Github user koeninger commented on the issue: https://github.com/apache/spark/pull/19789 My main comment is that if you have a situation where there's actually contention on the size of the cache, chances are things are going to be screwed up anyway due to consumers being recreated and losing any benefit of buffering. Are you running into this issue in situations where the cache is appropriately sized given the number of topicpartitions and executors?
--- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org