The labs in this repo are the practice. This roadmap is the curriculum: every skill area a strong DevOps / SRE / platform engineer needs, in what order, how deep, and how you know you've got it.
- Work level by level. Finish Level 1 in every domain before going deep into Level 3 of one. Breadth first, then depth.
- For each topic: learn → practice → prove.
- Learn: the resources listed, plus the lab's
knowledge.md. - Practice: the linked lab, typed by you, not copy-pasted.
- Prove: you can do the checkpoint without notes, and explain it to someone else.
- Learn: the resources listed, plus the lab's
- Write as you go. Notes in each lab's
knowledge.md("My notes") and README ("What broke", "Lessons learned"). Writing is where understanding gets tested. - Tick the checkpoints in this file as you reach them. It's your progress tracker.
| Level | Meaning | Roughly |
|---|---|---|
| 1. Foundation | You can use it and explain the basics | Junior |
| 2. Practitioner | You can build and run it on your own, and fix common problems | Mid-level |
| 3. Advanced | You can design it, run it in production, and debug the hard problems | Senior |
| 4. Expert | You know the internals and trade-offs, and you teach others | Staff / principal |
Levels 1–3 are the goal of this roadmap. Level 4 comes from years of running real systems.
| Level | Learn | Practice | Checkpoint: you can… |
|---|---|---|---|
| 1 | Shell, files and permissions, users and sudo, packages, processes, services, logs | 01, 02 | - [ ] Harden a fresh server without notes, and explain each step |
| 2 | systemd in depth, signals, cron/timers, storage (partitions, LVM, mounts), boot process | 17, 19, 03 | - [ ] Fix a server whose disk is full, a service that won't start, and a bad fstab |
| 3 | Namespaces, cgroups, kernel tuning, performance analysis (USE, perf), security modules |
18, 20, 21 | - [ ] Find why a server is slow within 15 minutes, with evidence |
Resources: The Linux Command Line (free book) · How Linux Works (Brian Ward) · Systems Performance (Brendan Gregg) · Julia Evans' zines · man pages.
| Level | Learn | Practice | Checkpoint |
|---|---|---|---|
| 1 | IP, ports, TCP vs UDP, DNS, HTTP(S), firewalls | 01, 02, 05 | - [ ] Explain what happens when you type a URL and press Enter, down to TCP and TLS |
| 2 | Subnets/CIDR, routing, NAT, load balancing, TLS and certificates, packet capture | 13, 15, 16 | - [ ] Debug "can't connect" by layer: DNS, route, firewall, port, TLS, app |
| 3 | Container and Kubernetes networking internals, VPNs, BGP basics, cloud networking (VPCs) | 14, 28, 35 | - [ ] Draw how a packet reaches a pod in Kubernetes, rule by rule |
Resources: Computer Networking: A Top-Down Approach (Kurose & Ross) · High Performance Browser Networking (free) · Beej's Guide to Network Programming · Julia Evans' networking zines · Wireshark.
| Level | Learn | Practice | Checkpoint |
|---|---|---|---|
| 1 | Bash: variables, loops, conditionals, pipes, set -euo pipefail, exit codes |
every lab's scripts (deploy.sh, smoke-test.sh) | - [ ] Write a script that passes shellcheck and fails safely |
| 2 | One real language: Go (you already have hello-api) or Python; HTTP, JSON, tests, CLIs | apps/hello-api | - [ ] Add a feature to hello-api with tests, through CI |
| 3 | Automation tools and services: API clients, a small CLI, a Kubernetes controller | your own project | - [ ] Ship a tool other people use |
Resources: A Tour of Go · Learning Go (Jon Bodner) · Automate the Boring Stuff with Python (free) · ShellCheck wiki.
| Level | Learn | Practice | Checkpoint |
|---|---|---|---|
| 1 | Commit, branch, merge, PRs, .gitignore, never committing secrets |
this repo: a branch and PR per lab | - [ ] Write commit messages and PRs others understand |
| 2 | Rebase, conflicts, revert, bisect, tags and semantic versioning, code review | — | - [ ] Find a bad commit with git bisect; recover a "lost" commit with reflog |
| 3 | Trunk-based development, release strategy, monorepos, branch protection | Phase 14 | - [ ] Design a branching and release process for a team |
Resources: Pro Git (free) · Conventional Commits · trunkbaseddevelopment.com.
