zhijiangW commented on pull request #11687: URL: https://github.com/apache/flink/pull/11687#issuecomment-623538520
Thanks for the further reviews @pnowojski ! > I'm still not sure if this is the right approach. It's better compared to the original proposal, at a cost of much more added lines of code, while the fundamental issues still remain the same: input channels are even more complicated compared to master branch. Now input channels are interconnected with BufferManager (both Local/RemoteInputChannel and BufferManager are calling/accessing one another many times during a single method invocation) and they are intertwined with RecoveredInputChannel. Actually I am also a bit torn while implementation and I am also not quite sure which option is the best approach, unless we can fully implement every option to compare through. When I tried to implement the option of working on `CheckpointedInputGate`, I also encountered with some troubles/confusing in other different aspects to make me quit finally. ---------------------------------------------------------------- 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. For queries about this service, please contact Infrastructure at: us...@infra.apache.org
