16.3.0 Patch Releases

Profiling jobs with regex-based detection rules no longer get stuck on columns containing large text values. The maximum input length for regex evaluation can be limited using the MAX_REGEX_INPUT_LENGTH DPE property.

Lineage Scanning Service no longer fails to start in self-managed deployments due to a missing edge management configuration.

Lineage Scanning Service now starts correctly after upgrading to 16.3.0-patch24.

Data load no longer fails with IllegalStateException: Queue full during party matching when a large number of matched pairs are produced.

Fixed incorrect matching results when using multiple matching rules with filters in a key rule.

Source detail screen now loads correctly when reopening a previously viewed source.

Job submission no longer occasionally fails with a "Job already exists" exception.

Job submission no longer fails with an "Engine not available" error when DPE is temporarily reconnecting.

ONE Data tables open correctly without TypeError error notifications.

Saving Data Observability configuration no longer times out or takes over a minute in environments with large data volumes.

Lineage diagram edges now display correctly after refreshing the page with All selected for the datastores field.

Faster DPM resource allocation queries, reducing database load in high-throughput environments.

Large profiling plans are now stored via object storage instead of the database, and job data is cleaned up after completion, reducing database load and storage usage.

Improved security of AI service configuration.

In Ataccama Cloud environments, restarting the DQ Issue Tracker application from the Admin Center now correctly restarts the service. Previously, the restart did not fully take effect in containerized deployments.

Lineage scans now correctly handle private keys with newline characters.

Improved AI Evolution service performance.

Fixed an issue where ADLS mounting would fail when using hybrid DPE with Spark pushdown and Azure Key Vault credentials.

Faster loading times for Source and Catalog screens through optimized metadata fetching.

Faster catalog item metadata imports through reduced database overhead.

Databricks lineage assets now correctly link to catalog items instead of appearing as generic database assets.

Fixed an upgrade issue where certain roles could end up with duplicate access levels, preventing modifications.

Improved hybrid DPE stability during high memory usage.

Optimized system validations to reduce resource usage on large environments.

Reduced log noise from analytics-related network errors.

Power BI reports with errors now show "No preview available" in ONE instead of a TypeError.

Audit records in audit_affected_asset table now show only the unique user ID in the user field, no longer including user roles.

Data observability overview metrics now load on demand, improving page load times for large sources.

Fixed an issue where MS SQL lineage scans could fail when logging certain parsing errors.

Attribute IDs on SQL catalog items imported from content packs are now preserved when editing the query.

Fixed 'user is null' messages flooding logs on some environments.

PostgreSQL application_name connection parameter in logs no longer contains correlation and transaction IDs, resulting in more useful query statistics.

Security fixes for Lineage Scanning service: CVE-2025-48976, CVE-2025-55163.

Set the default logging level for hybrid DPEs to DEBUG.

When importing lineage from a ZIP file, you can now import up to 50 connections at once, preventing incomplete diagram imports.

MS SQL scan plans no longer fail due to datetime parsing issues. If parsing fails, the current time is used instead (or null if allowed) and a warning is logged.

Scan plans screen now refreshes automatically when updates are available.

Null values in lineage data no longer cause parsing errors when loading diagrams.

Updated documentation link to the correct version.

Improved MDM server startup time through database optimization.

Filtering works as expected on fields using WINDOW lookup type.

Usernames are correctly displayed in MDM Web App Admin Center instead of appearing hashed.

Removed an incorrect ssh:// prefix from the GIT_CLIENT_URL field, which was preventing SSH connections to repositories in Ataccama Cloud environments.

Upgraded Tomcat to v10.1.50 for improved security.

Monitoring project results in email notifications are now rounded to two decimal places.

Loading to ONE Data works correctly for data with null or empty values of DateTime data type.

Loading to ONE Data now works correctly when DateTime values are null or empty.

Databricks lineage is now correctly labeled and mapped to catalog items, preserving the lineage hierarchy after import.

