curcur edited a comment on pull request #16844: URL: https://github.com/apache/flink/pull/16844#issuecomment-900431151
Thanks @wsry for fixing up! But I have a small concern about this change. 1. It is necessary to recycle BufferBuilder in the BufferWritingResultPartition#close method, but not sure why we add `closeBufferPool()` in the `fail` method. 2. Originally, `fail()` is responsible for propagating failure and release `**buffers**`, but not the `**buffer pool**`. `closeNetworkResources` is responsible to close network-related resources after `fail()` is called. In the `TaskCanceler`, `closeNetworkResources` is called to recycle resources and finally, BufferWritingResultPartition#close() will also be called as well as `closeBufferPool()` in `super.close()`. I am confused why we break this contract and include resource clean-up in fail(), and have to include fail in a cancler? Put in another way, why simply add point 1 (the close method added in BufferWritingResultPartition) is not enough? Also, how about the `release` method (the same method to release buffer consumers) instead of `close` method. That may be a better place? Let's discuss and sync up offline tomorrow. -- 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]
