Hadoop Preinstall - Parallel Upgrade Tool (PUT)

Hadoop Preinstall is the first part of the upgrade sequence, which has several functions. The first time it runs (immediately after the Package Designations step), PUT divides the packages listed on the Confirm Package Selections screen into groups that will be upgraded sequentially, one group at a time.

Before scripts are run on each package group, a pause screen appears. This screen displays the ID of the group it is currently running. Once users reach the first pause screen, their options are limited to aborting the operation or continuing the operation. For this step the back button is disabled, as returning to a previous plugin is not supported. On each subsequent pause screen after the first pause screen, users can proceed with the operation; the Back and Abort buttons are disabled.

The Package Rules text file is in the /opt/Teradata/TDput/data directory by default. If the user wants to use a customized rules file, it can be placed in the /var/opt/teradata/pkgs/HDP directory. This flexibility helps when an upgrade requires a different sequence or grouping of the RPMs that are being upgraded.

If a group has a pre-upgrade script, Hadoop Preinstall runs a job on one of the Hadoop master nodes to execute the appropriate script. Once the script has completed (or if there is no script to run), the list of .rpm files is built in preparation for the installation step. If an upgrade script fails, there is an option to retry. Retrying is only recommended if the problem that caused the script to fail can be resolved before retrying. Attempting to run the script multiple times could render the cluster unstable. The Hadoop Preinstall step runs multiple times, once for each package group. Typical Hadoop upgrades require four or five package groups to complete the upgrade of the cluster depending on whether the teradata-hadoop-tools package is included or upgraded ahead in advance.