This node split rows from a DB Data table. The dialog enables you to specify the number of rows to split and the splitting strategy.
The created partitions might overlap depending on the database and the selected sampling option. KNIME is not storing the data of the first partition in any way but is executing the query that represents the first partition also as part of the query to retrieve the second partition. If the query that defines the first partition returns a different result for each execution the two partitions might overlap. This is most likely the case for random sampling without a fixed seed.