RocMarshal commented on PR #25734: URL: https://github.com/apache/flink/pull/25734#issuecomment-2522825677
With offline discussion( Thanks @zhuzhurk @1996fanrui for the help), we got 2 strategy options to implementation the features. I make 2 PRs: - [implementation-1](https://github.com/apache/flink/pull/25737) based on condidate-1(conclusion in the offline discussion) stratey. - The main diff is 'Add a greedy strategy for PipelinedRegionSchedulingStrategy to get all slots requests in once to support all slot requests could be sent to PhysicalSlotProvider in once.' - Non-refactor the existed Batching logic for resources requests at DeclarativeSlotPool side. - [implementation-2] (curent PR) based on candidate-2 strategy. - Refactor resource request wait mechanism to DeclarativeSlotPoolBridge side for Default Scheduler, which breaks a littile of the semantic of the original 'Batching logic for resources requests '. Hi, @zhuzhurk Could you help take a look for it if you had the free time ? Many thanks ! -- 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]
