Github user squito commented on the issue: https://github.com/apache/spark/pull/21698 > What if the user does't provide a distributed file system path? E.g., you can read from Kafka and write them back to Kafka and such workloads don't need a distributed file system in standalone mode. yeah that is a good point. I think we want a solution which is correct without checkpointing (eg. always sort), but perhaps can leverage checkpointing when possible to avoid the overhead.
--- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org