Cloudera Docs

Prerequisites for CDSW to CML migration

Before migrating from Cloudera Data Science Workbench (CDSW) to Cloudera Machine Learning (CML) in Cloudera Data Platform Private Cloud, you must meet a number of prerequisites to succeed. A prerequisite for migration is the installation of Cloudera Machine Learning on your Cloudera Data Science Workbench base cluster.

The following table presents the supported migration version combinations for Cloudera Data Science Workbench and Cloudera Machine Learning:

Table 1. Supported migration versions for Cloudera Data Science Workbench and Cloudera Machine Learning
Supported CDSW versions Target CML PVC version

CDSW 1.10.0

CDSW 1.10.1

CDSW 1.10.2

1.5.1

CDSW 1.10.5

1.5.2

1.5.3

1.5.4

Migration from CDSW, configured with LDAP, SAML, or LOCAL authentication to Cloudera Machine Learning, is supported, but the automatic migration is supported only if CDSW is running with LDAP. The migration process does not automatically migrate your authentication configurations. Therefore, setting up LDAP in CDSW prior to migration is part of the migration procedure.

The migration does not migrate your CDSW endpoint connections. Therefore, post-migration instructions include setting up LDAP, endpoint connections, and DNS on Cloudera Machine Learning, as well as downloading CDSW-related Grafana dashboards, so you can upload them after migration to Cloudera Machine Learning.

  1. You must have a CDSW 1.10.0 or later version cluster in Cloudera Data Platform; otherwise, choose one of the following options:
    • If you have a CDSW installation in either CDH or HDP, migrate to Private Cloud 1.5.1 or later version, and then migrate CDSW to Cloudera Machine Learning.
    • If you have CDSW installation earlier than 1.10.0, upgrade to CDSW 1.10.0 or later versions.
  2. If you do not have LDAP set up in your CDSW cluster on Cloudera Data Platform, set up LDAP before pre-migration tasks. For guidelines on setting up LDAP, see Configuring External Authentication with LDAP and SAML.
    The migration process cannot succeed without authentication.
  3. Meet the Cloudera Machine Learning software requirements for Private Cloud, including storage, for installing Cloudera Machine Learning on Cloudera Data Platform Private Cloud 1.5.1 or later version. For Cloudera Machine Learning software requirements for Private Cloud, see CML software requirements for Private Cloud.
  4. Backup CDSW data. For details on how to backup CDSW data, see Backup and Disaster Recovery for Cloudera Data Science Workbench.
  5. In CDSW, export your Grafana dashboards. For details on how to export Grafana dashborads, see Export and import | Grafana documentation.
  6. Note the connections of endpoints in your CDSW cluster, note your custom settings.
    You need to use this information after migration to set up endpoints in your Private Cloud cluster.
  7. If you customized your DNS configuration, make notes your custom settings to be able to customize your DNS configuration after migration.
    If you did not customize your DNS configuration, the migration tool configures DNS in your Private Cloud cluster.
  8. Gather information about your LDAP configurations on CDSW.
    After migration, you must set up LDAP again on the Cloudera Machine Learning cluster. The LDAP configuration is not migrated.
  9. In CDSW, manually back up the custom DNS configuration for Kube-DNS, and then migrate your custom configuration to Cloudera Machine Learning.
    Cloudera Machine Learning uses the core-DNS, which is incompatible with the CDSW Kube-DNS.
  10. In Cloudera Manager, select install and upgrade to CDP Private Cloud 1.5.1 or later version using the Embedded Container Service on your CDSW cluster.
    Migration of your CDSW workloads to Cloudera Machine Learning on OpenShift is not supported.
  11. During the installation of Cloudera Data Platform Private Cloud Data Services using Embedded Container Service set up a network connection between CDSW and the Cloudera Data Platform Private Cloud cluster if you select Airgap.
  12. Enable those Cloudera Machine Learning features during installation that you were using in CDSW.
    For example, enable model metrics and monitoring.
    If you do not enable the same, or similar, Cloudera Machine Learning features during installation that you were using in CDSW, you will not be able to use the Cloudera Machine Learning features.