Hi all, I have created a PR https://github.com/apache/hudi/pull/12856 for RFC 90 for enabling clustering plans to be "cancellable"
Background: Clustering is a table service that assists with optimizing table/files layout in Hudi to speed up read queries. Clustering table service plans can delay ingestion writes from updating a dataset with recent data if potential write conflicts are detected. Furthermore, a clustering plan that isn't executed to completion for a large amount of time (due to repeated failures, application misconfiguration, or insufficient resources) will degrade the read/write performance of a dataset due to delaying clean and archival. This RFC proposes to support "Cancellable" Clustering plans. Support for such cancellable clustering plans will provide HUDI an avenue to fully cancel a clustering plan to allow other table service and ingestion writers to proceed and avoid possible starvation (based on user needs). Thanks, and any feedback will be appreciated. -- From, Krishen Bhan
