The Data Engineering Stack for 2027: What I’d Keep, What I’d Drop

Fewer tools, more of them owned by the same company, and a “skip it if” for every recommendation I make.

On 22 April 2026, Airflow 2 reached end of life. The version table in the Airflow README has listed it as EOL since that day, after limited maintenance started in October 2025. The last release of the line was 2.11.2. Anything still running on it gets no security fixes.

My team and I got there before the deadline. We run Airflow 3 internally and at our clients, and every Airflow 2 instance we were responsible for is now on 3.x. I expected this to hurt. It didn’t, at least not on our side. For the platform itself, the migration came down to defining a handful of new environment variables and updating the dependencies in our Airflow image. That was it. What our clients had to change inside their own DAGs, I honestly can’t tell you in detail, because they owned that part. But none of them came back with a migration that stalled. If you’ve been putting off the upgrade because you expect a rewrite, my experience says the platform side is the small part. Budget your time for testing your own DAGs, not for the infrastructure.

That migration made me look at the rest of the stack with the same question. Every tool you…