uros-b commented on code in PR #57651: URL: https://github.com/apache/spark/pull/57651#discussion_r3865311637
########## core/src/main/scala/org/apache/spark/ExecutorAllocationManager.scala: ########## Review Comment: Looking at https://github.com/j1wonpark/spark/blob/c170af8c3c04cedf139562d20afef3d0f4b511ce/core/src/main/scala/org/apache/spark/ExecutorAllocationManager.scala#L948-L953, late failed TaskEnd leaves a phantom backlog timer. The new guard ignores late TaskStart bookkeeping, but a subsequent non-TaskKilled failure still calls onSchedulerBacklogged(). Since the stage is already completed, no retry occurs and addTime remains armed. A later stage can consequently bypass schedulerBacklogTimeout and request executors prematurely. Gate this path on the stage attempt still being active and add a failed-late-task regression test. -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
