Major features and updates for the Cloudera Machine Learning experience.
October 27, 2021
New Features / Improvements
- ML Flow - Cloudera Machine Learning now supports experiment tracking using open MLflow standards. For more information, see CML Experiment Tracking through MLflow (Preview) in Preview Features.
- Spark 3 - Users can now use Spark 3 in Sessions, Jobs, Models or Applications in their projects configured to use ML Runtimes.
- cdswctl - Users can now use the cdswctl CLI client to create Sessions with Spark.
- Root volume size - On existing CML workspaces based on AWS, there is a known issue where the provisioned root volume capacity is not enough to accomodate CML deployments with K8s versions >= 1.21. To mitigate this issue, when upgrading existing workspaces with a K8s version older than 1.19 on AWS, the root volume size is automatically increased/updated to 128 GB. Please note that this is not an issue on Azure where the pre-provisioned capacity for workspaces is already 128GB
- Workbench Editor - The Workbench now supports wrapping lines in the editor window. Users can enable this by selecting the Line Wrapping Enabled option under the View menu item in the Workbench.
- Private AKS clusters - Support for Private AKS clusters is now available in Preview mode. With Private AKS cluster support, CML workspaces are provisioned with a Private K8s API server making the deployments more secure.
- Pre-flight checks - NFS Server pre-flight checks have been added for Azure where
validations are done to make sure that the specified BYONFS server directory meet the
following criteria NFS server directory specified should be:
- The directory must be owned by CML user 8536.
- The directory must be empty.
- Atlas (DSE-16706) - Fixed an issue with reporting data lineage of CML model deployments to Atlas.