Reviewing Runtime add-on compatibility

As an on premises Administrator, you must ensure that Runtime add-ons used on your site are compatible with the base cluster version. This involves disabling any incompatible versions that may be installed.

The Java version used in the Cloudera AI Workbench must match, or must be higher than, the Java version running on the base cluster. If the base cluster uses a higher Java version than the one provided by the Hadoop add‑on, follow the steps in Troubleshooting Custom Runtime Add‑ons to upgrade the Java version in their Workbench.

The following rules apply to Java compatibility:

Hadoop CLI 7.3.2.0 and higher Hadoop CLI versions mount Java 17 into workloads.
Hadoop CLI add‑on versions lower than 7.3.2.0 mount Java 8.
Hardened Runtime images always use Java 17, regardless of the Hadoop CLI version in use.

To check and update the compatibility of ML Runtime add-ons, follow the instructions:

Go to the Cloudera Management Console, and determine the base cluster version.
In each Cloudera AI Workbench, in Site Administration > Runtime:
- Set the Hadoop add-on as default that has the base cluster version in its name.
- Keep those Spark add-ons enabled that have the base cluster version in their name and disable other Spark add-ons.
- If some workloads have been configured to use a disabled Spark add-on, the affected workloads must be reconfigured to use an enabled Spark add-on. This can happen in the event a workbench is upgraded.
Use the applicable Java version for the Cloudera AI Workbench.