I have a fairly simple CDC job that I'm trying to put together. My source table has a record type code of "I" for Inserts, "D" for deletes, "UB" for Update Before and "UP" for Update After. I use a Map_CDC_Operation transform to update the destination table based on those codes.
I am not using the Transaction Control feature (because it just throws an error when I use it)
My issue is as follows.
Let's say I have a set of 10,000 Insert records in my source table. Record number 4000 happens to be a duplicate of record number 1. The job will process the records in order starting with record 1 and begin happily inserting records into the destination table. Once it gets to record 4000 however it runs into a duplicate key issue and then my try/catch block catches the error and the dataflow will exit. All records that were inserted prior to the error will be rolled back in the destination.
But the same is not true for updates or deletes. If I have 10000 deletes and 1 insert in the middle that happens to be an insert of a duplicate key, any deletes processed before the insert will not be rolled back. This is also the case for updates.
And again, I am not using Transaction Control, so I'm not sure why the Inserts are being rolled back, but more curiously Updates and Deletes are not being rolled back. I'm not sure why there isn't a consistent result regardless of type of operation. Does anyone know what's going on here or what I'm doing wrong/what my misconception may be?
Environment information: both source and destination are SQL Server 2008 databases and the Data Services version we use is 14.1.1.460.
If you require more information, please let me know.