Airflow 3: data_interval_end == data_interval_start?

[Djoren]

hi all,

I upgraded to Airflow 3.0.1, from Airflow 2.10.5, and read one of the breaking changes is redefining logical_date (from being equal to data_interval_start before to run_after now). However, I'm also finding the breaking change that for every DAG the data_interval_start == data_interval_end == logical_date.

I could not find anything in documentation about interval start and end dates being made equal. I'd like to get help understanding if this finding is correct, and if so what the intention behind this was?

Thanks

[yxl1108]

set ENV AIRFLOW__SCHEDULER__CREATE_CRON_DATA_INTERVALS=True will revert AF3 back to old behavior, data_interval_start and data_interval_end cover the span of the interval

[novandwiatmaja]

I have same issue. after set env AIRFLOW__SCHEDULER__CREATE_CRON_DATA_INTERVALS=True how to resolve?

[vietvudanh]

I am having the same problem. Say if I do ETL which requires today run_id cover yesterday data: Now that logical_date = start = end = today, I need to change all code to {{ ds }} subtract 01 day?

[yxl1108]

I am using Astronomer airflow, all I did was to add the following line to Dockerfile, and now in airflow 3, the behavior is the same as what's in airflow 2
ENV AIRFLOW__SCHEDULER__CREATE_CRON_DATA_INTERVALS=True

[novandwiatmaja]

If you set AIRFLOW__SCHEDULER__CREATE_CRON_DATA_INTERVALS=True, I think that’s not the best decision, as it’s not future-proof.

Instead, I took a different approach using:

from airflow.timetables.interval import CronDataIntervalTimetable

schedule = CronDataIntervalTimetable("0 0 * * *", timezone="UTC")

For existing tasks, the interval may not change unless you delete the historical task data from the database. However, future DAG runs will follow the correct data interval start and end.