Airflow DAG and DAG Run

An Airflow DAG defines a directed acyclic graph of tasks. Edges express dependencies: direction determines what is upstream, and the graph cannot contain a cycle. Independent tasks can run concurrently when their conditions and available capacity permit.

A DAG Run is one execution of that definition. Each run has its own task instances and scheduling context. The same task definition can therefore be successful in one run and failed in another. Dependencies within a run do not automatically serialize all runs.

TaskFlow calls made while constructing the DAG create task references and dependencies; they do not immediately execute the decorated function body. Keep repeated parsing free of source queries and external side effects. A graph edge does not transfer a worker-local file or establish data quality.

See Apache Airflow Fundamentals: DAGs, Tasks, and Scheduling for worked examples.

Further reading

• Airflow 3.1.0 Documentation, Dags