We are using Hive version 0.10.0+198 and Hive Server version 1.cdh4.4.0.p0.15.el6, according to our RedHat packages.
Consecutive Hive queries build up the number of connections over time. For every query we execute, the command
/usr/sbin/lsof -p _hiveserverpid | grep -i tcp
grows by two lines at the beginning of a query, then shrinks by one sometime during or after the map portion of the query. Over time this number grows to 65 lines, at which point the Hive server presumably runs out of connections and gives up.
This seems likely to be module-specific behavior, although I am unsure of this. I'm willing to do a lot to fix this, or to correct my use of the module to avoid the issue.
We are using Hive version 0.10.0+198 and Hive Server version 1.cdh4.4.0.p0.15.el6, according to our RedHat packages.
Consecutive Hive queries build up the number of connections over time. For every query we execute, the command
/usr/sbin/lsof -p _hiveserverpid | grep -i tcp
grows by two lines at the beginning of a query, then shrinks by one sometime during or after the map portion of the query. Over time this number grows to 65 lines, at which point the Hive server presumably runs out of connections and gives up.
This seems likely to be module-specific behavior, although I am unsure of this. I'm willing to do a lot to fix this, or to correct my use of the module to avoid the issue.