Folks, There have been many requests from users around supporting concurrency control for Hudi tables. I'm proposing we break down this ask into 2 phases. The first phase will focus on providing the ability to perform parallel writes to Hudi tables - this means as long as writes touch non-overlapping files, users will have the flexibility to mutate the tables from multiple writers. In phase 2, we will explore the idea of providing concurrency control semantics when 2 different writers want to mutate the same file or even record and how to provide serializability.
Phase 1 solves some of the outstanding requirements while is relatively less complex to implement and achieve than Phase 2. With such a breakdown, we will make progress on multi-writers for Hudi tables while also re-structuring the code and paving way for Phase 2. Please chime in. Thanks, Nishith
