Hi Folks,
To close this one off I want to mention some more information we were able to
acquire. This may help, in particular people running Airflow on K8s.
If you define a custom XCom backend in your values.yaml configuration and
Airflow fails to load the class, the entire Chart deployment will
Looks like you replied just before me.
You should not need to do anything beyond confirming that airflow config
resolves the right xcom. Its usage should be the same. E.g. ti.xcom_push
etc.
Your dags and tasks should remain unchanged.
If you can get a terminal in the container then you should be able to do
with something like
from airflow.models.xcom import XCom
print(XCom.__name__)
i think that willl print the actual class that is being used
you can see how the import works in that module
see resolve_xcom_backend
I just discovered 'resolve_xcom_backend' and looked at the source code and it
appears to explicitly load the XCom class from the configuration.
https://airflow.apache.org/docs/apache-airflow/stable/_api/airflow/models/xcom/index.html#airflow.models.xcom.resolve_xcom_backend
Should this be called