CI/CD Pipelines — Learning Roadmap
A structured path from zero to production-ready pipeline design. Each stage builds on the last — don't skip to Advanced patterns before Stage 2's fundamentals are second nature; canary deployments and GitOps are much harder to reason about if the basic build→test→deploy flow isn't already intuitive.
Stage 1 — Manual first, automate second (1 week)
Before automating anything, manually run your project's full build → test → deploy sequence from the command line, end to end, and write down every step. This sounds like a detour but isn't — you cannot correctly automate a process you haven't done manually and understood completely. Most broken first pipelines are broken because a step that "just works" on a developer's machine (an implicit .env file, a globally-installed tool) was never actually understood as a dependency.
Stage 2 — Your first real pipeline (1-2 weeks)
Set up GitHub Actions or GitLab CI (see Installation) on a real project — ideally one you already have running locally. Build a pipeline that: runs on every push, installs dependencies, runs your test suite, and fails loudly if tests fail. Get comfortable reading pipeline logs to debug a failure — this is the single most-used skill day to day, more than writing new pipeline YAML.
Stage 3 — Add deployment (1-2 weeks)
Extend Stage 2's pipeline to actually deploy somewhere — a free-tier PaaS (Render, Railway, Vercel) is a fine target for learning, no need to provision real cloud infrastructure yet. Add a manual approval gate before deployment so you internalize the Continuous Delivery vs. Continuous Deployment distinction hands-on, not just conceptually.
Stage 4 — Multi-environment + secrets discipline (2 weeks)
Add a staging environment distinct from production, and practice promoting the same built artifact through both (see Intermediate). Set up real secrets management — cloud credentials via your CI platform's secrets store, never hardcoded. This stage is where most of the "intermediate DevOps engineer" interview questions live.
Stage 5 — Speed and reliability (2 weeks)
Add dependency caching and, if your test suite is large enough to matter, parallelized/sharded test execution (see Intermediate). Add a genuine rollback path — practice actually triggering it, not just designing it on paper.
Stage 6 — Production-grade patterns (ongoing, project-driven)
Pick one advanced pattern and implement it for real rather than reading about all of them shallowly: blue-green or canary deployment on a real (even personal-scale) project, or a GitOps setup with ArgoCD/Flux against a small Kubernetes cluster. Depth on one advanced pattern, actually built and broken and fixed, teaches more than surface familiarity with all of them.