| Level | Learn | Practice | Checkpoint |
|---|---|---|---|
| 1 | Images, containers, volumes, ports, Compose | 04, 06 | - [ ] Containerize an app with its database from scratch |
| 2 | Multi-stage, small and non-root images, layer caching, multi-arch, healthchecks, registries | 04, 07 | - [ ] Get an image under 30 MB that runs as non-root |
| 3 | What a container is (namespaces, cgroups, overlayfs), runtimes (containerd), image security | 18, 37, Phase 13 | - [ ] Build a container by hand with unshare |
Resources: Docker docs · Docker Deep Dive (Nigel Poulton) · What even is a container?.
| Level | Learn | Practice | Checkpoint |
|---|---|---|---|
| 1 | Pipelines: test, build, push, deploy; secrets in CI | 07 | - [ ] Build a pipeline from an empty repo to a running deploy |
| 2 | Zero-downtime deploys, rollback, immutable artifacts, GitOps | 08, 12 | - [ ] Deploy under load with zero errors, and roll back in one command |
| 3 | Multiple environments and promotion, canary and feature flags, DORA metrics | 27, 42, Phase 14 | - [ ] Measure your deployment frequency, lead time, change failure rate, recovery time |
Resources: Continuous Delivery (Humble & Farley) · Accelerate (Forsgren, Humble, Kim) · dora.dev · GitHub Actions docs.
| Level | Learn | Practice | Checkpoint |
|---|---|---|---|
| 1 | Declarative vs imperative, idempotency, Ansible playbooks | 09 | - [ ] Rebuild a server from zero with one command |
| 2 | Terraform: state, plan, providers, modules; secrets (Vault, ansible-vault) | 10 | - [ ] Create and destroy a cloud environment from code, with remote state |
| 3 | Reusable modules, testing IaC, policy as code, drift detection at team scale | 27, Phase 13 | - [ ] Write a Terraform module others reuse, with tests |
Resources: Ansible docs · Terraform: Up & Running (Yevgeniy Brikman) · Terraform tutorials.
| Level | Learn | Practice | Checkpoint |
|---|---|---|---|
| 1 | Resource hierarchy, projects, billing budgets, IAM (roles, service accounts), gcloud, Compute Engine, Cloud Storage |
26, 29 | - [ ] A budget alert exists before anything else; SSH to a VM with no public IP (IAP) |
| 2 | Project structure as code, VPC and Cloud NAT, load balancing, instance groups, Artifact Registry, Cloud Run, Cloud SQL, Secret Manager, Workload Identity Federation | 27–30 | - [ ] CI deploys to GCP with no service account key anywhere |
| 3 | GKE in production, Workload Identity, Cloud Logging/Monitoring, FinOps, multi-region design | 31, 32 | - [ ] Explain the bill line by line, and design for a given budget and availability target |
How the projects are laid out (one per environment, state, IAM, CI access): ADR 0004.
Resources: Google Cloud Skills Boost (hands-on labs) · Google Cloud documentation · Google Cloud Architecture Framework · Terraform Google provider · Associate Cloud Engineer, then Professional Cloud DevOps Engineer certification.
terraform destroy dev after each lab. Cloud mistakes cost real money. The concepts (IAM, VPCs, managed services) carry over to AWS and Azure.
| Level | Learn | Practice | Checkpoint |
|---|---|---|---|
| 1 | Pods, Deployments, Services, namespaces, labels, kubectl fluency; ConfigMaps, Secrets, volumes, StatefulSets, Jobs | 12, 33, 34 | - [ ] Deploy an app with a database from YAML you write yourself, without looking anything up |
| 2 | Networking (Services, DNS, Ingress, Gateway API, NetworkPolicies), scheduling and resources, Helm, GitOps, troubleshooting | 35, 36, 38, 23 | - [ ] Fix CrashLoopBackOff, Pending, ImagePullBackOff, and OOMKilled pods in minutes; pass CKAD |
| 3 | Security (RBAC, Pod Security, policy as code), control-plane internals, kubeadm, upgrades, etcd backup, GKE | 37, 39, 40, 31 | - [ ] Build a cluster with kubeadm, upgrade it, and restore etcd; pass CKA |
| 4 | Operators and CRDs, admission control, progressive delivery, multi-environment GitOps | 41, 42 | - [ ] Write an operator in Go; canary with automatic rollback; pass CKS |
Resources: Kubernetes docs (also allowed during CKA/CKAD/CKS) · Kubernetes Up & Running · Kubernetes in Action (Marko Lukša) · Killercoda (free browser labs) · killer.sh (exam simulator) · Kubernetes the Hard Way · The Kubebuilder Book · Programming Kubernetes (Hausenblas & Schimanski).
