So interesting, a lot of comments seem to be negative experiences. I haven’t used Airflow at scale yet but would love to convert our extremely limited, internally built orchestrator + jobs over to Airflow. I think it would allow us TO scale, at least for some time. I think a lot of companies are still really behind the times. Our DAGs are fairly simple, and Airflow has been a major improvement in my testing. The UI is great for helping me debug jobs / monitoring feed health / backfilling. DAG writing has been a bit frustrating but is much improved format over the internal systems we have. Am I just naive? Is everyone writing extremely complex graphs? Is this operational complexity due mostly to K8s (I’ve just been playing with Celery)? Anyone enjoying using Airflow?