Cloudera Enterprise 5.15.x | Other versions

Upgrading from an Earlier CDH 5 Release to the Latest Release

  Important:
  • If you are using Cloudera Manager to manage CDH, do not use the instructions in this section. Follow the directions in Upgrading CDH Using Cloudera Manager to upgrade to the latest version of CDH 5 in a Cloudera Manager deployment.
  • MRv1 and YARN share a common set of configuration files, so it is safe to configure both of them. Cloudera does not recommend running MapReduce MRv1 and YARN daemons on the same hosts at the same time. If you want to easily switch between MapReduce MRv1 and YARN, use Cloudera Manager to manage these services.
  Note: Running Services

Use the service command to start, stop, and restart CDH components, instead of running scripts in /etc/init.d directly. The service command creates a predictable environment by setting the current working directory to / and removing most environment variables (passing only LANG and TERM). With /etc/init.d, existing environment variables remain in force and can produce unpredictable results. When you install CDH from packages, service is installed as part of the Linux Standard Base (LSB).

Important Tasks

  • Upgrading from any release earlier than CDH 5.4.0 to CDH 5.4.0 or later requires an HDFS metadata upgrade.
  • Upgrading from a release earlier than 5.2.0 requires all of the following:
    • Upgrade HDFS metadata
    • Upgrade the Sentry database
    • Upgrade the Hive database
    • Upgrade the Sqoop 2 database
Make sure you also do the following tasks that are required for every upgrade:
  • Upgrade the Oozie database and shared library.
  • If you have uploaded the Spark assembly JAR file to HDFS, upload the new version of the file.

Each of these tasks is described in context as you proceed through the upgrade. The following sections provide information and instructions:

Page generated May 18, 2018.