Hi team, As a follow up to the discussion in https://lists.apache.org/thread/vff0q9oogxrsp0w1nzco2lqhy9b02pwz, I want to call for a lazy consensus on what we document about metadata database access approaches in the Airflow 3 upgrade guide for Airflow 2.x users.
We will be documenting: Approach 1: Airflow Python Client (REST API) - Use `apache-airflow-client` package to interact with Airflow metadata via REST API - Document as the primary/recommended solution Approach 2: Database Hooks (PostgresHook/MySqlHook) - Create a connection to metadata DB and use DB hooks to execute SQL directly - Document as a "known workaround" (NOT recommended) with explicit warnings based on suggestions on the other mail thread: - Can break in future versions (especially 3.2+) - Users are responsible for adapting when schema changes occur - Database schema is NOT a public API and can change without notice - Breaks task isolation (one of Airflow 3's core features) - No support guarantees - Performance implications Both approaches will be documented with pros and cons. Approach 3 (Direct SQLAlchemy Access) will NOT be documented at all as discussed. The lazy consensus will end on Friday (14 Nov) at 4 PM UTC (~72 hours from now). If there are no objections or significant concerns by that time, I will proceed with updating the documentation accordingly in this PR: https://github.com/apache/airflow/pull/57479 Thanks & Regards, Amogh Desai
