I seem to have thousands of empty pipeout files left behind under /tmp. They are all under /tmp/hive. What are they, do I need to worry about them?'

[#1]

hiveserver2 create operational logs which are added part of https://issues.apache.org/jira/browse/HIVE-4629 to report the progress of query like below.

"Parsing command", "Parse Completed", "Starting Semantic Analysis", "Semantic Analysis Completed", "Starting command"

these pipe file are created to write intermediate result into it to report the query progress,you need not to worry about them, if you want to disable them there is an option in hive to disable them by using

• hive.server2.logging.operation.enabled

[#2 · Accepted solution]

I've seen this before - we created a /tmp/hive/<id>.pipeout file for every JDBC connection, but if memory serves we failed to clean up those where no operation had been done on that JDBC connection.

I believe it was fixed in the latest versions of HDP (>2.5.0) so if you're running an older version you could be running into this: BUG-46108 is what I am referring to