Rolling Upgrade Guide
Also available as:
PDF
loading table of contents...

Chapter 3. Upgrading the Cluster

In this stage, services will be switched to the new software version.

[Important]Important

Perform all upgrade steps in the exact sequence listed in this section, for components deployed in your cluster.

  1. Upgrade ZooKeeper

  2. Upgrade Ranger (if Ranger is enabled on your cluster)

  3. Upgrade core-cluster master processes, including HDFS JournalNodes and NameNode, YARN ResourceManager, and HBase Master(s)

  4. Upgrade core-cluster slave processes: HDFS DataNodes, YARN NodeManagers, and HBase RegionServers

  5. Upgrade non-core cluster components: Hive, Oozie, and Falcon

  6. Upgrade Knox

  7. Upgrade components outside of the cluster, such as Storm topologies that run on Storm standalone

  8. Upgrade Slider

  9. Upgrade clients on cluster (core and non-core) components, such as MapReduce, Tez, and Phoenix client

  10. Finalize the rolling upgrade process. Caution: rollback and rolling downgrade operations are not possible after this step!

After upgrading each component, you’ll validate that the component is running successfully.

[Note]Note

The hdp-select script requires root access. The instructions in this section assume that you start the upgrade process logged on as root.