The April 2026 release of Data Ingestion and Replication includes the following changed behavior.
Empty strings no longer replicated as NULL to Databricks targets
For application ingestion and replication tasks and database ingestion and replication tasks that have a Databricks target, empty strings ("") in a source string column now remain as empty strings in the target. Previously, any empty string in the source data was incorrectly replicated as NULL to the Databricks target.
With the new logic, a new default null marker is used. If a source string column contains NULL data, the data is replicated as infa_null in the Databricks staging file in CSV format. Data Ingestion and Replication interprets any occurrence of infa_null as an actual NULL in the target table.
To enable this behavior for new tasks created in the current release and later, set the writerEmptyStringsAsNulls custom property to false on the Task Details - Target Details page of the task configuration wizard.
For existing tasks deployed in a version earlier than April 2026, set both the writerEmptyStringsAsNulls and disableWriterDatabricksEmptyHandling custom properties to false on the Task Details - Target Details page of the task configuration wizard. Then redeploy the task.
SAP HANA Log-based CDC support for update SPS08
Database ingestion and replication jobs that use Log-based CDC can now process SAP HANA 2.x SPS08 sources. To use SPS08, the database ingestion and replication user must have read access to the SYS.M_SYSTEM_OVERVIEW system monitoring view. This view is required to fetch system overview information, such as the SAP HANA database version, that's needed to detect some changes in the structure of the binary files that SPS08 introduced. Without this information, jobs might encounter file-parsing failures or data discrepancies due to the incorrect parsing of data.
Improved processing of Db2 for z/OS compensation records
Previously, if database ingestion and replication incremental load or combined load jobs with a Db2 for z/OS source encountered unneeded Db2 compensation records in the log, a slowdown in CDC processing might occur. Now, CDC handling of compensation records is more efficient. Instead of removing the compensation log records at the end of the UOW, Database Ingestion and Replication stores the compensation records as key:value pairs in a hash map, which allows these records to be ignored when the log data is released to the log parser.
Improved handling of unload tasks across upgrade boundaries
If any unload tasks in database or application ingestion and replication jobs that use the initial load type are active when an IDMC upgrade occurs, the Database Ingestion agent service now waits for the running unload tasks to complete, up to a maximum of 24 hours, before it shuts down. This behavior helps ensure that the unload tasks have no unexpected shutdown during the upgrade and don’t need to re-read data after the upgrade. Previously, the Database Ingestion agent service stopped the unload tasks during the upgrade and restarted them from the beginning after the upgrade completed.