I85PREP
⌘ K
Airflow DAG Authoring study materials
Airflow DAG Authoring study resources

Airflow DAG Authoring: DAG Scheduling and Time Semantics

Airflow DAG Authoring · Work through your exam, one topic at a time.

All guide sections

DAG Scheduling and Time Semantics

Logical date, data interval, and run time

Airflow schedules by data interval. A run's logical_date (formerly execution_date) is the start of the interval it processes. With schedule='@daily' and start_date=2024-03-01, the run with logical date 2024-03-01 covers data from 2024-03-01T00:00 to 2024-03-02T00:00 and is triggered at the end of that interval (2024-03-02T00:00). Templates like {{ ds }} resolve to the logical date, which is exactly the partition the run should process.

Schedules can be cron strings, timedelta, None (manual only), or timetables for complex calendars. catchup=True backfills every interval since start_date; max_active_runs limits concurrency; catchup=False starts from the latest interval.

Traps candidates hit

  • Expecting today's data in today's run. A daily run for 2024-03-01 does not start until 2024-03-02.
  • start_date in the future means no runs until that time passes; start_date with catchup=True can flood the scheduler.
  • schedule_interval is the legacy name; modern Airflow uses schedule. Both may appear in questions.
  • Time zones matter. Naive datetimes are interpreted in the deployment's default timezone; use pendulum/timezone.utc explicitly.
  • end_date stops new runs but does not cancel running ones.

Rule of thumb

"The run processes the interval that ends when it starts." If a question asks which date a run handles, answer with the interval start, not the wall-clock trigger time.

Worked example

start_date=pendulum.datetime(2024,3,1,tz='UTC'), schedule='@daily', catchup=False. The first run has logical date 2024-03-01, and {{ ds }} renders 2024-03-01 even though execution begins on 2024-03-02.