zhijiangW edited a comment on issue #8242: [FLINK-6227][network] Introduce the DataConsumptionException for downstream task failure URL: https://github.com/apache/flink/pull/8242#issuecomment-490011255 Yes, the blacklist could solve the hardware problems future. As for `d`, you are right that the `IOException` might be caused by multiple factors, then we do not cover it in this PR. As for `f`, if I understand correctly, the scenario is similar as the case `a` I mentioned above. If the TM is unreachable, the first phase of establishing connection would fail by `ConnectionTimeoutException` or `ConnectionRefusedException` etc. I proposed `DataConnectionException` not `PartitionNotFound` in `a`, because it might be caused by network temporary problem. After retrying the connection later it might be success. In other words, from the `ConnectionTimeoutException` or `ConnectionRefusedException` received by consumer, it could not estimate whether the producer TM is really lost or the temporary network issue. After the connection established, the second phase is to send logic `PartitionRequest` message, if at this time the producer TM is unreachable to cause any network exceptions, we still can not distinguish whether it is caused by TM really lost or temporary network issue, only if the consumer receives specific `PartitionNotFoundException` via network which is the cases of b, c. So we might only focus on b, c currently?
---------------------------------------------------------------- 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 With regards, Apache Git Services