Does Datastax dsbulk tool duplicates or upsert data when previously loaded file reloaded?

I'm assuming you are referring to the feature that allows a failed operation to be resumed, which was introduced in 1.10.

If so, yes, there is a risk of inserting the same row twice. There is no risk of missing a row though.

As a consequence, you should only use this feature if your data is idempotent, or if you don't care about having duplicates in the database.