All releases
Apache Airflowv3.0
Apache Airflow 3.0 — The Biggest Release Since 2.0
Airflow 3.0 is the most significant release since Airflow 2.0 in 2020. A completely rewritten React UI, a new multi-language TaskSDK, decoupled execution architecture, and formal removal of all 1.x legacy baggage make this a production-altering upgrade.
What's New
- 1Brand new React-based UI — faster, mobile-friendly, and redesigned from scratch (old Angular UI is gone)
- 2TaskSDK — author DAGs in Python, Java, Go, or any language via a standardised execution API
- 3Decoupled execution architecture — workers, scheduler, and API server can be scaled independently
- 4REST API v2 (OpenAPI 3.1) replaces the legacy Experimental API — all integrations should migrate
- 5Assets (formerly Datasets) with improved lineage tracking, versioning, and cross-DAG dependency graphs
- 6Backfill UI overhauled — trigger and monitor historical backfills directly from the UI without CLI
- 7Dropped support for Python 3.8 and 3.9 — Python 3.10+ required
Breaking Changes
- Stable REST API v1 removed — all integrations must migrate to REST API v2
- Python 3.8 and 3.9 no longer supported
- SubDagOperator removed — migrate to TaskGroups
- contrib providers removed from core — install separately from PyPI
- airflow.cfg parameter names changed in multiple sections — review migration guide
Upgrade Notes
Do not attempt a direct upgrade from 2.x in production. Use the official 2.x → 3.0 migration guide. Upgrade to Airflow 2.10 first, resolve all deprecation warnings, then migrate. Allow 2–4 weeks for large deployments.