I am trying to implement yarn user:queue mapping where user1:queue1 and user2:queue2 and mapreduce.queue.name=default.
Now all user1 MR jobs are going in default queue whereas user2 MR jobs are going to queue2. I'm confused with the behavior, either both user's MR job should go to default queue or to their respective queue.
I have ranger hive policy enabled for the users and yarn queue-override-mapping flag is set to false.
HDP version: 2.5.3
Update: The above worked after correcting the unix user:group mappings and OS level caching.
There was a caching issue in RHEL6, which linux admin fixed and after YARN restart everything was in place.