Homepage

Cloudera Data Services on premises 1.5.5 SP4 Release Summary

This release summary for Cloudera Data Services on premises 1.5.5 SP4 introduces key updates and functionality enhancements across the Cloudera Control Plane, Cloudera Data Warehouse, Cloudera Data Engineering, Cloudera AI, Cloudera Observability, and Cloudera Data Catalog.

Platform Cloudera Control Plane

  • Support for Multi-Instance GPU (MIG) nodes: Multi-Instance GPUs (MIG) from NVIDIA can now be partitioned into smaller slices to increase GPU resource utilization across workloads.
  • Setting up single / multi-rail cluster for Multi-Instance GPU (MIG) nodes: Cloudera Control Plane supports NVIDIA Hyperscale Graphics eXtension (HGX) clusters through structured operational phases to discover rails, define YAML configurations, apply updates, and verify setup.
  • Support for multi-hop Longhorn upgrade option: Direct Cloudera Embedded Container Service (ECS) upgrades are supported from Cloudera Data Services on premises SP2 to SP4, bypassing intermediate releases.
  • High Availability (HA) Embedded Database cluster: High availability is introduced for the embedded database to minimize service disruptions. The system automatically detects primary node unavailability and selects a new primary node.
  • Support for configurable UID/GID ranges: You can specify custom UID and GID ranges for container execution across both Cloudera Embedded Container Service (ECS) and Red Hat OpenShift Container Platform (OCP) during installation and upgrade processes.
  • Notifications: You can subscribe to receive real-time announcements and notifications about product updates and service-specific alerts directly within the Cloudera Management Console Notifications page or from the notification bell menu.
  • Enhanced user interface to manage Diagnostics data: Navigation workflows and user interface enhancements offer a streamlined experience for managing diagnostics data handling.
  • Data Recovery Service (DRS): The Data Recovery Service (DRS) provides a holistic backup and recovery solution for the core Cloudera Control Plane and integrated Data Services, supporting backup and recovery on a secondary cluster.

Cloudera Data Warehouse

  • Configuration change notifications integrated with Cloudera Management Console: Cloudera Data Warehouse integrates with the Cloudera Management Console notification service to send real-time alerts when Virtual Warehouse configuration changes occur, such as updates to Kerberos configurations or security certificates.
  • Integrate with Velero-based Disaster Recovery Service (DRS): Enhanced DRS capabilities support Velero-based backup and restore operations. You can back up Virtual Warehouses, Cloudera Data Visualization, and shared Cloudera Data Explorer (Hue) instances, and restore them onto a secondary ECS cluster connected to the same base cluster.
  • Configuring query-level fault tolerance and coordinator high availability for Trino: Supports fault tolerance with query-level retries (up to four times using in-memory buffer spooling) and Active-Passive coordinator High Availability (HA) for Trino Virtual Warehouses. You can configure both settings when creating or editing a Trino Virtual Warehouse in the web UI.
  • OpenTelemetry export support: The Enable OpenTelemetry export feature allows Hive and Impala Virtual Warehouses to route telemetry metrics and traces directly to the Cloudera Control Plane OpenTelemetry gateway.
  • Add specific stable labels to Cloudera Data Warehouse workload namespaces: Enclosing namespaces for Impala, Hive, and Trino Virtual Warehouses, Database Catalogs, and Hue instances are automatically tagged using the [cdw.cloudera.com/application](https://cdw.cloudera.com/application):<application_name> format to enable third-party integration and precise resource filtering.
  • Enrolling Cloudera Data Warehouse Virtual Warehouse Namespaces into Istio service mesh: Virtual Warehouse namespaces can be enrolled into an Istio service mesh to enforce mutual TLS (mTLS) security in Ambient mode by updating the istio-mesh-mode property in the Cloudera Data Warehouse ConfigMap.

Iceberg on Cloudera Data Warehouse

  • Trino Iceberg deletion vector support: Trino supports deletion vectors on Iceberg V3 tables using roaring bitmap deletion vectors in Puffin files, reducing small-file overhead and accelerating row-level delete queries.
  • Trino Iceberg row lineage support: Trino supports tracking row history on Iceberg V3 tables using the $row_id and $last_updated_sequence_number metadata columns.

