If you're already using partitioned dataset then just writing a zero row dataframe should work