When importing Databricks or Dremio lineage, connections are correctly preselected in the import mapping settings.

When providing a private PEM key for a scan plan, newlines are no longer replaced by spaces for values spanning multiple lines.

Transition Name in custom workflow configuration must now be unique.

MDM server no longer fails on startup with WorkflowServerComponent not found error when WorkflowTaskListener is configured in nme-executor.xml.

Fixed database connection leaks in event handlers.

Improved support for long-running requests in Ataccama Cloud, preventing large data imports from timing out.

Fixed connection leaks in RDM that were causing sessions to remain idle, potentially leading to lock or wait issues.

Improvements to SAP RFC connections:

Notification handler now resubscribes automatically after five minutes of inactivity, preventing high-traffic environments from getting stuck.

In Data Observability, attributes are no longer individually logged when calculating domain statistics, improving platform performance.

Changes in terms applied to schema views are now reflected in search immediately.

In Ataccama Cloud environments, uploading large lineage scan results into ONE no longer gets stuck or fails.

ONE Metadata Reader step correctly reads properties of Referenced Object Array type.

Collibra Writer step now properly groups relations to prevent overwrites in REPLACE mode and allows specifying Source or Target asset as the relationship identifier.

Added a JSON Assigner step for creating and modifying JSON data, outputting valid JSON payloads as STRING or JSON type.

After upgrading to 16.3.0, Sweep documentation flow now appears in the documentation flow menu.

Rule suggestions load correctly in monitoring projects for catalog items with assigned rules or terms you don’t have access to.

Term Suggestions services no longer overload application logs with zero-confidence term suggestions.

Term Suggestions metrics are no longer collected when the services are not configured.

BigQuery pushdown processing now displays the correct source name in error messages.

Improved monitoring project notification logging for faster troubleshooting.

Upgraded OpenSearch for enhanced security.

Updated Kafka client library in Kafka Provider Component to prevent compatibility errors.

Save option now appears alongside Publish and Discard when creating or updating records with direct publishing configured.

Rule suggestions now correctly respect your access permissions when using the Check for rule suggestions option. You’ll see suggestions for catalog items you have access to (even without term access), while rules requiring higher permissions are excluded.

Improved error messaging when creating an ADLS Gen 2 connection, making it easier to troubleshoot connection failures.

Power BI reports are now correctly matched to reports in Data Catalog instead of datasets.

Scan plans executed on an edge instance are no longer shown as Finished if artifacts failed to upload.

When running MS SQL scan plans, empty statements are removed during preprocessing.

When running SQL scan plans, synonyms are now correctly recognized and processed for assets already in Data Catalog.

When a connection name is referenced through a function in a scan plan, a fallback name is now used instead of causing the scan to fail.

Assets with long concatenated names now appear correctly on the lineage diagram after import.

Lineage imports can now be canceled regardless of their status, including stuck or failed imports.

MDM Web App widgets no longer show a blank screen on load.

Resolved memory issues in the Matching step with parallelism configured that could cause the application to crash.

Email notification links now work correctly in Ataccama Cloud environments. Previously, the $detail_href$ variable was incorrectly resolved, causing HTTP 401 Unauthorized errors when opening links.

When reimporting the same monitoring project configuration but with a post-processing plan attached, related catalog items are no longer duplicated and the project runs as expected.

When importing monitoring project configuration containing filters, filter references to catalog items and attributes are correctly updated after remapping. Previously, filters would appear invisible after remapping but were still submitted to DQ evaluation as duplicates, causing exponential growth in filter combinations and processing failures.

Data quality rules containing regular expressions now produce consistent results between rule debugging and DQ evaluation.

Pausing an Ataccama Cloud environment running ONE Data with PrivateLink configured works as expected.

Updating individual data quality checks using the Update rule option in the DQ check modal works as expected. Previously, the update would fail to be propagated.

Editing an asset as a non-admin user now correctly creates a review task instead of returning a GraphQlTransport error.

