If it were me, I think that I would just have each process open the file-name in shared read/write mode and call it good enough. Of course rows will be added to the file in unpredictable order. I'm not sure that adding complexity is really buying you much.