What's New in Apache Atlas

Cloudera Runtime 7.3.2:

New React-Based User Interface for Apache Atlas

Apache Atlas now features a redesigned React-based user interface that offers enhanced usability and streamlined metadata management. You can switch between the Classic and New UI experiences. The new interface introduces an improved search panel that automatically lists all available entity types, classifications, and glossary terms, with one-click access to relevant members. Enhanced filtering capabilities allow users to show empty service types, unused classifications, and toggle between category or term views in the glossary. Additionally, entities and classifications can now be displayed in a collapsed flat tree view for simplified navigation of complex metadata hierarchies. For more information, see Apache Atlas dashboard tour.

Apache Atlas component upgraded to 2.4.0

The Atlas runtime component is upgraded from 2.1.0 to 2.4.0. Several stability and correctness fixes are included from the upstream release for bugs, including user interface improvements for classification propagation settings.

Atlas Auto-Purging introduced

The automated entity auto-purging feature addresses potential performance and storage issues caused by the previous manual purge strategy. Atlas preserves metadata by only marking entities as deleted. This leads to query performance degradation and increased storage usage as soft-deleted entities accumulate. The soft-deleted items could be only manually deleted by using the PUT /admin/purge/ API call, however, this API call leaves behind the column lineage entities for soft-deleted process entities.

The new, cron-based system can be configured to clean up obsolete process entities, including their column lineage entities, that are no longer relevant. This prevents sparse graphs and significantly improves metadata hygiene and query performance. For more information, see Atlas Auto-Purging overview.

Support replication of Atlas data from on-prem to on-prem

Atlas now supports asynchronous import of metadata using Kafka. Previously, the only import mechanism was synchronous: the HTTP connection remained open until the entire import completed, causing timeouts for large datasets and making concurrent imports fragile. With asynchronous import, a client submits an import request that is immediately staged and queued as a Kafka message, and receives an import ID in response without waiting for processing to finish. Atlas processes the import in the background and persists the request state, including received time, processing start time, completion time, and outcome.

The following new REST API endpoints are available:

• POST /api/atlas/admin/async/import — submit an asynchronous import; returns immediately with an import ID
• GET /api/atlas/admin/async/import/status — list all async import statuses
• GET /api/atlas/admin/async/import/status/{importId} — get the status of a specific import
• DELETE /api/atlas/admin/async/import/{importId} — abort a specific queued import

Atlas upgraded to use JDK 17

Atlas now runs on Java 17, upgraded from Java 8. JDK 17 is a Long-Term Support (LTS) release that brings improved performance, enhanced security, and better long-term maintainability. Key benefits for Atlas users include:

• Improved garbage collection, resulting in lower latency and more efficient memory usage for metadata-intensive workloads.
• Stronger cryptographic algorithms reducing security vulnerabilities.
• Long-term support guaranteed until at least 2029, ensuring continued security patches.

Logback introduced as logging framework

Apache Atlas now uses Logback as its logging framework, replacing Log4j2. This change enhances security and simplifies log management. It also enables the user to add any new properties overriding existing properties.

• Simplified Configuration: Streamlined logging setup is introduced with native XML configuration instead of the .properties file.

  • Go to Cloudera Manager > Atlas Server XML Override to replace the complete configuration file.

• Configuration still remains the same, as file size and rotation.