/k8dev-labs

The labs

Lab	Pairs with	What you build
Lab 1	Phase 1	An ingress-ready kind cluster + kubectl drills
Lab 2	Phase 2	whoami: Deployment, Service, rolling update, rollback
Lab 3	Phase 3	Postgres + Adminer with ConfigMap, Secret, PVC, probes
Lab 4	Phase 4	Two services behind one Ingress on a real hostname
Lab 5	Phase 5	Break-and-fix drills, CronJob, autoscaling
Lab 6	Phase 6	Helm install of podinfo + Kustomize dev/prod overlays