Rolling back a failed upgrade

Once you've determined why the upgrade failed, you can roll it back and restore your old cluster.

The rollback script removes all of the changes made by the upgrade script. It also restores your old cluster from backup, starts it, and performs a healthcheck.

Important: The rollback script removes the logs created by the upgrade script, so be sure to check them beforehand to determine why the upgrade failed.

When the script finishes running, your cluster will be back to its pre-upgrade state and running. You can then either rerun the upgrade process or restore your old cluster.

If you decide to rerun the upgrade process, you must rerun the merge script, as the merged version of bdd.conf wasn't included in the backup.

To roll back a failed upgrade:

If you upgraded from BDD 1.0, downgrade your Hadoop cluster to the version you were running before.
This is required because BDD 1.0 doesn't support CDH 5.5.2. For instructions, refer to the documentation for your Hadoop distribution.
On the Admin Server, open a command prompt and go to the upgrade source directory.
Run the rollback script:
```
./rollback.sh bdd.conf [--silent]
```
You can optionally include the --silent flag to avoid the confirmation step.
Confirm that you want to roll back the upgrade.
When prompted, enter the absolute path to your old cluster's backup TAR file.