Is there any way to get informative errors from python scripts in hadoop streaming?
if you have access to the jobtracker for the cluster where you are running you can get to the stderr/stdout of the script by finding the job and looking for the tasks that failed.