> For the complete documentation index, see [llms.txt](https://docs.collibra.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.collibra.com/data-quality-self-hosted/release-notes.md).
# Release notes
{% hint style="warning" %}
* **Failure to upgrade to the most recent release of the Collibra Service and/or Software may adversely impact the security, reliability, availability, integrity, performance or support (including Collibra’s ability to meet its service levels) of the Service and/or Software.** For more information, read our Collibra supported versions policy.
* Collibra provides most configuration parameters as `--set` flags or a custom **values.yaml** override file. If you have modified any Helm chart templates directly, those modifications are not preserved during upgrades and must be reapplied manually. Before upgrading, consider comparing your current chart files against the new release chart files to identify any customizations you need to reapply. You can use the `helm diff upgrade` command or any file comparison tool.
* Some items included in this release may require an additional cost. Please contact your Collibra representative or Collibra Account Team with any questions.
* For information about Data Quality & Observability for Collibra Platform, go to [About Data Quality & Observability](https://productresources.collibra.com/docs/collibra/latest/Content/UnifiedDataQuality/co_about-data-quality.htm).
{% endhint %}
{% tabs %}
{% tab title="2026.09 (latest)" %}
## Release 2026.09
* Release date of Data Quality & Observability (self-hosted) 2026.09: October 13, 2026
* Release notes publication date: October 6, 2026
## New and improved
### Platform
* You can now generate a support bundle utility for Data Quality & Observability (self-hosted) that consolidates all relevant troubleshooting information into a single compressed file. This utility is available across all deployment types including Kubernetes, standalone environments, and Spark clusters, enabling you to share comprehensive diagnostic data with support teams through Zendesk more efficiently.
### Admin configuration
* You can now configure the system-wide maximum number of concurrent web pushdown jobs per deployment using the environment variable `DQ_PUSHDOWN_MAX_CONCURRENT_JOBS`. This allows you to optimize pushdown throughput for your infrastructure capacity while maintaining the existing per-tenant concurrency limits.
* You now see clear error messages when your organization's AI feature usage reaches its Collibra Unit (CU) limits. At 100% of your base allowance, a notification informs you that performance is rate-limited. At 120% of your total allowance, AI features are disabled, and an urgent message explains the next steps.
### Security
* Helm deployments that configure a custom Java truststore for the DQ Web application now store the truststore password securely in a Kubernetes Secret. This replaces the previous method of exposing the password in cleartext through the ConfigMap or command line arguments. The upgrade applies this security improvement automatically, so no configuration changes are required.
* You can now configure the Referrer-Policy in response headers for enhanced security through ConfigMap or `owl-env.sh` settings. The default behavior maintains the current implementation, which uses the browser's default referrer policy when not explicitly set.
* You now benefit from enhanced password encryption in Data Quality & Observability (self-hosted), aligning with updated security compliance standards. This improvement requires no action from you and does not impact your current workflows.
* Several Helm chart values that were ignored before now take effect, and the chart no longer passes four unused SAML settings. Before you upgrade, remove `global.persistence.metastore.mount.data.storageClassName` if you set it, and move `global.cloudStorage.HADOOP_CONF_DIR` to `global.cloudStorage.hdfs.HADOOP_CONF_DIR`. Spark History Server with HDFS log storage also deploys again without the old value set.
### Connections
* The column type for the Results column in the DQ\_COLLIBRA\_\* tables has been updated from TEXT to NCLOB to align with SAP HANA Cloud's deprecation of the TEXT type. This change ensures compatibility with the latest SAP HANA Cloud standards.
### Jobs
* You can now archive rule break records when running Data Quality jobs for Dremio connections, enabling more efficient validation of your data quality through enhanced rule result tracking.
* You can now better identify root causes for Data Quality jobs stuck in the "SETUP" state. Enhanced error messaging provides clearer insights into issues such as memory settings, image mismatches, and staged states like "submitted" or "terminated." A 5-minute grace period ensures jobs with large memory requests display expected errors after a delay. Additionally, job statuses are now accurate, preventing incorrect statuses like "FINISHED" due to disk full errors.
### Collibra Platform integration
* When you delete a Data Quality connection through the UI or API, all associated mappings are now automatically deleted. This enhancement eliminates the need for additional actions, ensuring a seamless deletion process. If you're using a release version prior to DQ 2026.09, you must manually delete mappings using the API endpoint `DELETE /dgc/integrations/mapping/connections?connectionName={connectionName}`. For versions DQ 2026.09 and later, mappings are deleted automatically.
## Fixes
### Platform
* Secure connections and encryption no longer intermittently fail in FIPS-enabled deployments on Kubernetes.
* The standalone installation script, `setup.sh`, now applies the `-owlpackage` option to control where the installer reads its files from. If the directory you specify is wrong, the installation stops with an error instead of reporting `install complete`.
