This message was deleted.

looking at code where the ERROR printed “failed to stop in a timely manner, killing task” I’ve traced to call of this function https://github.com/apache/druid/blob/cc89c661d0aebf95bb85f01bfbedf76251e2f307/inde[…]xing/seekablestream/SeekableStreamIndexTaskClientAsyncImpl.java which seems to return false, either when there is http error or when it’s notAvailable(not sure what it means) anyway I think in case where task already finished, it can’t respond to http requests…so it’s not surprising that it’s ‘failed’ to stop it I’m thinking there is some missing synchronisation piece since the task should have already marked as ‘done’/success when it was paused at first place, or alternatively the race between 2 threads above should be solved in some way? Any thoughts?

Thanks, Igor. Let me review this further

thanks @Brandon Tan , appreciate it

Hi @Igor Berman. Thanks for the thorough analysis. There does seem to be a race condition when it comes to killing the task replicas

thanks @Amatya Avadhanula I’ll open ticket than