Trino on Cloudera Data Warehouse

  • Trino runtime upgraded: Upgraded from version 479 to 481, providing performance optimizations and stability improvements.
  • Apache Kudu connector available in federation connectors: Trino supports querying, inserting, and deleting data directly in Apache Kudu using the Apache Kudu connector within Federation Connectors.
  • Connection pooling for Hive and Impala federation connectors: Supports JDBC connection pooling for Hive-jdbc and Impala federation connectors. Enable pooling by setting connection-pool.enabled to true in the connector configuration.
  • Support for the Trino Lakehouse connector: You can configure and manage Lakehouse connectors in the Trino UI with integrated Apache Ranger properties for simplified security management.
  • Kerberos authentication for the Trino Oracle connector: The Trino Oracle connector supports Kerberos authentication for connecting to Kerberos-secured Oracle databases.

Cloudera Data Engineering

  • In-place upgrade support: Supports In-Place Upgrades and Backup & Restore capabilities on premises, allowing seamless service upgrades within existing environments.
  • Notifications: You can subscribe to product announcements and service-specific alerts.
  • Data Recovery Service (DRS): Integrates with the new Data Recovery Service for holistic backup and recovery across the Cloudera Control Plane and Data Services.
  • Expanded data connector support for Kafka, HBase, and Phoenix: Data Connector framework support expands beyond Apache Ozone to include native connectivity for Apache Kafka, Apache HBase, and Apache Phoenix.
  • Support for Resuming and Re-running Airflow DAG Tasks: Administrators can resume or rerun failed DAG tasks using the Clear Task action in the Airflow UI without requiring full DAG execution.
  • Automated Airflow Log Management: Supports automated log retention and cleanup policies for task logs and run metadata, configurable during Virtual Cluster creation.
  • Export Spark event logs to third party observability tools: Enables exporting finalized Spark event logs to third-party observability tools through the OpenTelemetry (OTel) gateway.
  • Update TLS Certificates for Cloudera Data Engineering Repositories: Supports updating TLS certificates for existing repositories after creation.
  • Enhanced security compliance with single-purpose EKU certificates: Enforces strict Single-Purpose Extended Key Usage (EKU) for digital certificates (such as serverAuth or clientAuth).

Cloudera AI

Cloudera AI Workbench

  • NVIDIA GPUDirect Remote Direct Memory Access: Supports NVIDIA GPUDirect RDMA to enable applications to directly access remote GPU memory, reducing latency and kernel overhead.
  • Multi-instance GPU support in Cloudera AI Workbench: Enables partitioning physical GPUs into isolated virtual instances (MIG) to run concurrent model deployments on shared hardware.

Cloudera AI Registry

  • PostgreSQL upgrade: The underlying PostgreSQL database is upgraded to version 17 (PG17).

Cloudera AI Inference service

  • Persistent Volume usage for model storage in Cloudera AI Inference service: Supports dedicated persistent volumes for hosting large model artifacts, preventing repeated model downloads across workload restarts.
  • Multi-instance GPU support in Cloudera AI Inference service: Supports dynamic fractional GPU allocation (MIG) using the NVIDIA GPU Operator, allowing you to select predefined GPU memory profiles for model deployment.
  • Enhanced Replica monitoring for Model Endpoints: The Model Endpoint Details page header includes a real-time Replicas status breakdown for Total, Healthy, Pull Image, and Pending replicas.
  • Low-Rank Adaptation adapters with Cloudera AI Inference service: Supports serving Parameter-Efficient Fine-Tuning (PEFT) adapters, such as LoRA adapters, with NVIDIA NIMs and Hugging Face models to serve multiple fine-tuned model variants from a single endpoint.
  • Filter and discover certified models in Model Hub: Filter and identify optimized models in Model Hub by using workload certification badges, custom YAML labels, and advanced catalog filters across Cloudera AI tools.

ML Runtimes

  • ML Runtimes version 2026.08.1: Version 2026.08.1 is available for deployment.

Agent Studio

  • Standardized prompts for Large Language Model (LLM) validation in Agent Studio: Standardizes non-editable validation prompts for Test Model and Test Function Calling actions when testing registered LLMs.

Cloudera Observability

  • Telemetry export to third party for Cloudera Data Services: Supports exporting Cloudera Data Engineering Spark event logs as well as Cloudera Data Warehouse Hive metrics and Impala traces to external third-party observability destinations through the OpenTelemetry gateway.

Cloudera Data Catalog

  • Data Recovery Service (DRS) integration: Supports backup and restore operations for Cloudera Data Catalog as part of control-plane-wide backups on Cloudera Embedded Container Service (ECS).