DevOps Learning Labs¶
Welcome to DevOps Learning Labs — a comprehensive, structured learning journey through the entire DevOps and Kubernetes ecosystem.
DevOps Games¶
K8s Games¶
🎯 What This Project Is¶
A practical, open-source, teachable learning environment that takes you from Docker basics all the way through multi-region Kubernetes deployments with GitOps automation, observability, and advanced patterns.
📋 Learning Path (Standard Track: ~1-2 weeks)¶
Phase 1: Containerization (Days 1-2)¶
- Theory: DevOps fundamentals, Docker concepts
- Labs: 00-01 (Docker basics, image building, pushing to registry)
Phase 2: Kubernetes Fundamentals (Days 3-4)¶
- Theory: K8s architecture, workloads, networking
- Labs: 02-05 (Pods, Deployments, Services, ConfigMaps, Secrets)
Phase 3: Helm & Packaging (Days 5-6)¶
- Theory: Helm charts, templating, dependencies
- Labs: 06 (Create and deploy Helm charts)
Phase 4: Multi-Region & Sidecars (Days 7-8)¶
- Theory: Advanced patterns, multi-cluster architectures
- Labs: 07-08 (2 Minikube clusters, inter-cluster networking, sidecars)
Phase 5: GitOps & Observability (Days 9-10)¶
- Theory: GitOps principles, Flux, observability
- Labs: 09-10 (Flux deployment, Prometheus, Grafana, Loki)
Phase 6: Troubleshooting & Integration (Days 11+)¶
- Theory: Advanced debugging, security, performance
- Lab: 11 (Hands-on troubleshooting scenarios)
🛠️ Technology Stack (All Open-Source)¶
| Component | Tool | Why |
|---|---|---|
| Local Kubernetes | Minikube + kind | Lightweight, multi-cluster capable |
| Container Runtime | Docker CE | Industry standard |
| K8s Package Manager | Helm | De facto standard for K8s deployments |
| GitOps Automation | Flux CD | CNCF project, enterprise-ready |
| Metrics | Prometheus | Time-series DB, Kubernetes-native |
| Dashboards | Grafana | Open-source, powerful visualization |
| Logging | Loki | Lightweight log aggregation |
| Tracing | Jaeger (optional) | Distributed tracing |
| Container Registry | Docker Hub / ghcr.io | Free image hosting |
| K8s UI (Built-in) | Minikube Dashboard | Visual pod/service management |
| K8s Terminal UI | k9s (optional) | Real-time cluster monitoring |
| K8s IDE | Lens (optional) | Enterprise visual IDE for K8s |
🎓Interview & Self-Assessment¶
- 📄 Interview Prep: Q&A for self-study and validation
🔗 Resources & References¶
- Kubernetes Official Docs: https://kubernetes.io/docs/
- Helm Documentation: https://helm.sh/docs/
- Flux CD Documentation: https://fluxcd.io/flux/
- Prometheus: https://prometheus.io/
- Grafana: https://grafana.com/
🎯 Quick Links¶
First Time? → Start with Setup Guide
Want Theory? → Read Theory Modules
Ready to Code? → Jump to Lab 00
Preparing for Interview? → Check Interview Prep
updated plan¶
Blog Angles by Week
| Week | Item | Angle |
|---|---|---|
| 1 | Real 3-node Standard GKE cluster (not Autopilot) | "What Autopilot hides from you" — node pool sizing, machine types, real kubectl describe node output |
| 1 | GKE auth via Workload Identity vs static kubeconfig | Practical writeup of the auth model most tutorials skip |
| 2 | Live HPA event under load (hey/k6) | Capture before/during/after pod counts + latency — hard to make interesting on a toy cluster, easy on real infra |
| 2 | Helm chart templating decisions | "Every values.yaml choice, defended" — good technical-depth post |
| 3 | Bad image tag pushed through GitLab CI pipeline | "Here's what actually breaks" — rollout failure + automatic rollback, the stuff generic tutorials skip |
| 3 | GitLab CI → GKE deploy pipeline | Reconciliation model explained in your own words (Flux vs GitLab, same underlying idea) |
| 4 | Cost observability — GCP billing export → Grafana panel | "I hooked up cost monitoring to my own dashboard" — not on the original plan, differentiated content |
| 4 | Live diagnosis of a deliberately broken deployment | Real troubleshooting narrative, not a staged tutorial |
Here's a project that threads every concept from your plan into one continuous build — not isolated labs, but a single app you keep extending, so each new concept has to interact with everything before it (the way it actually would in a real role).
Project: "Orders API" — a small service you ship through the full real-world path¶
The shape of it: a tiny HTTP API (2-3 endpoints — could literally be /orders, /orders/:id, /health) that you deploy, expose, scale, break, fix, and monitor — on your real GKE cluster, through your real GitLab pipeline.
| Stage | What you add | Concepts it forces you to actually use |
|---|---|---|
| 1 | Containerize it, deploy raw manifests to GKE | Docker, Pod, Deployment, ClusterIP Service |
| 2 | Expose it properly | NodePort → LoadBalancer → Ingress w/ a real hostname, TLS |
| 3 | Externalize config | ConfigMap + Secret (e.g. a fake DB connection string) |
| 4 | Package it | Helm chart — templatize everything above |
| 5 | Automate it | GitLab CI/CD pipeline: build → push to Artifact Registry → Helm deploy to GKE on every push |
| 6 | Make it scale | Add a CPU-heavy endpoint, load-test it, watch HPA react in real time |
| 7 | Make it observable | Prometheus scraping custom metrics, Grafana dashboard, alert rule |
| 8 | Make it break (on purpose) | Bad image tag through the pipeline → failed rollout → automatic rollback |
| 9 | Make it accountable | GCP billing export → cost panel in the same Grafana instance |
| 10 | Prove you own it | Kill a pod mid-traffic, diagnose a deliberately misconfigured probe, explain the entire request path from your own memory |