rdblue commented on issue #23401: [SPARK-26513][Core] : Trigger GC on executor node idle URL: https://github.com/apache/spark/pull/23401#issuecomment-450700817 Cached data or broadcast tables could easily take a large portion of the heap, making it more likely that the full GC is triggered, but unlikely that it will collect any space. That's one reason why I think this would need to interact with spark.memory.fraction or maybe the memory manager directly to see how much potential free space there is. Also, the paper's results are based on setting the threshold to 0.3, so I thought it was likely that the threshold would be considerably lower than 0.8.
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
