Github user cloud-fan commented on the issue: https://github.com/apache/spark/pull/20490 AFAIK, there is no KILL mechanism for speculative tasks, which means, if multiple task attempts are running at the same time and one of them finishes earlier, the others will keep running until finished. But your concern is valid, users can manually cancel a job/stage, and then the task thread will be interrupted. Fortunatelly this case is correctly handled. The writing logic is wrapped by `try-catch`. If the task is interrupted, `InterruptedException` will be thrown and caught, and `DataWriter.abort` will be called.
--- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org