What's new in Cloudera Data Services on premises 1.5.5 SP4
This Cloudera Data Services on premises release provides you with several new capabilities.
Support for Multi-Instance GPU nodes
The partitioning option for NVIDIA Multi-Instance GPUs is now supported. For better utilization, you can partition your GPUs into smaller slices. For more information, see About Multi-Instance GPU.
Support for multi-hop Longhorn upgrade option
You can enable multi-hop Longhorn upgrades to allow direct Cloudera Embedded Container Service upgrades starting from Cloudera Data Services on premises SP2 to Cloudera Data Services on premises SP4, bypassing intermediate releases. For more information, see Enabling multi-hop Longhorn upgrade.
High Availability (HA) Embedded Database cluster
The HA embedded Database cluster minimizes impact of services dependent on the HA enabled embedded database. HA Embedded Database cluster detects the unavailability of the primary node that renders the primary node being unavailable and automatically selects a new primary. For more information, see Managing High Availability Embedded Database cluster.
Support for configurable UID/GID ranges
Users can specify the UID and Cloudera Control Plane system supports the UID by enabling the containers to run those UID as mentioned. The UID is configurable for both Cloudera Embedded Container Service and OpenShift Container Platform. Supports both Installation and Upgrade processes. For more information refer to the note section under Embedded Container Service installation requirements, Embedded Container Service upgrade requirements, OpenShift Container Service installation requirements, and OpenShift Container Service upgrade requirements.
Setting up single / multi-rail cluster for Multi-Instance GPU (MIG) nodes
Setting up the HGX cluster involves multiple operational phases that includes discovering rails, defining YAML configuration, applying the changes, and verifying configurations. For more information, see Setting up Single or Multi-Rail cluster.