* Web pods no longer fail to start when SAML configuration contains an incorrect application ID that prevents metadata retrieval from identity providers.
### Admin configuration
* The owlmanage.sh script now uses the updated `server.servlet.session.timeout` property instead of the outdated `server.session.timeout`. For Kubernetes deployments, the `SERVER_SERVLET_SESSION_TIMEOUT` property is now exposed in the Helm chart, allowing you to configure the session timeout.
### Connections
* Databricks connections that use Microsoft Entra ID service principal authentication now handle case-sensitive duplicate properties correctly when using the native Databricks JDBC driver 3.x, preventing connection validation failures that could occur with differently-cased property names.
* Parallel JDBC connections on SQL Server and Teradata pullup jobs now process complex SQL queries with multiple table joins and DISTINCT clauses without causing syntax errors. This enhancement ensures smoother query execution and improved reliability.
* You can now create jobs using Google Cloud Storage (GCS) remote file connections on standalone installations without encountering "Path traversal attempt detected" errors.
* You can now save a Databricks connection after correcting a previously invalid OAuth variable value, without needing to create a new variable.
### Jobs
* Data Quality jobs against BigQuery no longer fail with type-mismatch errors when dataset query has quoted table names and joins.
* In standalone FIPS deployments that use the default Spark submit mode, Check and Profile jobs no longer fail when generating data shape previews or recording rule breaks. You can use the new `DQ_FIPS_BC_JAR_PATH` and `DQ_FIPS_SPARK_EXTRA_JAVA_OPTIONS` environment variables to set a custom location for the BC-FIPS JAR or add custom JVM options for Spark jobs. If your Spark cluster runs on separate hosts from the DQ Agent, place the BC-FIPS JAR at the same path on the agent host and on every Spark node, and set DQ\_FIPS\_BC\_JAR\_PATH to that path.
* Parallel rule execution has been improved to ensure that a single rule with an invalid filter query no longer causes other rules in the same job to fail. Job logs now accurately report the failure reason, including the rule name and filter query, instead of showing a misleading "Timeout" label.
* Pullup jobs with break record archival configured to S3 connections no longer fail entirely when the S3 credentials are invalid or stale. Jobs now run to completion and display data quality results normally, with archival failures shown as rule-level exceptions in the Findings section.
* Pushdown jobs on certain data sources no longer report stack overflow errors when running rules and archiving break records. This improvement ensures that rules referencing a large number of columns now function correctly, and the driver accurately reports the number of break records inserted.
* The `archiveBreaks` value in a pullup job's JSON payload now accurately reflects the Archive Breaks setting after the job runs. Previously, the value always showed "false," even when the setting was enabled in the UI. This fix ensures the JSON payload aligns with the selected configuration, while break-record archival functionality remains unaffected.
### Findings
* Pulse View no longer shows a green "Complete" cell for a run that failed due to a breaking rule or anomaly when the row count itself did not fail.
* You can now download CSV files of rule breaks for datasets with a long history of runs, even when no explicit export limits are set, without encountering timeout errors. The download process has been optimized to prevent hanging when retrieving the total count of break records.
### Rules
* Data quality rule templates that reference the `dataset` or `@dataset` variable on SQL Server pushdown connections now run successfully without SQL syntax errors. These rule templates complete as expected during both manual runs and scheduled jobs. The syntax has been updated to ensure compatibility with SQL Server.
### Alerts
* Alert notifications now appear in the notification logs even when email delivery fails due to SMTP configuration errors.
### Dataset Manager
* You can now clone datasets and compile edited scope queries without encountering the "Error in query compilation - An error occurred" message. This issue, caused by a missing table parameter in API calls, has been resolved.
* You can now create query-based datasets with accurate source table detection. When you create or edit a Pullup job using custom SQL queries, the system correctly identifies source tables from complex queries including subqueries, CTEs, and joins, and automatically populates the database and table name fields in the catalog.
### Collibra Platform integration
* Auto-mapping jobs now use optimized database queries that reduce source system load and include configurable timeouts to prevent individual table failures from blocking entire jobs. You can now successfully cancel running auto-mapping jobs in multi-replica environments, and job status updates are more resilient to temporary connectivity issues. Additionally, failed table mappings no longer generate fabricated column mappings, ensuring data accuracy in your mapping results.
* Integration jobs now complete successfully when same-named rules are renamed or moved in the Collibra Platform. The synchronization flow correctly resolves and authorizes the target rule representation, preventing permission failures during attribute updates or deletions. You can now perform Data Quality synchronization jobs without encountering issues for affected rule assets.
* You can now scroll through long SQL queries in the SQL View dialog box when mapping datasets. A scroll bar appears when the query exceeds the visible editor height, allowing you to view the entire query.
