Sometimes it makes no sense to backfill old DAG runs. For example, when we retrieve data from a REST API, which always returns the current state, when we use Airflow to send a newsletter, or when the DAG run computes the entire data history every time it runs, so the execution date does not matter.

In Airflow, there are two ways to prevent the DAG from backfilling old runs.

We can set the catchup parameter of a DAG to False. In this case, Airflow will never create DAG runs with the execution date in the past.

1
2
3
4
dag = DAG('example_dag',
        catchup=False,
        ... # other parameters
    )

The second method is to include the LatestOnlyOperator operator inside the DAG. This operator stops DAG execution if the current run is not the latest one. This approach is useful when we want to backfill only some of the tasks and skip others. To understand how to use the LatestOnlyOperator, take a look at this blog post.

Parsing machine learning logs with Ahana, a managed Presto service, and Cube, a headless BI solution

Check out my article published on the Cube.dev blog!