A Brev Launchable that stands up a complete, self-paced Kubernetes 101 course on a single CPU instance — an interactive website with a live in-browser shell wired to a real Kubernetes cluster, animated diagrams, guided labs, and auto-graded exercises + CKA/CKAD-style debug challenges.
Brev VM-mode launchable (Ubuntu 22.04)
└─ Docker ── kind (Kubernetes-in-Docker), latest upstream k8s, multi-node
├─ kubectl / helm on the host
├─ metrics-server + ingress-nginx + local-path storage (default StorageClass)
└─ Workshop website (Next.js) on :3000
├─ MDX lessons + animated SVG/framer-motion diagrams
├─ live lab shell (xterm.js ↔ node-pty) — real `kubectl` against the cluster
└─ /api/check grader — labs & debug challenges verified by exit code
No GPU, no nested virtualization, no cloud account beyond the Brev instance. Everything the learner types runs against a genuine multi-node Kubernetes cluster living in Docker on the same VM.
A modern, comprehensive 101 — from "what is a container" to CNI internals:
| Part | Topics |
|---|---|
| I · Orientation | The lab environment, the live shell, how self-paced grading works |
| II · Foundations | Why orchestration, cluster architecture (control plane + nodes), kubectl, your first Pod |
| III · Workloads | Pods in depth (init/sidecar, probes, lifecycle), Deployments/ReplicaSets (rollout & rollback), DaemonSets, StatefulSets, Jobs, CronJobs |
| IV · Scheduling & Config | Labels/selectors, node affinity, taints & tolerations, requests/limits & QoS, ConfigMaps & Secrets |
| V · Storage | Volumes, PV / PVC / StorageClass, access modes, reclaim policies, CSI drivers, dynamic provisioning, volume snapshots, ephemeral volumes |
| VI · Networking | Services (ClusterIP/NodePort/LoadBalancer), DNS & endpoints, the CNI pod-network model, Multus & NetworkAttachmentDefinition, Ingress, NetworkPolicy |
| VII · Debug like CKA/CKAD | The debugging workflow (logs, exec, describe, events, port-forward) and auto-graded broken-cluster challenges |
| VIII · Reference | kubectl cheatsheet, exam tips, links |
| Resource | Minimum | Recommended |
|---|---|---|
| vCPU | 4 | 8 |
| RAM | 8 GB | 16 GB |
| Disk | 40 GB | 80 GB+ |
| GPU | none | none |
| Runtime mode | VM mode (Ubuntu 22.04) | VM mode |
| Nested virt | not needed | not needed |
| Port | Purpose |
|---|---|
| 3000 | Workshop website — start here; lessons + live shell + graded labs |
| 80 | Ingress-nginx HTTP (Ingress + Gateway labs) |
| 30080 | NodePort demo (Services lesson) |
| 6443 | kind API server (optional, for kubectl from outside) |
As a Brev Launchable setup script:
cd k8s-101-launchable
./scripts/setup.shLocally (macOS/Linux with Docker) to preview the site while iterating on content — the shell labs need the cluster, but the site + diagrams render without one:
cd web
npm install
npm run dev # http://localhost:3000Run provisioning phases individually for debugging — see scripts/.
kind delete cluster --name k8s101
sudo systemctl disable --now k8s101-workshop.servicescripts/ phased provisioning (Brev setup.sh calls these in order)
kind/ kind cluster config (multi-node, latest k8s)
manifests/ example YAML used by the labs (copied into the lab workspace)
web/ the Next.js workshop site + live-shell server
src/content/ one .mdx per lesson
src/lib/curriculum.ts the single source of truth for modules/chapters
src/components/ Terminal, LabSplit, LabCheck (grader), animated diagrams
The hero above is a static banner (docs/hero.svg). To capture a real demo GIF of the
running site (lesson → Run-in-shell → graded ✓), start the site, then screen-record the browser:
# 1. bring the site up (locally with Docker, or on the Brev VM)
cd web && npm install && npm run dev # http://localhost:3000
# 2. record the window to a GIF (macOS example — needs ffmpeg + gifski, or use QuickTime → gif)
# Open http://localhost:3000/learn/first-pod, hit "Run in shell", press "Check my work",
# and record ~15s. Then:
ffmpeg -i screen.mov -vf "fps=12,scale=1000:-1:flags=lanczos" -f gif docs/demo.gifDrop the result at docs/demo.gif and reference it here. (A GIF can't be generated headlessly —
it needs a screen recording of the live UI, which is why this is a recipe rather than a build step.)