zentol commented on pull request #15019: URL: https://github.com/apache/flink/pull/15019#issuecomment-785808932
One issues is that the semantics for the attempt numbers as a whole are not really well-defined; internally we use it to differentiate different deployments (for example, when we tried to create deterministic `ExecutionAttemptIDs`), but users probably think more of it like "how often did this subtask fail". I'm not even sure how valuable the attempt number is as a whole; I'd think that users either want to identify instable TaskManagers (where the question is "how many failures occur on this TM") or operators ("how often do subtasks of this operator fail?"), but the attempt number fulfills neither because it is also incremented for subtasks did not actually cause a failure. ---------------------------------------------------------------- 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