Good afternoon.
Typically, to not do full replaces of data pulls into Domo from our databases, we have used recursive dataflows with ETL. I am trying to think of a way of accomplishing this using the partitioning features in ETL and/or workbench to avoid the ETL still having to process the whole set of updated+historical.
I either can't wrap my head around it, or just don't have a strong enough grasp on the partitioning features, or I'm barking up the wrong function-tree.
I have attached a very small sample of what one of our raw jobs might look like when it gets into Domo and am wondering if anyone thinks partitioning is a good use case for this.
Each separate color column header represents a different table in the SQL Server database we're pulling into workbench from. Each table is connected with the column "donation_id". Each row represents a single, unique donation for which we would only ever want 1 row of in the ETL output.
Other than "donation_id", any of the columns can be modified at any time (the day after a donation or 10 years after a donation). When 1 table is modified in the database, it does not prompt the other tables to be modified.
Example, in the screen shot, the collection_date is 1/1/2020. If any one of these columns changed on 3/30/2023, we would want to bring in that new row, and eliminate the existing row.