Due to
HIVE-26283, it is noticed that some queries that
were running normally (with the default value of "-1") in the previous version of
Cloudera Runtime are now taking longer to complete or fail entirely (usually noticed in
large clusters).
As a workaround, you can revert the value of
hive.optimize.sort.dynamic.partition.threshold
to "-1" either at
the query session level or in hive-site.xml using the Cloudera
Manager Safety Valve.
For queries that ran in the previous version with the
value set to "0" or "1" (optimization enabled), you do not have to modify the default
value after upgrading to Cloudera Runtime 7.1.7 SP1 and the queries will continue to
run normally.