sqoop split by option is giving error while using a derived column in the split by option

Use subquery to wrap row_number calculation, then use derived column in the split-by.

   --query "select col1, ... colN, RANGEGROUP 
               from (select t.*, row_number() OVER (order by t.item_id ) AS RANGEGROUP
                      from table t ) s 
              where 1=1 and $CONDITIONS"

row_number should be deterministic, it means when executed multiple times, it should assign exactly the same number to all rows. What can happen if ORDER BY in the OVER contains not unique column or combination: row_number can return different numbers for the same rows. And if you are using it in the split-by, you will get duplication because the same row can be in split range 1, say 1-100, in mapper2 sqoop will execute same query with filter for range 2, say (101-200) the same row can appear also in that range. Sqoop runs the same query in different containers(mappers) with different condition to get split ranges in parallel.

If Id is int (and much better if it is evenly distributed), use that ID. Why you may need row_number is when it is STRING column. read this: https://stackoverflow.com/a/37389134/2700344, split-column is not necessarily a PK