@caisch The temp tables are created during the application run as intermediate data. These intermediate tables will not be removed in case the application fails and cleanup does not happen. The other reason might be if you are using beeline to run the query and if you abruptly disconnect the session without disconnecting properly by using '!q' then the file created on the '/tmp/hive' during the beeline initialisation will not be cleared. To clean up the /tmp directory automatically add the below properties in custom-hive-site.xml hive.start.cleanup.scratchdir - True // To clean up the Hive scratch directory while starting the HiveServer2. hive.server2.clear.dangling.scratchdir - true //This will start a thread in Hiveserver2 to clear out the dangling directories from the HDFS location. hive.server2.clear.dangling.scratchdir.interval - 1800s After adding the property kindly restart the hive service.  Reference link: https://cwiki.apache.org/confluence/display/Hive/Setting+Up+HiveServer2#SettingUpHiveServer2-ScratchDirectoryManagement    or you can run a cron job to delete the files periodically. Reference Link:  https://community.cloudera.com/t5/Support-Questions/Do-we-have-any-script-which-we-can-use-to-clean-tmp-hive-dir/m-p/156965    Please 'Accept as Solution' if my answers are really helpful to you. Thanks! ... View more