### APIs
* The ROLE\_DATA\_EXPORT permission is now enforced on the /v2/get-rule-with-details API endpoint. Users without this role can no longer retrieve rule break details through this legacy endpoint, ensuring consistent data export controls across all API versions.
{% endtab %}
{% tab title="2026.08" %}
## Release 2026.08
* Release date of Data Quality & Observability (self-hosted) 2026.08: September 8, 2026
* Release notes publication date: August 21, 2026
## New and improved
### Platform
* Data Quality now supports Amazon EMR 8.x, which runs Spark 4.0.x.
### Admin configuration
* You can now configure the system-wide maximum number of concurrent CDQ web pushdown jobs per deployment using the environment variable \`DQ\_PUSHDOWN\_MAX\_CONCURRENT\_JOBS\`. This allows you to optimize pushdown throughput for your infrastructure capacity while maintaining existing per-tenant concurrency limits.
* Collibra admins now see Collibra Units (CU) usage banners in the consumption dashboard. Banners appear at 80%, 100%, and 120% of the purchased CU allowance, with increasing restrictions at each level and show organization-wide usage and its impact on CU-connected features. In-app banners for additional CU-connected features will follow in an upcoming release.
### Rules
* You can now preview Snowflake datasets that use key-pair-based authentication in the Rules Workbench, enabled through Livy support for the "Run result preview" feature.
* Rule query results are now saved in the source system when a link ID is specified. You can also enable a new option to use the rule results for preview and download, providing greater flexibility in managing rule outputs.
### Dataset Manager
* The maximum string length for values in SQL Server datasets has been increased from 30 to 100 characters, allowing for more comprehensive data handling in profiling and other activities.
### Jobs
* Pullup jobs against Oracle connections now use `TO_DATE` instead of `TO_TIMESTAMP` by default for run date values, improving performance by enabling index range scans instead of full table scans. If your date column requires microsecond or nanosecond precision, you can edit the query in the SQL Query editor to use `TO_TIMESTAMP`.
## Fixes
### Security
* We removed the default keystore and password from the installation to enhance security. To complete the installation, follow the steps in the Cloud Native Deployment documentation ([Create a Java keystore](/data-quality-self-hosted/security-configuration/setting-up-ssl-https.md#create-a-java-keystore) and [Create an SSL keystore secret](/data-quality-self-hosted/installation/installing-data-quality-observability-self-hosted-on-self-hosted-kubernetes/install-on-self-hosted-kubernetes.md#create-an-ssl-keystore-secret)) to create a custom keystore secret and update the corresponding password in the Helm charts.
### Connections
* You can now browse tables and schemas for CData JDBC connections in the Data Quality Explorer. The "Generate Report" function also works correctly for CData schemas without errors.
### Collibra Platform Integration
* Auto Map jobs no longer fail on schemas with more than 1 million table and column elements. Auto Map now retrieves only the tables you select for mapping, instead of querying the entire schema's asset tree.
* The schema drop-down in the Integration Mapping step now lists all available schemas when mapping a database with more than 1,000 schemas to a Data Quality connection. Previously, the drop-down was cut off at 1,000 records, causing some schemas to be missing.
* The Integration Mapping feature now displays accurate mapped and total counts for schemas and tables, resolving an issue where the mapped count incorrectly exceeded the total count.
### Rules
* Hive connections no longer generate parse errors when running rules previews on datasets that contain reserved words in column names.
* You can now access the Rules page without errors, even when duplicate rule history entries exist at the same timestamp. The most recent modification state is applied to the rule, ensuring consistent page loading.
* Column detection in multi-line SQL data quality rules now works correctly. The entire SQL rule is searched for referenced columns, ensuring no columns are missed. You may see additional columns discovered and associated with existing rules, especially when the SQL Pretty option is enabled. After upgrading, review the newly discovered columns and configure the primary column as needed.
### Jobs
* Job memory estimates no longer return a value of 0 in certain situations. A default value is now applied to the "parallelism factor" configuration key when no valid value is set, ensuring accurate memory estimates during job execution.
* You can now edit multiple remote file datasets in sequence without encountering the "Cannot read remote file system" error. Datasets from two remote file connections pointing to the same storage account but using different credentials, such as separate Service Principals, can now be edited without triggering this error. Each connection now uses its own Livy session, ensuring proper functionality across connections.
* Data Quality jobs no longer fail when you change a connection's host between an IP address and a DNS name. Collibra now correctly matches the job to its existing dataset, ensuring the job runs without interruption.
* Pushdown jobs using JDBC connections no longer cause the Data Quality service to run out of memory under heavy job loads. The service now handles a high volume of concurrent Pushdown jobs more efficiently, improving stability and eliminating the need for manual restarts.
* Netezza and other non-BigQuery database connections no longer fail when submitting Pullup jobs in Kubernetes deployments. The correct driver path is now identified without incorrectly appending "/core/" to the library location, ensuring successful job execution.
* You can now create Data Quality jobs for Google Cloud Storage (GCS) files in Kubernetes environments without encountering the "Cannot Read Remote File System" error. Previously, GCS credential files were not properly mounted to Spark pods, preventing job creation even when data preview worked correctly.
* Data Quality jobs on Dremio connections no longer fail with parse errors when profiling columns with unsupported data types. Unsupported data types are now automatically disabled during job creation, ensuring smoother job execution.
* Data Quality jobs with multiple outlier configurations now correctly preserve the **time bin** and **by** parameters during execution, ensuring accurate and consistent processing of outlier configurations.
* Profiling scores for completeness and consistency now display with consistent formatting and accurate breakdown percentages.
* Data Quality import job errors are now parsed and reported correctly even when a batch contains an error not tied to a specific asset. Previously, one such error caused the entire batch to fall back to the unparsed error format instead of the normal per-asset breakdown.
### APIs
* The `PUT /v2/updatecatalogobj` API now enforces proper role permissions, preventing users without the ROLE\_DATA\_GOVERNANCE\_MANAGER role from updating catalog metadata through direct API calls.
### Dataset Manager
* Additional access checks now ensure that only users with edit permissions can update dataset configurations. Users with read-only access can no longer edit scope queries in the Data Quality application. However, users with the "Data Preview" role and read-only access can now preview data without needing full dataset permissions.
* Dataset caching in Data Quality now correctly handles dataset names with capital letters, preventing unnecessary reprocessing when the same dataset is referenced multiple times across rules.
* You can now view columns correctly in Dataset Overview when working with Google BigQuery tables that have names containing hyphens, spaces, or leading digits without manually editing the SQL query.
### Findings
* You now see only the connections you have access to when selecting an archive location, as "Connection Security" filters the connections. However, "role\_admin" users bypass the "Connection Security" filter and see all connections in the explorer and archive drop-down list.
* The maximum string length for values in SQL Server has been increased from 30 to 100 characters, ensuring accurate profile data and resolving issues where unique counts and breaking record counts were incorrectly calculated due to string truncation.
* You can now view breaking records and preview breaks for rules in Pushdown jobs, even when the archive is written to a different database and schema than the source database. Additionally, you can now successfully download CSV files using the "Archive Break Records" option without encountering errors.
* You can now view archived break records in Data Quality Pullup jobs when using Azure Data Lake Storage Gen2 connections with Service Principal authentication that include subfolders in the connection path. Previously, clicking "Archived Breaks" resulted in a "File does not exist" error for these connection configurations.
* You can now export rule details with LinkIDs for Pushdown Data Quality jobs. Previously, the export returned empty results despite rule breaks existing for the run.
## Patch release
### 2026.08.1
* Data preview tables no longer display corrupted values when unencrypted dataset fields containing hex strings are mistakenly processed as encrypted data.
* You can now create, rename, and run Data Quality jobs as a non-administrator when "Connection security" is enabled, provided your role grants access to the connection and dataset. Previously, non-administrators received a generic error unless they had the ROLE\_ADMIN permission. Additionally, saving newly created jobs now works correctly for all users, including administrators, when "Dataset security" is enabled.
{% endtab %}
{% tab title="2026.07" %}
## Release 2026.07
* Expected release date of Data Quality & Observability (self-hosted) 2026.07: August 5, 2026
* Release notes publication date: July 31, 2026
## Announcement
{% hint style="warning" %}
**In this release (2026.07)**, Data Quality & Observability (self-hosted) is only available on Java 21. Depending on your installation of Data Quality & Observability (self-hosted), you can expect the following:
* **Kubernetes installations**
* Kubernetes containers automatically contain Java 21.
* If you use custom drivers, ensure they are compatible with Java 21.
* **Standalone installations**
* To install Data Quality & Observability (self-hosted) 2026.07, you must upgrade to Java 21.
* If you use custom drivers, ensure they are compatible with Java 21.
* Follow the latest [steps to upgrade](/data-quality-self-hosted/installation/upgrade-data-quality-observability-self-hosted/upgrade-collibra-data-quality-observability.md) to Data Quality & Observability (self-hosted) 2026.07 with Java 21.
* We encourage you to migrate to a Kubernetes installation to improve the scalability and ease of future maintenance.
The June release (2026.06) was the final release to contain Java 17 builds. Starting in this release (2026.07), only critical and high-priority bug fixes will be made for Java 17 versions of Data Quality & Observability (self-hosted) from 2026.06 and earlier.For a breakdown of Java availability in current and upcoming releases, click "See what is changing" below.
See what is changing
| Data Quality & Observability (self-hosted) version | Java 17 | Java 21 | Additional notes |
| -------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------- |
| 2026.06 and earlier | 
Yes | 
No | |
| 2026.07 and newer | No | Yes | Only critical and high-priority bug fixes will be made for Java 17 versions of Data Quality & Observability (self-hosted) from 2026.06 and earlier. |
{% endhint %}
{% hint style="warning" %}
**Dataproc (Managed Service for Apache Spark) Version 3.0 — Advisory** Dataproc 3.0 (Apache Spark 4.1) is not yet available or certified for production use. Customers may use it in non-production environments at their own risk, but we cannot provide full support until it is available.
* **Roadmap**: Official support for Amazon EMR 8 (Spark 4.x) is planned for the 2026.08 release.
* **Compatibility** Customers who want to test Dataproc 3.0 before it is available can use the 2026.08 DQ Spark cluster package, subject to the support limitations above.
* **Recommendation**: We recommend migrating to Cloud Native (Kubernetes) to avoid version compatibility issues in future releases, as we will continue upgrading Spark versions going forward.
{% endhint %}
{% hint style="warning" %}
**Breaking change: Replace owlmanage.sh when upgrading from 2026.06 to 2026.07** When upgrading from Data Quality & Observability (self-hosted) 2026.06 to 2026.07, the JVM will fail to initialize if the old `owlmanage.sh` is not replaced. To resolve the issue:
{% stepper %}
{% step %}
After extracting the DQ 2026.07 tar archive, replace the `owlmanage.sh` in the bin folder with the one from the new tar output before starting the service.
{% endstep %}
{% step %}
Extract the DQ 2026.07 tar archive to a staging directory.
{% endstep %}
{% step %}
Copy the new `owlmanage.sh` from the tar output to the bin/folder, overwriting the existing file.
{% endstep %}
{% step %}
Verify the custom class loader reference is no longer present in the file.
{% endstep %}
{% step %}
Start the DQ service as normal.
{% endstep %}
{% endstepper %}
**Note**: This is a one-time manual step required only during the upgrade. Fresh installations of DQ 2026.07 are not affected.
{% endhint %}
## New and improved
### Platform
* Auto map jobs now use the latest Collibra Platform IDs when determining table and column relationships, ensuring that mapping data remains current and accurate.
### Admin configuration
* You can now mask truststore passwords in Data Quality Agent configurations when using Spark on Kubernetes. Passwords no longer appear in clear text in job execution commands or logs, improving security for SSL configurations.
* When you manually edit a SQL query in the Explorer SQL editor, you now see a warning before compiling. The warning explains that manually edited queries are sent to the data source as-is and that identifiers must be properly escaped for the target database.
* Only users with explicit ROLE\_DATA\_EXPORT role assigned will be able to download archived break records. This includes the ADMIN user.
### Jobs
* You can now mask sensitive profile values in pullup jobs to ensure they are no longer stored in the Metastore in readable form. Enable the "Descriptive Statistics" setting for individual jobs in the job settings, or enforce it for all jobs at the system level. The administrator setting overrides the job-level setting.
### Security
* Security audit trail logs now include new events for rule templates: "Create new template," "Update existing template," and "Delete existing template." This enhancement improves traceability and accountability for rule template actions.
* We improved the security of the application to better protect your data.
### Connections
* You can now successfully compile dataset queries with capital letters in column aliases when using SAP HANA and PostgreSQL connections.
* You can now use AWS Athena JDBC driver version 3.8.0 with Data Quality & Observability (self-hosted), which provides improved compatibility and performance for your AWS data connections.
* You can now install and run Data Quality & Observability (self-hosted) in standalone mode on JDK 21 systems.
* Data Quality & Observability Classic (self-hosted) now uses PostgreSQL 18 as the default database, providing enhanced TLS 1.3 support for improved security compliance.
* You can now profile ".sas7bdat" files stored in Azure Blob Storage. These files are treated like other supported formats, such as CSV, Parquet, and JSON. Profiling outputs for ".sas7bdat" files are now accurate, and other file formats are unaffected.
### Rules
* You can now customize the column order in Freeform (SQLF) Template Rules by specifying columns in your SELECT clause, and the "Preview Breaks" dialog will display them in the exact sequence you defined rather than alphabetical order. This enhancement works with dynamic variables and maintains consistent column ordering across all supported data sources including Databricks, Redshift, Snowflake, and Oracle.
## Fixes
### Jobs
* You can now successfully download archive records from Azure Blob storage locations when using Pull Up connections in Data Quality jobs. The "File Download" button and signed links now work as expected without authorization errors.
* Snowflake schema and table names that start with digits are now automatically enclosed in double quotes in Data Quality job queries, eliminating SQL compilation errors and the need for manual corrections.
* The "Estimate" button in the Sizing tab within Explorer now works correctly for BigQuery connections when source queries include backticks around schema or column names with special characters. However, if a job query fails to compile, the job estimation will also fail.
### Dataset Manager
* Non-admin users now correctly receive queries with TOP N syntax when creating SQL Server datasets with Partial Scan enabled. Partial Scan queries for SQL Server datasets no longer use LIMIT syntax, preventing job failures when using pull-up connections.
* Time zone settings are now preserved during export and import processes when using APIs, ensuring consistent scheduling across workflows.
### Admin configuration
* When the security setting is enabled, users without ROLE\_DATA\_EXPORT can no longer see data rows in the Source tab. Instead, the message "You do not have permission to view preview data" is shown in place of the data grid.
### Rules
* When you create a Data Quality rule from a template and run an integration job, the "Predicate" field in the DQ Rule asset now shows the complete SQL query instead of only the primary column name.
* Rules using NOT IN conditions with reference datasets now return the correct non-matching records instead of inverted results.
* Rule break data now exports with correct column alignment.
### Security
* You can now log back into the web interface after a session timeout without encountering an invalid CSRF token error or needing to manually refresh the page.
## Patch release
### 2026.07.1
* You can now view breaking records and preview breaks for rules in pushdown jobs, even when the archive is written to a different database and schema than the source database. Additionally, you can now successfully download CSV files using the "Archive Break Records" option without encountering errors.
* Pushdown jobs using JDBC connections no longer cause the Data Quality service to run out of memory under heavy job loads. The service now handles a high volume of concurrent pushdown jobs more efficiently, improving stability and eliminating the need for manual restarts.
* Data Quality & Observability (self-hosted) now reliably tracks schema changes, even when a monitored table has a large number of column updates. Previously, excessive schema changes could exceed a database index limit and interrupt monitoring. Collibra now caps the list of altered columns to prevent this issue.
* Upgrading to Data Quality & Observability (self-hosted) no longer fails due to a B-tree index error. Previously, this issue occurred when a monitor status row's index key exceeded the database's limit.
{% endtab %}
{% tab title="2026.06" %}
## Release 2026.06
* Expected release date of Data Quality & Observability (self-hosted) 2026.06: July 6, 2026
* Release notes publication date: June 30, 2026
## Announcement
{% hint style="warning" %}
**In the July release (2026.07)**, Data Quality & Observability (self-hosted) is only available on Java 21. Depending on your installation of Data Quality & Observability (self-hosted), you can expect the following:
* **Kubernetes installations**
* Kubernetes containers automatically contain Java 21.
* If you use custom drivers, ensure they are compatible with Java 21.
* **Standalone installations**
* To install Data Quality & Observability (self-hosted) 2026.07, you must upgrade to Java 21.
* If you use custom drivers, ensure they are compatible with Java 21.
* Follow the latest [steps to upgrade](/data-quality-self-hosted/installation/upgrade-data-quality-observability-self-hosted/upgrade-collibra-data-quality-observability.md) to Data Quality & Observability (self-hosted) 2026.07 with Java 21.
* We encourage you to migrate to a Kubernetes installation to improve the scalability and ease of future maintenance.
The June release (2026.06) will be the final release to contain Java 17 builds. Starting in the July release (2026.07), only critical and high-priority bug fixes will be made for Java 17 versions of Data Quality & Observability (self-hosted) from 2026.06 and earlier.For a breakdown of Java availability in current and upcoming releases, click "See what is changing" below.
See what is changing
| Data Quality & Observability (self-hosted) version | Java 17 | Java 21 | Additional notes |
| -------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------- |
| 2026.06 and earlier | Yes | No | |
| 2026.07 and newer | No | Yes | Only critical and high-priority bug fixes will be made for Java 17 versions of Data Quality & Observability (self-hosted) from 2026.06 and earlier. |
{% endhint %}
{% hint style="warning" %}
**Databricks JDBC driver compatibility** The Databricks JDBC driver version (3.4.1) bundled with Data Quality & Observability (self-hosted) 2026.06 has a known issue causing column-metadata retrieval to fail for any rule or dataset whose SQL query contains a WITH ... AS clause. Affected jobs fail with the error "Unable to retrieve column type metadata." Queries without CTEs are unaffected.The workaround is to downgrade to driver version 2.6.36, Simba 2.8.1, or native 2.7.6 until a fix is available in Data Quality & Observability (self-hosted) 2026.07.**Redshift JDBC driver compatibility**The latest Redshift JDBC driver 2.2.6 has a known issue with column type support for TIMESTAMP\_WITH\_TIMEZONE and TIME\_WITH\_TIMEZONE. This causes Pushdown jobs to silently skip these column types, and Pullup jobs to fail with an UNRECOGNIZED\_SQL\_TYPE exception.The workaround is to use driver version 2.1.0.28 until a fix is available in Data Quality & Observability (self-hosted) 2026.07.
{% endhint %}
## New and improved
### Platform
* You can now view Collibra Platform version information in log messages when starting the integration import processes. Spark Standalone customers must replace their log4j2.xml file with the updated version included in this release. No action is required for Kubernetes-based deployments.
