aleksandra_ch
Databricks Employee
Databricks Employee

Hi @Andolina ,

Indeed, prefer using natural keys in your use case. Identity columns have limitations, it is better to use natural keys if available. 

Alternatively, you can generate surrogate key from the row values itself (with hash or concat function). 

If identity columns are strictly necessary, you'll have to generate them in the bronze table, and put it into theTRACK HISTORY ON in AUTO CDC so that changes on that column are not tracked.

Hope it helps