AI-generated transcription of the lineage transformation context is correctly cleared after opening a different flow.

Resolved the issue with failing health checks for system validations when upgrading to 16.3.0-patch5.

Updated MS SQL JDBC driver to 13.2.1.

Infrastructure improvements to Ataccama Cloud environments.

Increased the default client_max_body_size in Ataccama Cloud environments, enabling larger bulk record imports in RDM without application timeouts.

Plans using Lookup step no longer fail with a NoSuchMethodError message.

Profile inspector no longer shows mixed results when monitoring projects target the same catalog item with and without data slices. Results now correctly reflect whether the project uses the full dataset or a specific data slice.

Searching to add a data steward doesn’t remove all previously assigned stewards. New stewards are now added without affecting existing assignments.

Databricks pushdown processing works as expected when operational credentials are used.

Keycloak now starts successfully when using a custom monitoring port.

Users with MDM_Viewer role correctly see only attributes of cmo_type when querying the MDM REST API.

During publishing, the streaming event handler now correctly populates the meta_origin column as a string instead of incorrectly treating it as a number.

In RDM Server runtime configuration (server.runtimeConfig), predefined connections now use OpenID authentication instead of Basic authentication to enforce 2FA-only access to the application. Predefined connections can be overridden by custom connections of the same name if needed.

IgnoredComparisonColumns ignores columns only on the selected entities.

Batch load and workflow operations no longer fail when the parallel write strategy is enabled and matching is added to an entity after initial data load.

Workflows no longer fail when runtime configuration contains empty passwords or other secrets.

Selecting Details in the Orchestration section of MDM Admin Center expands rows correctly.

MDM server logs events related to setup of PostgreSQL database.

MDM server logs events related to Lookup items and VFS configuration.

Security fix (CVE-2025-59250).

Security fix (CVE-2025-59250).

Long term names do not overflow the Relationships widget.

When using Snowflake OAuth credentials, the token validity is derived from the OAuth response.

On the lineage diagram, the transformation context panel now stays closed when dismissed and won’t show outdated content from previous diagrams.

Using the Start lineage from here option on attribute-level lineage correctly shows all edge assets.

Virtual file system is used as expected in MDM cleansing plans. Previously, lookup files in these plans were incorrectly loaded from a temporary local folder instead and the VFS was ignored, leading to recent unsafe memory access operation error messages.

Improved error messages for missing autocorrect configuration when using enrichments.

When rejecting a matching proposal with the Exclude this decision from AI matching learning option selected, the web application now clearly reports an error if there are no records related to the proposal. If the linked entity exists, the proposal is rejected without issues.

MDM Web App Admin Center correctly reports on the event handler execution times (started and finished), allowing you to more accurately assess the handler status.

In HA setups, scheduled tasks use a dedicated thread pool, allowing the second processing node to remain stable.

The default workflow and scheduler ID in MDM Server components is set to MDM.

Streamlined the web application upgrade template in ONE Desktop: redundant workflow step checkboxes were removed and group names are automatically converted to valid technical formats.

Improved table loading performance for non-admin users in environments with extensive role configurations.

Added confirmation prompt (type "delete") to prevent accidental deletion of the active RDM configuration in the Admin Center.

In Ataccama Cloud environments, some properties (such as spring.mail.*) from the application.properties configuration file can be successfully overridden using extraProperties key in Helm charts.

ONE Desktop job results are no longer retained in the Executor Bucket (MinIO). The Delete MinIO Job result cleanup option is applied by default.

Unity Catalog column and data type mapping works correctly with Map to Query functionality in Databricks.

ONE Desktop successfully connects to MinIO instances behind a proxy.

SQL Commands are available in the table context menu.

Metadata Reader step correctly filters embedded object array data by specific properties instead of returning all values.

Metadata Reader step correctly retrieves projectStats data for monitoring projects.

CSV viewer displays a message when encountering lines exceeding the maximum record length (65,536 characters) instead of silently skipping or truncating them.