Hi all, I am running sparksql in hiveql dialect, the sql is like "select * from (select * from t1 order by t1.id desc) as ff". The sql succeed when it runs only once, but it failed when I run the sql five times at the same time. It seemed that the thread is dumped and executors are lost. The problem is not caused by memory or gc, the shufflle data is relative large, but the whole shuffle size is less than 3g onceand 15g five times. Does anyone have a good idea?
Thanks qinggangwa...@gmail.com
