Apache Airflow version

3.0.1

If "Other Airflow 2 version" selected, which one?

No response

What happened?

Hi,

normally I am cleaning up at the end of each DAG execution the DAG XCom's. I have now upgraded to the newest Airflow 3.0.1 version and found that this is not any more possible because of the following restriction: Direct database access via the ORM is not allowed in Airflow 3.0

Normally I was using the following code to clean up the XCom's:

from airflow.utils.db import provide_session
from airflow.models import XCom

@provide_session
def cleanup_xcom(session=None):
    logging.info('Cleaning XCom')
    session.query(XCom).filter(XCom.dag_id == "airflow_activity_context_migrator").delete()
    logging.info('XCom clean successful!')

Now I have tried also this one, which is based on the official XCom Documentation :

from airflow.models.xcom import XComModel

def cleanup_xcom():
    logging.info('Cleaning XCom')
    XComModel.clear(dag_id='airflow_test')
    logging.info('XCom clean successful!')

Still getting the same error and my DAG fails.

I am quite sure that, my first approach was not the best one, because I was querying and deleting elements directly, which I though is the problem. When I tested with the XComModel.clear() function and still getting the same error, I expected it to work, because it is a predefined function and the user passes only the required arguments, which is not so harmful.

What you think should happen instead?

I think there should be definitely a way, how the people can delete their XCom's inside a DAG and the XComModel.clear() function should have probably an exclusion for this new restriction, which forbites the direct db access via ORM.

How to reproduce

Just execute the following in any DAG:

from airflow.models.xcom import XComModel

def cleanup_xcom():
    logging.info('Cleaning XCom')
    XComModel.clear(dag_id='airflow_test')
    logging.info('XCom clean successful!')

Operating System

Red Hat Enterprise Linux 8.10 (Ootpa)

Versions of Apache Airflow Providers

No response

Deployment

Other Docker-based deployment

Deployment details

I am running Airflow 3.0.1 with KubernetesExecutor on Openshift

Anything else?

No response

Are you willing to submit PR?

• Yes I am willing to submit a PR!

Code of Conduct

• I agree to follow this project's Code of Conduct