Hi @ManojkMohan @

Thanks for your note about Spark’s distributed retries causing PK violations. I’ll go with the staging table approach, but I have a quick question: does this issue happen only with composite primary keys (like my three-column PK), or could it also occur with a single-column PK?

If it’s only for composite keys, I might be able to merge my three columns into a single PK and avoid the staging table. Does that sound feasible?

Thanks!