After downloading the latest stable Hive release, we copied it to the desired location and uncompressed the archive file. This created a directory, hive-<version>.

Similarly, as we previously defined HADOOP_HOME and added the bin directory within the installation to the path variable, we then did something similar with HIVE_HOME and its bin directory.

We then created two directories on HDFS that Hive requires and changed their attributes to make them group writeable. The /tmp directory is where Hive will, by default, write transient data created during query execution and will also place output data in this location. The /user/hive/warehouse directory is where Hive will store the data that is written into its tables.

After all this setup, we run the hive command and a successful installation will give output similar to the one mentioned above. Running the hive command with no arguments enters an interactive shell; the hive> prompt is analogous to the sql> or mysql> prompts familiar from relational database interactive tools.

We then exit the interactive shell by typing quit;. Note the trailing semicolon ;. HiveQL is, as mentioned, very similar to SQL and follows the convention that all commands must be terminated by a semicolon. Pressing Enter without a semicolon will allow commands to be continued on subsequent lines.