Major features and updates for the Cloudera Machine Learning service.
March 22, 2021
- CML Model DNS resolution - For CML on Azure and AWS, DataHub and CDE connectivity to CML services such as workspaces and model endpoints is now possible. This is supported by enabling internal DNS resolution for CML workspace endpoints and models on private IPs. However, any existing workspaces must be upgraded to the current release to obtain this functionality.
- RAZ integration - CML now integrates with RAZ for S3.
- Cdswctl support - ML Runtimes is now supported in cdswctl.
- External engine repositories - Admins can now add basic Docker credentials for external engine repositories.
- ML workspace provisioning - In case of validation errors, it is possible to skip preflight validation for ML workspace provisioning. See Provisioning ML Workspaces for more information.
- Cloud permissions - CML now supports configuring AWS environment cross-account roles with fine-grained cloud permissions and policies for provisioning and running workspaces. See AWS restricted policies for more information.
- DSE-14481 - CML no longer needs Network Security group ports to be open for EFS on AWS (port 2049).
- DSE-14685 - Clarified impact of changing Load Balancer Source Range settings on Security Group settings. See Provisioning ML Workspaces for details.