2026.4 release notes
July 2026
The 2026.4.0 release adds support for a new datasource, Databricks, now available as an Early Access feature in Hyperscale. It also adds Virtual Compliance support for the Snowflake connector.
On this page:
New features
-
New Datasource support - Databricks (Early access)
Added support for Databricks environments hosted on Azure. The connector supports masking of Delta tables backed by external Azure Blob Storage or ADLS storage and uses Unity Catalog APIs for metadata processing. It also includes profiling capabilities, which scan data through the Databricks connector, identify sensitive columns using profile set domain classifiers, and generate a masking inventory with recommended masking algorithms.To learn more, see Databricks connector and Databricks profiler. -
New VM Appliance deployment method (Snowflake)
Hyperscale now supports a VM-based appliance deployment option in addition to the existing Kubernetes model for new deployments of Hyperscale Compliance.
The VM appliance model follows the structure of the traditional engine based hosting model and provides a more straightforward management experience for administrative teams.-
Administrators can perform core system management tasks through the appliance interface, including:
-
Storage management
-
Network management
-
Network security (KeyStore and TrustStore management)
-
Support bundle generation
-
VM-based deployment is available to Oracle, MSSQL, MongoDB, and Snowflake connectors.
Fixed issues
| Issue number | Summary |
|---|---|
| HM-5707 | Fixed an issue where masking jobs on Oracle datasets with more than 1000 indexes in a table group could fail during the pre-load Drop Indexes step with an ORA-01795 error. Index lookups are now batched so jobs complete successfully. |
| HM-5887 |
MongoDB connector: Fixed an issue where load and unload jobs could fail under load due to |
| HM-6089 | Masking service: Fixed an issue where masking jobs could hang because sub-jobs remained in a pre-assignment state when a load-balancer race condition triggered ConcurrentModificationException and NullPointerException errors. |
| HM-6104 | MongoDB Load Service: Added support for optional mongoimport tuning through MONGOIMPORT_MAINTAIN_INSERTION_ORDER, MONGOIMPORT_NUM_INSERTION_WORKERS, and MONGOIMPORT_BATCH_SIZE. Existing behavior is unchanged when these settings are not configured. |
| HM-6153 | MongoDB connector: Fixed an issue where unloaded source (/src) files were not deleted during execution cleanup, which could exhaust the staging mount. Cleanup now removes source files for each execution and honors the RETAIN_EXECUTION_DATA setting. |
| HM-6193 | MongoDB connector: Fixed an issue where masking jobs that used a multi-column algorithm (algorithm_field_name) could fail with the error "File Name cannot have special characters." The unload service no longer sends masking_inventory in the /masks payload. |
Known issues
| Issue number |
Summary |
Workaround |
|---|---|---|
| HM-11280 | A Databricks masking or data transfer job can fail when it loads data into an existing Delta table that uses column mapping. In some cases, the job reports success, but the target table contains NULL values in every column except the partition column. This issue affects tables with the table property delta.columnMapping.mode set to name or id.
|
None |
| HM-5101 | Dataset creation does not currently support payloads or file uploads containing more than 10,000 items. Any attempt to upload files exceeding this limit will fail. |
None |
|
HM-3636 |
MSSQL: Unexpected movement of count from constraints (Primary & Unique) to index in postload response. |
None |
| HM-1397 | Oracle Load fails for table having triggers only with SQL*Loader-937 error. | None |
EOL (end-of-life) notice
Docker Compose - September 2026
Docker Compose deployments for Hyperscale Compliance will reach End of Life (EOL) in September 2026. For more information, see Deprecated and EOL notices.