Pre-Upgrade Steps for Hadoop and YARN

Complete the following steps before you upgrade Hadoop and YARN with or without the Installer.

About this task

Before release 6.2.0 and EEP 7.0.0, Hadoop and YARN services were part of the Data Fabric repository for core packages. Upgrading Hadoop was not possible because releases 5.2.x, 6.0.x, and 6.1.x all used the same Hadoop version (version 2.7.0). Beginning with release 6.2.0 and EEP 7.0.0, Hadoop and YARN services were removed from Data Fabric core and delivered as ecosystem components in the EEP(MEP) repository. For more information, see Installing Hadoop and YARN.

Delivering Hadoop and YARN services in an EEP makes it possible to upgrade the packages independently of the HPE Ezmeral Data Fabric. Hadoop Protocol Versions shows the currently supported Hadoop versions. The following table describes the supported Hadoop upgrades:
If your current Hadoop version is . . . . To upgrade, you must
2.7.4.0 or later Upgrade to an EEP that provides a newer Hadoop version.
IMPORTANT Rolling upgrades from Hadoop 2.x to Hadoop 3.x are not supported. Only offline upgrades to Hadoop 3.x are supported.
2.7.0 Upgrade to core 6.2.x and install Hadoop components from a EEP that is supported on core 6.2.x. See Upgrade Workflows (Release 6.1.x to 6.2).
To prepare to upgrade Hadoop 2.7.4.0 or later:
  1. Stop the following services if they are installed:
    maprcli node services -name resourcemanager -action stop -nodes <IP-address>
    maprcli node services -name nodemanager -action stop -nodes <IP-address>
    maprcli node services -name historyserver -action stop -nodes <IP-address>
    maprcli node services -name timelineserver -action stop -nodes <IP-address>
    

You do not need to back up configuration files manually. Hadoop configuration files are automatically backed up to this directory: /opt/mapr/hadoop/hadoop-<timestampversion>