Airflow DAGs Live in Git

Promoting DAGs from development to production is risk-free with no surprises.

The DAG code is Consistent across all Karrots tools

If the underlying code runs elsewhere in Karrots, e.g. Jupyter, then it will run correctly in Airflow — no unpleasant surprises!

Breakpoint Debug Airflow DAGs!

Airflow is a dominant, powerful tool, but it’s complex! One area where data people struggle: runtime errors. DAGs are not Python programs! You write them in a domain-specific-language. Even where elements of the DAG run is unknown to the developer! Karrots allows a data analyst to project their laptop into the cloud, in developer environments only, so that they can breakpoint-debug the DAG elements as they run in the cloud! This reduces the cost and complexity of debugging Airflow DAGs by about 100x. We find that data analysts, who generally not developers, cannot understand why a DAG has failed, let alone debug it. With Karrots, they are trivially able to see what their code does and are able to fix errors in minutes!