Configure Hive Directories

You can configure the following Hive directories:

  • Hive Scratch Directory
  • Hive Warehouse Directory
  • Hive Error Logs Directory

Hive Scratch Directory

In Hive 1.0, HPE Data Fabric configures the Hive scratch directory to be /user/<user.name>/tmp/hive/<user.name>. In Hive 0.13, the default scratch directory is /user/<user.name>/tmp/hive. The hive user must have write access to the /user folder.

To modify this parameter, perform one of the following operations:

  • Set this parameter in the hive-site.xml.Copy the hive.exec.scratchdir property elements from the $HIVE_HOME/conf/hive-default.xml.template file and paste them into an XML configuration element in the $HIVE_HOME/conf/hive-site.xml file. Then, modify the value elements for these directories in the hive-site.xml file.

  • Set this parameter from the Hive shell. Example:

    hive> set hive.exec.scratchdir=/myvolume/tmp
NOTE
You will see better performance when queries import data from a table that is in the same Data Fabric volume as Hive scratch directory.

How Hive Handles Scratch Directories on HPE Data Fabric

When a query requires Hive to query existing tables and create data for new tables, Hive uses the following workflow:

  1. Create the query scratch directory hive_<timestamp>_<randomnumber> under the Hive scratch directory.
  2. Create the following directories as subdirectories of the scratch directory:
    1. Final query output directory. This directory's name takes the form -ext-<number>.

    2. An output directory for each MapReduce application. These directories' names take the form -mr-<number>.

  3. Hive executes the tasks, including MapReduce applications and loading data to the query output directory.
  4. Hive loads the data from output directory into a table. By default, the table's directory is in the /user/hive/warehouse directory. You can configure this location with the hive.metastore.warehouse.dir parameter in hive-site.xml, unless the table DDL specifies a custom location. Hive renames the output directory to the table directory in order to load the output data to the table.
  5. The scratch directories are automatically deleted after the query completes successfully.

HPE Data Fabric uses Administering Volumes, which are logical units that enable you to apply policies to a set of files, directories, and sub-volumes. When the output directory and the table directory are in different volumes, this workflow involves moving data across volumes. Moving data across volumes is slower than moving data within a volume. Therefore, HPE Data Fabric sets hive.optimize.insert.dest.volume to true to automatically create a scratch directory in the same volume as the target table.

Hive Warehouse Directory

Hive tables are stored in the Hive warehouse directory. By default, HPE Data Fabric configures the Hive warehouse directory to be /user/hive/warehouse under the root volume. This default is defined in the $HIVE_HOME/conf/hive-default.xml.template file.

To modify this parameter, perform one of the following operations:

  • Set this parameter in the hive-site.xml.Copy the hive.metastore.warehouse.dir property elements from the $HIVE_HOME/conf/hive-default.xml.template file and paste them into an XML configuration element in the $HIVE_HOME/conf/hive-site.xml file. Then, modify the value elements for these directories in the hive-site.xml file.
  • Set this parameter from the Hive shell. Example:
    hive> set hive.metastore.warehouse.dir=/myvolume/mydirectory
NOTE
You will see better performance when queries move data between tables in the same volume.

Hive Error Logs Directory

The log files are stored in /opt/mapr/hive/hive-<version>/logs/<user> by default.

To modify the log location:

  1. Configure hive.log.dir in $HIVE_HOME/conf/hive-log4j.properties file. Example: 
    hive.log.dir=<other_location>
  2. Set the sticky bit on the new directory. Example: 
    chmod 1777 <other_location>