Description
When query 17 is run, some mappers from Map 1 currently fail (due to unwrap issue, and also due to HIVE-10112).
This query has 1000+ reducers; if they are ran in llap, they all queue up, and the query locks up.
If only mappers run in LLAP, query completes.
Related: https://issues.apache.org/jira/browse/HIVE-10029
This is expected at the moment. Until we support pre-empting tasks / removal of tasks from queues.