Upgrading the Watson OpenScale service
Upgrading the Watson™
OpenScale service on
IBM Cloud Pak® for Data requires that a cluster
administrator must first upgrade the cluster environment for the service, and then a project
administrator can upgrade the service on IBM Cloud Pak for Data. If you have integrated services, such as
OpenPages® or IBM® Analytics Engine powered by Apache Spark, you must take additional steps to ensure that
these services work correctly with the new version of Watson
OpenScale.
Upgrade scenarios for Watson OpenScale with IBM Analytics Engine powered by Apache Spark
The following upgrade scenarios are supported for Watson OpenScale with Analytics Engine powered by Apache Spark. Depending on your specific upgrade path, you might need to complete tasks in Watson OpenScale to re-establish the connection with Analytics Engine powered by Apache Spark.
- Upgrade Watson OpenScale version 4.6 and Analytics Engine powered by Apache Spark version 4.6 to Watson OpenScale version 4.8 and Analytics Engine powered by Apache Spark version 4.8.
- You must rebuild the wos_env virtual environment and upload the archive. For more information on batch processing environments for Analytics Engine powered by Apache Spark, see Preparing the Watson OpenScale batch processing environment in IBM Analytics Engine powered by Apache Spark.
- Upgrade Watson OpenScale version 4.7 and Analytics Engine powered by Apache Spark version 4.7 to Watson OpenScale version 4.8 and Analytics Engine powered by Apache Spark version 4.8.
- You must rebuild the wos_env virtual environment and upload the archive. For more information on batch processing environments for Analytics Engine powered by Apache Spark, see Preparing the Watson OpenScale batch processing environment in IBM Analytics Engine powered by Apache Spark.
Known issues with upgrading Watson OpenScale and Watson Machine Learning
After upgrade, Watson Machine Learning runtime pods are hibernated and restart only after the first scoring request. Because of the time that it takes to restart the runtime pods, the initial request might timeout. The workaround is to reattempt the scoring evaluation.
Next steps
To start the upgrade process, choose one of the following tasks: