Unlock a world of possibilities! Login now and discover the exclusive benefits awaiting you.
Hello Qlik Community,
We would like discuss about a suspected issue in Qlik Replicate. After a source table is recreated and explicitly reloaded, Full Load completes successfully, but subsequent changes to that table are not captured. CDC resumes only after an additional Stop → Resume.
Environment
Qlik Replicate: 2026.5.0.423. Sourse and Target use same Microsoft SQL Server 2019, Microsoft SQL Server (MS-CDC)
The test uses two tables:
dbo.VersionTest — the table being recreated.
dbo.CdcHeartbeat — an unaffected control table. A stored procedure continuously generates committed updates to this table so that the task’s LSN position continues to advance.
The heartbeat procedure remains running throughout the experiment, including periods when the Replicate task is stopped.
VersionTest has a primary key and simple text/numeric columns, without LOB or rowversion columns. We reproduced this behavior in a laboratory while investigating replication recovery after BAS application table restructuring.
Steps to reproduce
Start the Replicate task and confirm that CDC works for both tables. Start the procedure generating continuous committed updates to CdcHeartbeat.
While the Replicate task is running, drop dbo.VersionTest.
Create a new table named dbo.VersionTestNG.
Rename dbo.VersionTestNG to dbo.VersionTest. The table now has the original name but a different SQL Server object_id.
Manually enable MS-CDC for the new dbo.VersionTest. Verify that changes to this table appear in the source cdc.dbo_VersionTest_CT.
Stop the Replicate task and wait until it is Stopped. Mark VersionTest for Reload while the task is stopped, then select Resume Processing.
Wait for Full Load to complete successfully. Execute and commit an UPDATE setting the test value to PROBE_A. This event appears in the source CDC table but does not appear in the target VersionTest__ct. Meanwhile, heartbeat changes continue reaching the target and advance the saved task position beyond the LSN of PROBE_A.
Perform another Stop → Resume Processing, without an additional Reload.
Execute and commit another UPDATE setting the value to PROBE_B. This change is captured successfully. The target VersionTest__ct contains the transition PROBE_A → PROBE_B, but the previously missed event that changed the value to PROBE_A is not recovered.
Expected behavior
After an explicit Reload refreshes the recreated table’s metadata, CDC should resume for that table.
If this DDL scenario requires additional recovery steps, we need a documented, supported procedure that prevents missed changes and clearly identifies when capture remains inactive.
Observed mechanism
At the first Resume, the capture component detects the missing old object and removes its registration. Reload subsequently refreshes the metadata for the new object, but the new CT is not added to the running capture component’s polling list.
The additional task restart establishes the correct registration and restores polling.
The task continues processing other tables and advancing its LSN while capture of VersionTest remains inactive. This creates a gap in the change history, which is critical for downstream historical processing.
An additional Stop → Resume restores future CDC processing but does not recover the missed interval. The current workaround therefore requires keeping writes to the recreated tables paused until Reload, the additional restart, and CDC verification have completed.
The issue was resolved using the procedure recommended by Qlik Support.
If Full Load succeeds after recreating a SQL Server source table, but MS-CDC changes are no longer captured:
Stop the Replicate task and wait until its status is Stopped.
Select Advanced Run Options → Metadata only → Create missing tables and stop. Run it and wait for the task to stop automatically.
Mark the affected table for Reload, then select Resume Processing.
The metadata-only step refreshes internal metadata while preserving the saved processing position. In our test, Full Load and subsequent CDC worked correctly without an additional stop/resume cycle.
This restores ongoing capture; it does not automatically recover previously missed changes.
Our concern is the recovery of change capture following an explicit table Reload, after the replacement source table has been created under the original name and MS-CDC has been enabled for it. We understand the documented limitations of automatic RENAME TABLE handling.
Reload successfully discovers the new source object, persists its updated metadata, and completes Full Load. However, the corresponding source CDC change table remains absent from the active capture component’s polling queries, even after the table’s event handling mode changes to APPLY.
An additional Stop → Resume, without any changes to the task configuration, source table, or MS-CDC configuration, registers the correct CT and restores capture.
The issue was resolved using the procedure recommended by Qlik Support.
If Full Load succeeds after recreating a SQL Server source table, but MS-CDC changes are no longer captured:
Stop the Replicate task and wait until its status is Stopped.
Select Advanced Run Options → Metadata only → Create missing tables and stop. Run it and wait for the task to stop automatically.
Mark the affected table for Reload, then select Resume Processing.
The metadata-only step refreshes internal metadata while preserving the saved processing position. In our test, Full Load and subsequent CDC worked correctly without an additional stop/resume cycle.
This restores ongoing capture; it does not automatically recover previously missed changes.