azagrebin commented on issue #7186: [FLINK-10941] Keep slots which contain unconsumed result partitions URL: https://github.com/apache/flink/pull/7186#issuecomment-468605747 @zhijiangW Thanks for your answer. The consumer will close the connection on its side. If consumer fails or somehow realises that `CloseRequest` failed then I think the job and producer will be canceled by `JobMaster` and producer will get `sendFailIntermediateResultPartitionsRpcCall` to release partitions. My question is if `CloseRequest` is not received by producer and there can be a scenario when consumer still treats the close operation as success (if I understood you correctly) and reports success to `JobMaster`. I might miss a code path which would prevent the partition from lingering. How does it become aware about consumer is done?
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services
