Control Flow for Apache Airflows
Airflow control flow decides the order and conditions under which tasks run. Dependencies are set with the bitshift operators, trigger rules change when a downstream task starts, branching selects paths, and retries, timeouts and SLAs make workflows reliable when tasks fail or hang.
Stage: Authoring and Scheduling · Part of the Apache Airflow roadmap · Last reviewed 19 November 2026
What to learn in Control Flow
- Set task dependencies using >> and the chain helper
- Choose a trigger rule such as all_done or one_failed
- Branch a workflow with the BranchPythonOperator
- Configure retries, retry delay and execution timeout
- Make tasks idempotent so a retry does not duplicate data
Control Flow interview questions
What are trigger rules in Airflow?
A trigger rule decides when a task may run based on the state of its upstream tasks. The default is all_success. Others include all_done to run regardless of failures, one_failed for alert tasks, and none_failed_min_one_success, which is commonly needed after a branch.
Why should Airflow tasks be idempotent?
Tasks are retried after failures and may be rerun for backfills, so running a task twice must give the same result as running it once. For example, overwrite a partition or use upserts rather than blindly appending rows, or retries will create duplicates.
Where Control Flow fits in the Apache Airflow roadmap
Control Flow belongs to the Authoring and Scheduling stage (intermediate level). Build workflows.
Other units in this stage
Back to the full Apache Airflow roadmap