Github user jerryshao commented on the issue: https://github.com/apache/spark/pull/19396 IMO I think it might be better to throw an exception instead of shifting to another shuffle. Since user want to use external shuffle explicitly, letting user to know the issues and fix the issue would be better. Besides, this will lead to an issue if Cluster manager changing to not start shuffle service, while Spark application still assume external shuffle is on, and tries to connect to external shuffle service.
--- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org