zhijiangW commented on issue #8242: [FLINK-6227][network] Introduce the DataConsumptionException for downstream task failure URL: https://github.com/apache/flink/pull/8242#issuecomment-491258890 @tillrohrmann thanks for your further great suggestions! I totally agree the above points and the new exception category would make the semantics more clearly. To double confirm If my understanding is correct: - The new proposed `PartitionException` has three sub-type instances as `PartitionNotFoundException`, `PartitionCorruptionException` and `PartitionLookupException`. The job master would check the `PartitionException` for determining to restart the producer. Especially for instance of `PartitionLookupException`, the producer might need to be restarted in another machine. - Considering the current `SpilledSubpartitionView` would be replace by Stephan's new `BoundedBlockingSubpartition`, do you think it is necessary to throw `PartitionCorruptionException` in current code or focus on it after `BoundedBlockingSubpartition` is merged? - The producer could send `PartitionException` if fails to create sub partition view. On consumer side, if the exception is instanceof `PartitionNotFoundException`, it would keep the current logic of retrying to request partition. If instanceof `PartitionCorruptionException`, the consumer could fail directly. - I already created a separate jira for covering `PartitionLookupException`, and the connection lost after established would not be considered in this issue.
---------------------------------------------------------------- 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: [email protected] With regards, Apache Git Services
