Apache Airflow — Prerequisites
Airflow is a workflow orchestrator, not a data processing engine itself — it schedules and monitors other tools (Spark jobs, SQL scripts, API calls) running in the right order. That framing matters for what you need to know first.
What you need first
•Solid Python — Airflow DAGs (workflow definitions) are written in Python, and you'll write real Python logic (custom operators, sensors, branching conditions), not just declarative YAML. Comfortable functions, classes, and decorators are assumed.
•Understanding of what the tools you're orchestrating actually do — Airflow coordinates other systems (a Spark job, a dbt run, a database query, an API call); you need to understand those individual pieces before orchestrating them meaningfully. Airflow itself doesn't process data — it's the scheduler and dependency manager sitting above the tools that do.
•Basic scheduling/cron concepts — what a cron expression means, the difference between "run every hour" and "run once daily at 2am." Airflow's scheduling model builds on these concepts rather than replacing them entirely.
•Comfort with the command line and basic Docker — most modern Airflow deployments (including local learning setups) run via Docker Compose; basic Docker familiarity (starting/stopping containers, reading logs) is assumed.
Helpful but not required
•Prior experience with a different scheduler (cron, a different workflow tool) — useful context for what problems Airflow solves, but not required; this page explains Airflow's specific model directly.
•Kubernetes knowledge — relevant if you're deploying Airflow itself on Kubernetes or using the KubernetesExecutor, but not needed to learn Airflow's core concepts, which this page teaches assuming a simpler local/Docker Compose setup first.
A note on scope
This page teaches Airflow specifically — DAG authoring, scheduling, task dependencies, and operational patterns. It assumes you already know (or are learning separately) the actual data tools your DAGs will orchestrate — Airflow's own learning curve is entirely about coordination, not about teaching Spark, SQL, or any other specific data tool it might call.