fhan688 commented on PR #18897: URL: https://github.com/apache/hudi/pull/18897#issuecomment-4610108034
> > This improves bucket routing consistency for simple bucket index writes, especially when writer parallelism changes. > > Can you explain a bit more about this part, the record keys are routed based on bucket ids instead of tasks, even if the task parallelism changes, the same key should still route to the same bucket? Thanks for the correction. I agree that the current local simple bucket logic keeps the key-to-bucket mapping stable. A parallelism change only affects the bucket-to-task mapping, and both `partitionCustom` and `BucketStreamWriteFunction#bootstrapIndexIfNeed` recompute that mapping with the current parallelism, so it is not accurate to say that parallelism changes cause key routing inconsistency. I updated the motivation to be more precise: this change centralizes bucket-to-task assignment through the timeline service by reusing the existing remote partitioner capability. This makes the routing side and bucket-loading side use the same remote assignment source, instead of maintaining separate local assignment calculations. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