### Dataset Manager
* You can now automatically trigger a global webhook alert when a new dataset is created.
* The Dataset Overview now correctly shows null column values as "null" instead of blank cells. This improvement ensures that null values are distinguishable from empty rows, providing a more accurate representation of your data in the results table.
### Admin configuration
* You can now use the new "RULE\_CONDITION\_DISPLAY" feature flag in Configuration Settings. When set to TRUE, the Rules tab and exported XLSX files show ruleCondition instead of ruleValue for pushdown and pullup datasets. By default, the flag is set to FALSE, maintaining the current behavior.
### Security
* We improved the security of our Java libraries.
### Connections
* Databricks and Dremio drivers have been upgraded to the latest major versions to ensure continued compatibility and performance. The Databricks driver is now version 3.4.1, and the Dremio driver is now version 26.0.5-202509091642240013-f5051a07.
* The CDATA Bigquery JDBC driver has been updated to version 25.0.9539.0.
## Fixes
### Jobs
* You can now change the job size in the Job Estimator when the Default Master Path for k8s\:// contains the word “local”.
* The "Driver Memory" field now updates immediately based on the job estimation result. The value is automatically calculated as half of the "Executor Memory" value, with a minimum of 2 GB, ensuring accurate and timely updates.
* Jobs no longer hang when canceled during the initialization (INIT) stage, allowing other jobs to start without requiring a service restart.
### Alerts
* Data Quality alerts with configuration errors no longer prevent other alerts from firing. If an alert encounters an error during processing, it is now skipped and logged, allowing remaining valid alerts to trigger successfully.
* Webhook alerts now properly encode URLs in JSON payloads, ensuring complete URL generation including timestamps for findings\_url variables.
### Dataset Manager
* You can now copy and download query results from the Dataset Overview modal when dataset preview security is disabled, regardless of your user role.
* You can now see an accurate "Active" count that reflects the current dataset definition, even after switching between Add Layers and Review. Navigating to the Review page no longer locks the dataset configuration, allowing you to save additional edits. Added or deleted layers, such as Outliers and Shapes, now remain consistent across all screens until you modify them.
* Dataset audit logs now correctly show DATASET\_SETTINGS instead of SHAPE\_SETTINGS when you change dataset settings via the Explorer, providing more accurate information about the actual changes made.
### Admin configuration
* You can now save ServiceNow assignment queue configurations without setting the "Publish only" option to true. Previously, the system incorrectly required this setting before allowing a successful save.
* You no longer experience premature logouts after 15–20 minutes of inactivity when using local user authentication in Data Quality. Sessions now properly respect the configured 1-hour timeout setting. Local user authentication (non-SSO) is no longer affected by this behavior.
### APIs
* The v3/alerts API now validates that webhook alerts cannot be assigned a batch name, preventing invalid configurations and ensuring webhook recipients do not incorrectly appear in email batch selectors.
* When you create a rule using the POST /v3/rules API endpoint, the dimension name now appears correctly in the response instead of showing as null.
{% endtab %}
{% tab title="2026.05" %}
## Release 2026.05
* Release date of Data Quality & Observability (self-hosted) 2026.05: June 1, 2026
* Release notes publication date: May 25, 2026
## Announcement
{% hint style="warning" %}
**In the July release (2026.07)**, Data Quality & Observability (self-hosted) is only available on Java 21. Depending on your installation of Data Quality & Observability (self-hosted), you can expect the following:
* **Kubernetes installations**
* Kubernetes containers automatically contain Java 21.
* If you use custom drivers, ensure they are compatible with Java 21.
* **Standalone installations**
* To install Data Quality & Observability (self-hosted) 2026.07, you must upgrade to Java 21.
* If you use custom drivers, ensure they are compatible with Java 21.
* Follow the latest [steps to upgrade](/data-quality-self-hosted/installation/upgrade-data-quality-observability-self-hosted/upgrade-collibra-data-quality-observability.md) to Data Quality & Observability (self-hosted) 2026.07 with Java 21.
* We encourage you to migrate to a Kubernetes installation to improve the scalability and ease of future maintenance.
The June release (2026.06) will be the final release to contain Java 17 builds. Starting in the July release (2026.07), only critical and high-priority bug fixes will be made for Java 17 versions of Data Quality & Observability (self-hosted) from 2026.06 and earlier.For a breakdown of Java availability in current and upcoming releases, click "See what is changing" below.
See what is changing
| Data Quality & Observability (self-hosted) version | Java 17 | Java 21 | Additional notes |
| -------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------- |
| 2026.06 and earlier | Yes | No | |
| 2026.07 and newer | No | Yes | Only critical and high-priority bug fixes will be made for Java 17 versions of Data Quality & Observability (self-hosted) from 2026.06 and earlier. |
{% endhint %}
## New and improved
### Platform
* You can now configure a custom Spark executor pod template for Livy by setting global.configMap.data.sparkPodTemplate in your Helm values. When provided, the template is automatically mounted into the Livy deployment and applied to all Livy Session executor pods.
