Imports to Hive should use a nonce tmpdir for the HDFS import step.

Sqoop currently uses a two-tier model for importing to Hive. First it gets results into HDFS in some directory. Then it creates a Hive table and runs a LOAD DATA command to pull the data into the Hive warehouse subdir.

For a table-based import, this has the unfortunate side effect of leaving behind an empty directory named after the table in your hdfs home dir. For query-based imports, we can't infer a destination directory name based on a table, so you have to specify one with --target-dir. If you import into Hive, it'll then leave this directory empty as well.

Sqoop should recognize that when you've added --hive-import, it will be utilizing a temp directory, so it should pick a nonce directory target and then remove it when the hive import succeeds