Cloudera Docs

Enable work preserving recovery on NodeManager

Work preserving recovery is enabled by default for the ResourceManager in a Cloudera Manager managed cluster, but not for the NodeManager. Use Cloudera Manager to enable it on the NodeManager.

The default value for the recovery directory is /var/lib/hadoop-yarn/yarn-nm-recovery.

  1. In Cloudera Manager, select the YARN service.
  2. Click the Configuration tab.
  3. Search for yarn-site.xml.
  4. Find the NodeManager Advanced Configuration Snippet (Safety Valve) for yarn-site.xml.
  5. Select the NodeManager and click on the plus icon.
  6. Add the following:
    • Name: yarn.nodemanager.recovery.enabled
    • Value: true
  7. Search for NodeManager Recovery Directory.
  8. Enter the directory path in the NodeManager Recovery Directory field.
    For example: /var/lib/hadoop-yarn/yarn-nm-recovery
  9. Click Save Changes.