After upgrading from Airflow 2.1.4 to 2.2.0 (or 2.2.1 or 2.2.2), we have a number of DAGs that fail to execute. The error Airflow provides is "Negsignal.SIGKILL". This happens when running locally with the SequentialExecutor AND when we deploy to our EKS Kubernetes clusters.
We can't seem to find any issues with the DAGs in question, they are pretty simple. Some DAGs run fine, some don't.
Has anything like this been reported? At this point the problem definitely seems to be with the Airflow 2.2.x line as downgrading back to 2.1.x fixes the issues.
Any suggestions would be greatly appreciated as we need to upgrade to get a fix for a few other known issues.
[2021-12-14, 19:51:36 UTC] {local_task_job.py:154} INFO - Task exited with return code Negsignal.SIGKILL
UPDATE 12-14
We have narrowed the issue down to DAGs that utilize the flask-appbuilder module. In a few places we use the cached_app() method to query Airflow users and roles to perform maintenance tasks. It appears that something has changed in Airflow 2.2.x that is causing the containers to be immediately terminated.
