Migrating from UC4 Automic to Apache Airflow is far from a lift-and-shift exercise. UC4 offers advanced scheduling primitives that data teams rely on daily — and Airflow doesn’t replicate them out of the box.

At eBay, we migrated thousands of business-critical UC4 workflows onto our Airflow 2.10 platform. Rather than forcing teams to change how they operate, we built the missing capabilities natively into Airflow:

  • Breakpoints — pause a pipeline at a specific task for inspection without failing the run
  • Skip logic — dynamically bypass tasks or task groups at runtime
  • Calendar-aware scheduling — replicate UC4’s calendar model as custom Airflow timetables
  • Pipeline pause/resume — operator-triggered suspension of in-flight DAG runs with state consistency

We’ll share the engineering trade-offs, architectural constraints we hit, and patterns reusable beyond eBay’s stack.

Xiongfeng Song

Software Engineer at eBay