We give an overview of what rollbacks are, but no examples of where the file would be created and named (we do say it's in the dbpath though). We should probably describe a couple different scenarios of how to deal with the rolled back data. For example, in an insert only app, it would be appropriate to just re-insert the data into the primary. However, in many applications, it is a bad idea to blindly re-insert the data. The user will have to write the data into a temporary collection, then read each document, do some sort of checking and decide whether to re-insert or not. This is highly application dependent, so quite difficult to prescribe exact recipe for how to deal. This point should be made very clear.