-
-
Notifications
You must be signed in to change notification settings - Fork 3
Description
Current Situation
We offer the customer to inject DAGs
into airflow in two ways:
- Customers can mount a
configMap
into the Airflow pod and store theDAGs
there. - Customers can use GitSync to host
DAGs
in an git repository.
Both ways need a way to interact either with kubernetes itself or with git.
Proposed Situation
Customers can log into Airflow and edit DAGs
in the Airflow UI.
To archive this, we'd need to include an editor in airflow dockerimage and make it configurable through the operators CRD. I'd propose https://github.com/andreax79/airflow-code-editor as an candidate.
Advantage
Now where #573 is merged, we offer deeply integrated Open Policy Agent
support, which would lead to the advantage to offer role based editor access to data engineers or product owners. There wouldn't be a need to have knowledge of kuberentes or git to modify, develop and deploy DAGs
. In my opinion this would increase usability and accessibility of Airflow and could empower the customer.
Additionally, it would make the development process of DAGs
much easier as we would move from develop -> deploy -> test to develop -> test.
Evaluation
I expect this to be a rather large feature. We'd need to think about roles and how to configure those for editor access without OPA
too. We need to be able to integrate with GitSync and implement access, provisioning and usage of PVCs to store DAGs
written in the Airflow UI. Role based access may also apply to files stored in PVCs. We introduce at least another dependency we might need to maintain.