Zhihua Deng created HIVE-28456:
----------------------------------
Summary: ObjectStore updatePartitionColumnStatisticsInBatch can
cause connection starvation
Key: HIVE-28456
URL: https://issues.apache.org/jira/browse/HIVE-28456
Project: Hive
Issue Type: Bug
Reporter: Zhihua Deng
Assignee: Zhihua Deng
Since HIVE-26419, we have a secondary connection pool for schema generation,
and for value generation operations, the size of this pool is 2. However, based
on DataNucleus documentation on datanucleus.ConnectionFactory2, link:
[https://www.datanucleus.org/products/accessplatform_5_0/jdo/persistence.html]
the secondary pool also serves for nontransactional connections, which makes
the ObjectStore updatePartitionColumnStatisticsInBatch request the connection
from this pool, as it doesn't open a transaction explicitly. If there is a slow
on inserting or updating the column statistics, the pool will become
unavailable quickly(the pool reaches its maximum size), the ObjectStore cloud
see the "Connection is not available, request timed out" under such a situation.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