* Data Quality & Observability (self-hosted) now supports EDB Postgres Advanced Server 15.x. This ensures compatibility with the latest certified version of the database.
* All necessary environments now have equivalent Helm variables. This ensures consistency and simplifies configuration management across environments.
* The Redshift JDBC driver is now upgraded to version 2.2.6.
* We improved the security of our Java libraries.
### Jobs
* The Data Quality export functionality now handles large datasets with duplicate detection results more efficiently. The optimized process reduces API calls from thousands to one by using the batch endpoint, ensuring exports complete successfully. If an error occurs, user-friendly messages are now shown to assist with troubleshooting.
* Auto job resource estimation for jobs now allocates sufficient resources during job execution. This ensures that resource allocation matches the dataset size for each run, even if the dataset size changes.
### Collibra Platform integration
* When an integration job fails, the integration console now shows an exception message for the failing data set. You can hover over the message to view the full details, making it easier to troubleshoot job failures.
* You can now use the "Transfer List" component in the Integration Automap UI to view, search, and select multiple tables efficiently. This feature allows you to search tables by name, filter results accurately, and move one or multiple tables between available and selected lists without duplication.
## Fixes
### Platform
* After logging in, you are now redirected directly to the intended resource, such as the Dataset Profile page or the Findings page. Previously, you were redirected to the Home page and had to click the link again to reach your destination.
* As part of the integration process between Data Quality & Observability (self-hosted) and the Collibra Platform, inactive data quality rules now correctly show as "Suppressed" in the Collibra Platform.
### Jobs
* S3 jobs and Livy sessions using AssumeRole or Pod Identity no longer time out on "getFileStatus" after upgrading to Data Quality & Observability (self-hosted) 2026.03. Data Quality & Observability (self-hosted) now propagates the configured S3 region to Spark jobs and Livy sessions, resolving issues caused by cross-region access in environments with private S3 endpoints.
* The dataset source query compilation error in BigQuery Pushdown jobs after upgrading to Data Quality & Observability (self-hosted) 2026.03 has been resolved. The issue, caused by incorrect schema and table name parsing, is now fixed.
* The Metadata bar now correctly displays the schema and table name extracted from complex dataset source queries. This ensures accurate metadata representation for improved clarity and usability.
* Jobs no longer fail with a "value out of range: overflow" error when string or text values are processed as numeric. Statistical aggregations now exclude min, max, and mean absolute values exceeding 1.0e150, preventing PostgreSQL avg and stddev calculations from overflowing. Legitimately large numbers above this threshold are excluded, but infinity values are still preserved.
### Findings
* The Export Limit configuration now enforces limits on CSV downloads and signed links for Pullup job archives. The "Download CSV" button generates files that adhere to the configured export limit, excluding the header row. Similarly, the "Copy Signed Link" action ensures the downloadable file does not exceed the export limit.
* The "Copy Signed Link" and "Download CSV" actions now honor the export\_limit = 0 configuration. When this setting is applied, these actions are disabled, ensuring no records are exported.
### Rules
* The "Data Class Actions" menu no longer disables the UI when you don't select "Edit" or "Delete." You can now close the drop-down by clicking outside of it, ensuring the screen remains interactive.
* Pullup rules now correctly identify break records when you use the equals (=) operator on columns with comma-separated values.
### Dataset Manager
* Deleting or renaming a dataset with a common suffix now resolves only the intended dataset. For example, actions on "test\_dataset" no longer affect "another\_test\_dataset."
### Collibra Platform integration
* Custom rule statuses in Data Quality & Observability (self-hosted) now update to "NA" in the Collibra Platform when a rule moves to an exception state during integration. Once the exception is resolved, the status automatically reverts to "Passing" or "Breaking” in the Collibra Platform. While in the "NA" status, the rule's score is excluded from the Quality tab for that day, improving the handling of rules with exceptions.
* The "Passing Fraction" for the Outlier finding in the Collibra Platform now correctly reflects the corresponding outlier score from Data Quality & Observability (self-hosted). For example, if the outlier score is 70, the "Passing Fraction" is calculated as 30 (100 − 70).
## Patch release
### 2026.05.2
* We improved the security of the application.
### 2026.05.1
* Email notifications sent via the Microsoft Graph API now render correctly as HTML instead of displaying raw HTML source code in your email client.
* Scheduled Pushdown jobs on Snowflake data sources no longer fail with the "Unable to count rows for dataset" error that previously occurred sporadically.
* FIPS deployment no longer fails during password encryption initialization when using Bouncy Castle.
{% endtab %}
{% endtabs %}