When a user submits multiple queries in the same HS2 session (using thrift interface), the query goes through the same TezSessionState and ends up being submitted to the same Tez AM, and then fails with the error "App master already running a DAG"
Is this by design? I looked into the code, and the comments as well as the code suggest support only for serial execution of queries within the same HiveServer2 session (on tez). Our stack is Hue 3.5 - HiveServer2 (Hive 0.13.1) - Tez 0.4.1. We are using job queues, and have turned on doAs flag for impersonation. I am still familiarizing myself with the code, to determine possible fixes, but wanted to get input from folks who have any context on this, in the meantime. -pala
