Other: HVR6: Capture Rewind table isolation
During a recent incident when data integrity in one of our production tables were affected, we had to run a capture rewind to solve the issue. But in doing so, we found that running the activating a capture rewind also means that the integrate jobs for all the other tables went into a suspended state.
It is our understanding that currently the way to do it would be.
- Run activate > capture rewind for affected table for say 1 week
- Wait till the integrate for the affected table catches up
- Do validation checks on the affected table
- Run activate > capture rewind on ALL tables but for the timestamp we started step 1
If there was some table level isolation in how Activate jobs are triggered, this could be simplified to step 1-3 since the integrate jobs for other tables would not be suspended.
-
Harish,
HVR was designed to maintain transactional consistency between all tables in a channel. Whilst transactional consistency is no longer critical to many - primary analtytical - use cases, it is still ingrained in HVR. And because the transaction log is the basis for transactional consistency there is only a single starting point for change data capture. You cannot rewind just for one table; you rewind for all.
In most cases when a table goes out of sync the user will refresh the table (rather than replay changes). As of the time of this writing such refresh will halt the integration on all other tables until the refresh finishes. However, this blocking of integrate for all other tables will be lifted. We have been working on an isolated table refresh feature that I expect to be released for early adopters in the next few months (before the end of 2024). This would be a great feature to help address your issue.
Hope this helps.
Mark.
Please sign in to leave a comment.
Comments
1 comment