A mesh moves encryption, retries, traffic control, and per-call telemetry out of the app code and into proxies (or the kernel). Learn it after Kubernetes networking, security, and observability (35, 37, 38).
| Level | Learn | Practice | Checkpoint |
|---|---|---|---|
| 1 | Why meshes exist, sidecar pattern, L4 vs L7, Envoy listeners/routes/clusters, xDS | 43, 44 | - [ ] Route and retry traffic with an Envoy config you wrote yourself |
| 2 | Linkerd and Istio basics: injection, automatic mTLS, golden metrics, Gateway API, ingress gateway | 45, 46 | - [ ] Show encrypted pod-to-pod traffic and per-route success rates with zero code changes |
| 3 | Traffic management (canary, mirroring, fault injection, circuit breaking), zero-trust (SPIFFE, AuthorizationPolicy), tracing | 47–49 | - [ ] A 90/10 canary with default-deny policies, and one slow request traced across services |
| 4 | Sidecarless meshes (Istio ambient, Cilium), overhead, control-plane upgrades, multi-cluster, when not to use a mesh | 50 | - [ ] A measured comparison and a written recommendation; pass ICA |
Resources: Istio docs · Linkerd docs · Envoy docs · Istio in Action (Posta, Maloku) · Solo.io Academy and Buoyant Service Mesh Academy (free) · Isovalent labs for Cilium (free) · Gateway API (GAMMA).
| Level | Learn | Practice | Checkpoint |
|---|---|---|---|
| 1 | Metrics vs logs vs traces, dashboards, basic alerts | 11 | - [ ] Build a dashboard that answers "is it healthy?" at a glance |
| 2 | PromQL, LogQL, RED/USE, cardinality, alert design, SLOs | 11, 25 | - [ ] Write alerts that fire on user-facing symptoms, with no noise |
| 3 | Distributed tracing (OpenTelemetry), profiling, observability at scale | 38, 49 | - [ ] Trace one slow request across services to its cause |
Resources: Prometheus docs · OpenTelemetry docs · Observability Engineering (Majors, Fong-Jones, Miranda) · Google SRE books (free).
| Level | Learn | Practice | Checkpoint |
|---|---|---|---|
| 1 | SSH keys, firewalls, least privilege, secrets out of git, updates | 01, 07 | - [ ] Explain every secret in this repo: where it lives, who can read it |
| 2 | TLS/PKI, vulnerability scanning (Trivy), secrets managers (Vault), OWASP Top 10 | 21, 37, Phase 13 | - [ ] Make CI fail on critical vulnerabilities in your image |
| 3 | Supply chain (SBOM, cosign signatures, SLSA), policy as code (OPA/Kyverno), zero-trust service identity, threat modelling | 37, 48, Phase 13 | - [ ] Only signed images can run in your cluster |
Resources: OWASP Top 10 · SLSA · CIS Benchmarks · Container Security (Liz Rice).
| Level | Learn | Practice | Checkpoint |
|---|---|---|---|
| 1 | Running Postgres and Redis in containers, connections, basic SQL; what makes stateful services hard | 04, 51 | - [ ] Explain for each OpsForge data service how it's deployed, backed up, and monitored |
| 2 | Backups and restore tests, connection pooling, indexes, replication; MySQL vs PostgreSQL; Redis persistence and eviction; MongoDB replica sets | 03, 24, 52–55 | - [ ] Restore production data onto a fresh server, timed, for two different databases |
| 3 | Operators and automatic failover, point-in-time recovery, sharding; messaging with RabbitMQ and Kafka (acks, dead letters, partitions, consumer lag) | 52–57 | - [ ] Kill a primary or a broker under load and show what was (not) lost |
| 4 | Event-driven design (outbox, idempotent consumers), zero-downtime migrations, disaster recovery drills | 58, 59 | - [ ] A DR drill from zero with measured RPO and RTO |
Resources: Designing Data-Intensive Applications (Martin Kleppmann): the most valuable book on this list · PostgreSQL docs · PostgreSQL 14 Internals (Egor Rogov, free from Postgres Professional) · CloudNativePG docs · High Performance MySQL · Valkey / Redis docs · MongoDB University (free) · RabbitMQ tutorials · Kafka: The Definitive Guide · Strimzi docs · Enterprise Integration Patterns (Hohpe, Woolf).
