Validate that there are no broken symlinks to the configurations and that you are using the precise HDP version, from which you intend to upgrade. Be sure to check in the /etc/hadoop/conf directory.

Confirm that there is enough disk space in /usr/hdp, /tmp, and HDFS. The HDP 2.5.6 binaries require 6.5 GB of disk space. Do not underestimate your disk space requirements.

Clear the logs, especially if the logs are on a mount that is low on space and shared with the operating system.

Identify dead nodes and exclude or decommission them.

Remove any hosts in an abnormal state from the cluster. You can add them again after the upgrade.

Identify any special *-env.sh setting such as special bootstrap libraries. Document them in case you need to reapply them after the upgrade.

Identify corrupt blocks and either repair or make note of them.

Validate that you have access to each of the backing metastores used by HDP. These include, but are not limited to Ambari, Hive, Oozie, Ranger, and KMS. Each component has an upgrade process, some of which are fairly automatic. Ensure the connected user to these databases have sufficient privileges to make schema changes as required by the upgrade.

Confirm that Ambari recognizes all the components that are installed in the cluster. If not, redeploy the component by using Ambari.

Ambari will not proceed with the upgrade if any components are in “maintenance” mode. Take time now to get those services up, healthy and out of maintenance mode.