We switched from Airflow because we want engineers solving problems, not coding orchestration. Kestra delivers end-to-end automation with the robustness we need at our scale. Few companies operate at this level, especially in AI/ML.
Airflow 2 is No Longer Maintained. Reaffirm or Rethink Your Orchestration Model?
Airflow 2 enters end of life on April 22, 2026. Before you commit to upgrading, it's worth asking an important question: is Airflow 3 the right architecture for the next five years, or is this the moment to rethink your orchestration model entirely?
What's inside
- Exactly what changed in Airflow 3 and what stayed the same, so you can see the two approaches
- Where the Airflow model starts to work against you as your team scales
- How language-agnostic declarative orchestration differs, and which fits your team
- How to migrate incrementally so you don't have to stop everything and rebuild
Download the Free Guide
What you'll come away with
What changed in Airflow 3
Airflow 3 is a big rewrite. Schedules, executors, versioning, backfills, and UI all moved — we cover what's new, what broke, and what's just renamed so you don't get surprised mid-migration.
The architectural fork in the road
Code that runs fine in Airflow 2 will not port cleanly. Different DAG shapes, executors, and deployment models lead you to one of two choices: an in-place upgrade or a re-platform decision.
A proven migration path
Whether you're moving incrementally, running both in parallel, or rebuilding on a declarative orchestrator, we walk through concrete migration paths and the trade-offs of each.
Who this is for
Data Engineers
You're on Airflow 2 and have to make a call: upgrade, migrate, or modernize. You want a clear, defensible framework before you commit budget and roadmap time.
CTOs
You're balancing migration effort vs. platform debt. You need to understand how the switching cost compares to the ongoing cost of Airflow 2's operational quirks and the Airflow 3 upgrade.
Platform Engineers
You'll live with the decision day-to-day. You want to know what actually changes in your deploy pipeline, observability stack, and on-call rotation — beyond the release notes.
Kestra vs. Airflow 3.0
Choosing between two orchestrators is a team, deployment, and operational call as much as an architecture one.
See the full comparison| | | |
|---|---|---|
| Workflow definition | Declarative YAML | Python DAGs |
| Languages supported | Any (Python, SQL, Bash, R, Go, ...) | Python-first |
| Deployment | Single Docker Compose | API server + DAG processor + metadata DB + executor |
| Time to first workflow | ~5 minutes | ~30 minutes |
| Multi-tenancy | Namespace isolation + RBAC | Limited |
200engineers onboarded
2xfaster workflow iteration
0pipeline failures
Ready to explore your options?
Talk to our team about what migration looks like, or start building with Kestra for free.