| Level | Learn | Practice | Checkpoint |
|---|---|---|---|
| 1 | Availability, redundancy, health checks, graceful shutdown | 04, 08 | - [ ] Explain why a deploy can cause errors and how to prevent it |
| 2 | Load testing, capacity planning, horizontal scaling, caching, queues | 22–24 | - [ ] Find your system's breaking point and its bottleneck |
| 3 | SLOs and error budgets, chaos engineering, incident response and postmortems | 25, 47, Phase 14 | - [ ] Lead a (practice) incident and write a blameless postmortem |
Resources: Site Reliability Engineering + The SRE Workbook (free) · Release It! (Michael Nygard) · incident.io guide.
| Level | Learn | Checkpoint |
|---|---|---|
| 1 | Documentation, clear PRs, asking good questions | - [ ] Every lab README lets someone else repeat it |
| 2 | DevOps culture, DORA, blameless postmortems, on-call | - [ ] Write a runbook someone else can follow at 3 a.m. |
| 3 | Platform engineering (internal platforms, "golden paths"), mentoring, technical writing | - [ ] Teach a lab to someone, or publish it as a blog post |
Resources: The Phoenix Project (novel) · The DevOps Handbook · Team Topologies · platformengineering.org.
Phases 9 (GCP, labs 26–32), 10 (Kubernetes in depth, labs 33–42), 11 (service mesh, labs 43–50), and 12 (data services and messaging, labs 51–59) exist as lab stubs. Still listed here only:
| Phase | Labs |
|---|---|
| 13 — DevSecOps | Image scanning and SBOMs in CI · signing images with cosign · Vault for secrets · supply-chain levels (SLSA) · threat modelling one system |
| 14 — SRE and delivery | dev/staging/prod promotion · incident simulation and postmortem · DORA metrics for this repo |
| Stage | Months | Focus | Labs | Milestone |
|---|---|---|---|---|
| A. Foundations | 1–3 | Linux, networking, Bash, Git, containers | 00, 01–06, 13, 17 | All Level 1 checkpoints |
| B. Automation | 4–6 | CI/CD, IaC, observability, backups, a real VPS online | 03, 07–11 | The VPS rebuilt from code; alerts on your phone |
| C. Platforms | 7–10 | Kubernetes basics, Google Cloud, scaling | 12, 22–24, 26–31, 33–36 | Associate Cloud Engineer; CKAD |
| D. Depth | 11–15 | Kubernetes deep dive, OS internals, security, reliability | 14–16, 18–21, 25, 32, 37–42 | CKA, then CKS; a public portfolio |
| E. Data and messaging | 16–18 | PostgreSQL, MySQL, Redis, MongoDB, RabbitMQ, Kafka; event-driven design; DR | 51–59 | A DR drill with measured RPO/RTO; optional Kafka admin certification |
| F. Service mesh | 19–21 | Envoy, Linkerd, Istio in depth, sidecarless meshes | 43–50 | ICA (Istio Certified Associate); a measured mesh comparison in your portfolio |
| Time | Activity |
|---|---|
| ~2 h | Learn: read the lab's knowledge.md and one resource; answer "Before you start" |
| ~5 h | Practice: do the lab yourself; break things on purpose |
| ~1 h | Write: "What broke", "Lessons learned", "My notes"; tick checkpoints here |
| ~30 min | Review: what's next, and one thing you'd explain differently now |
Stages E and F are interchangeable: pick the one your next job needs most. Data services come up in almost every role; service meshes mainly in larger platform teams.
