Bryan Vanderhoof created AIRFLOW-1204:
-----------------------------------------
Summary: Scheduler ignores start_date if an earlier successful dag
run exists
Key: AIRFLOW-1204
URL: https://issues.apache.org/jira/browse/AIRFLOW-1204
Project: Apache Airflow
Issue Type: Bug
Affects Versions: Airflow 1.8
Reporter: Bryan Vanderhoof
I have repeatedly run into a problem with the scheduler automatically
scheduling DAGs before their start_date if a previous run of the DAG exists in
the database.
For example, if a new DAG has a start_date of 2017-05-01, it will run starting
on that date as expected. However, if there's a existing run for execution_date
2017-04-01, the scheduler will automatically generate dag runs for 2017-04-02
through 2017-05-01, completely ignoring the start_date.
This also happens with backfills. Today, I began backfilling data for January
2017 on a DAG with a start date of 2017-05-01. The backfill began as expected,
but as soon as some of the January tasks began completing, the scheduler also
created dag runs for every day in February, March, and April.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
