>> We're getting NULL values on key columns for some records replicated from HANA with CDC logs.
Those are always 'fun' problems to tackle and more often than not require detailed investigation of the data stream.
You'll need to be a lot, whole lot, more precise and concise with your information here to have any hope that anyone can magically provide a solution. We are NOT mind readers. More than likely you'll need a support ticket for this, which also needs all the data so you may as well start here.
"key columns" - as in IMPORTANT columns or columns which are part to the Primary Key (the only key Replicate cares about). What are the (common) attributes for those 'key' columns? Dates? Numbers? Precision? Name?
"from HANA " - Where is the data going to ? What versions for Replicate? Source DB, Target DB, Drivers?
"incremental load filter pattern." - What does that mean? The opening suggests CDC activity?
"ChangeSeq" - Do you mean 'changeSequence'. It's unlikely to be relevant but what do you do to it?
"Can you please help determine the possible issue and solution? " - No Way - Look back at the information you provided - how on earth can anyone help with that - UNLESS they had a possibly similar issue.
We can help you for you to determine the cause and perhaps a solution.
Have you checked the REPTASK_xxx.LOG files and APPLY_EXCEPTION tables for any Warnings?
"some records" - Here is the biggest clue. Some, but not all. Correct? So all YOU need to do is determine what make some records special. Since the method used (implicit conversions) works some of the time the most likely cause is that some specific data values cannot be handled. One classic example is an invalid date on the source. It might even be valid on the source, but not on the target. The best approach to this is to 'find' a handful of rows with the issue on the target, and then find the corresponding rows on the source and see what the data values for the bad columns look like. Do you see a pattern? Look at some rows 'next' to the bad rows which were valid. What is significantly difference? very high, very low? 0 vs value? SPACE vs values? Accented/UTF-16 very special characters versus 'regular' ascii?
If you cannot figure it out you may want to add a reply here (and to the support case once you have that) to attach a CSV file with SOURCE and TARGET values - with rows reduced to PK + the suspect column(s) and some good, and some bad rows.
Good luck!
Hein.