Small and steady beats long and rare: five sessions of 1.5 hours teach more than one 8-hour day.
| Certification | When | Why |
|---|---|---|
| HashiCorp Terraform Associate | Any time after lab 10 | Quick, validates IaC basics |
| Google Associate Cloud Engineer | After labs 26–31 | The entry GCP certification: hands-on cloud operations |
| CKAD (Certified Kubernetes Application Developer) | After labs 33–36 | Hands-on, about using Kubernetes; a good first Kubernetes exam |
| CKA (Certified Kubernetes Administrator) | After labs 37–40 | Hands-on, about running clusters; highly respected |
| CKS (Kubernetes Security) | After CKA (required) | Security depth |
| Google Professional Cloud DevOps Engineer | After Stage D | Senior-level: SRE practices on GCP |
| ICA (Istio Certified Associate) | After labs 46–49 | Hands-on Istio: traffic, security, observability |
| CCA (Cilium Certified Associate) | Optional, after lab 50 | eBPF networking and Cilium's mesh |
| Confluent Certified Administrator for Apache Kafka | Optional, after lab 57 | If your work involves Kafka |
| MongoDB Associate DBA | Optional, after lab 55 | If your work involves MongoDB |
Certificates open doors; the labs and your write-ups are what convince people once the door is open.
These aren't tools, but they explain why the tools behave the way they do. Study them alongside the labs, roughly when the "when" column says.
| Topic | What to learn | When | Resources |
|---|---|---|---|
| Distributed systems | CAP and PACELC, consistency models, consensus (Raft), leader election, idempotency, clocks and ordering, why "exactly once" is hard | With labs 39–40 (etcd) and 52–57 | Designing Data-Intensive Applications · MIT 6.5840 Distributed Systems (free lectures) · Raft visualisation · Martin Kleppmann's distributed systems lectures (free) |
| System design | Requirements, back-of-the-envelope estimates, trade-offs, diagrams; designing a whole system end to end | After Stage C, then ongoing (interviews) | System Design Primer (free) · System Design Interview vol. 1–2 (Alex Xu) |
| Identity for humans | OAuth2, OIDC, SSO, sessions vs tokens; a login (Keycloak + oauth2-proxy) in front of Grafana, Argo CD, Portainer | With labs 37 and 48 | OAuth 2.0 Simplified (free) · Keycloak docs · oauth2-proxy |
| APIs | REST design, gRPC and protobuf, versioning, API gateways, rate limits | With labs 43–47 | gRPC docs · Google API design guide |
| Disaster recovery | RPO and RTO, backup strategies (3-2-1), DR tiers, multi-region failover, DR drills | With labs 03 and 59 | Google Cloud DR planning guide |
| CLI fluency | jq, yq, awk, sed, regex, tmux, vim basics, shell history and aliases |
From day one; 15 minutes a day | The Missing Semester (free, MIT) · The Art of Command Line · jq manual |
| Skill | How to practise |
|---|---|
| Interviews | Live troubleshooting ("this pod won't start"), system design, and telling incident stories. Every lab's "What broke" section is a ready-made story: symptom, investigation, cause, fix |
| Portfolio | Finished labs with your own notes; a blog post every 2–3 labs; a clean README with the roadmap (this repo) |
| Technical English | Write all docs, commits, and notes in English (you already do); read postmortems and RFCs; explain a lab out loud |
| Staying current | Release notes of the tools you use; the CNCF landscape; one newsletter such as SRE Weekly |
- Edge and CDN: Cloudflare, caching, WAF, DDoS protection
- Compliance: SOC 2, ISO 27001, GDPR basics, audit trails
- MLOps and LLMOps: GPU workloads on Kubernetes, model serving
- eBPF in depth: Cilium, Tetragon, kernel-level observability
- A second cloud: AWS or Azure, mapped from GCP (most concepts transfer one to one)
- Run something real, for a long time. Keep the VPS online with real use (even just yours): upgrades, expired certificates, a full disk. Problems you didn't plan are the best teachers.
- Break things on purpose, then write down what you saw. Every lab does this; keep doing it.
- Explain it. If you can't explain it simply, you don't know it yet. Blog posts, a talk at a meetup, or helping someone online.
- Read other people's systems. Open-source Helm charts, Terraform modules, postmortems (danluu/post-mortems).
- Go general, then deep: be solid across all 14 domains (Level 2), and expert in two or three you enjoy most.