7.4. Downgrade the YARN ResourceManager process or HA process pair

This is the start of several steps related to the YARN downgrade process. All running jobs should continue to make progress during and after you downgrade JobHistoryServer, YARN Timeline Service, ResourceMaster, and NodeManager processes, without task retries or failures due to the downgrade.

MapReduce applications will continue to use the same MapReduce version during the downgrade.

If High Availability is not enabled for the YARN ResourceManager, downgrade the single ResourceManager process as follows:

Shut down the ResourceManager process, switch to the previous version, and start the process:

su - yarn -c "/usr/hdp/current/hadoop-yarn-resourcemanager/sbin/yarn-daemon.sh stop resourcemanager"

hdp-select set hadoop-yarn-resourcemanager 2.2.0.0-2041

su - yarn -c "/usr/hdp/current/hadoop-yarn-resourcemanager/sbin/yarn-daemon.sh start resourcemanager"

If High Availability is enabled for the YARN ResourceManager you’ll need to downgrade both ResourceManager processes, starting with the standby process.

In the following instructions, RM1 refers to the currently active ResourceManager. RM2 refers to the current standby ResourceManager.

Continue with the next downgrade step.