Recently we find that when multiple jdbc connection concurrently add jars, hiveserver will throw StackOverflowError when serializeMapRedWork to hdfs, I find the relate issue hive-2666 is similar, but I think it missed the concurrently scenario.
I find it is because the classloader is changed which will lead to the infinite loop. Any one met this issue? Any suggestion?
Wangwenli 2013-04-09, 07:08