Prasanth Jayachandran created HIVE-16687:
--------------------------------------------
Summary: LLAP: Hash table loader memory estimation should consider
current state of cluster
Key: HIVE-16687
URL: https://issues.apache.org/jira/browse/HIVE-16687
Project: Hive
Issue Type: Bug
Components: llap
Affects Versions: 3.0.0
Reporter: Prasanth Jayachandran
Hash table loader memory monitoring currently doesn't take into account the
state of the cluster (other fragments loading hash tables, overall overall
memory usage etc.). At current state, even when there is enough memory in the
daemon, a query can get killed as it had exceed the no conditional task size
for llap.
This is follow up for HIVE-16582.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)