Apache Airflow is the core of many data warehouses. It's a very mature technology, proven to be stable and flexible, a swiss-army knife of Data Engineers. In 2021 project bumped the major version from 1.X to 2.X, modifying all the components, making the upgrade process not trivial, especially if company-critical processes run there. And that's exactly the challenge we faced recently with one of the major Getindata clients. The legacy Airflow 1.10.5, extended with many plugins and custom operators was controlled by ansible and running on VMs. We not only bumped the version, but also moved components to run entirely on Kubernetes. During the webinar I'd like to share with you the challenges we faced, what did we do to mitigate the risks and what didn't go exactly as planned If you're still on Airflow 1.X, the content will help you to transition smoothly. But even if you're not Airflow 1.X user, I will convince you why upgrading components as often as possible is the key to the success.