我不知道它为spark编写的查询是否正确，但作为问题的一部分，下一个查询将从hive metastore（存储在mysql中）返回分区的位置：

SELECT 
      location 
    FROM 
      hive.DBS 
    LEFT JOIN hive.TBLS 
    ON 
      ( 
        DBS.DB_ID       = TBLS.DB_ID 
      AND upper(TBLS.tbl_name) = 'CG_CRDSTDLVRS_RAW' 
      ) 
    LEFT JOIN hive.PARTITIONS 
    ON 
      ( 
        TBLS.TBL_ID = PARTITIONS.TBL_ID 
      AND PARTITIONS.PART_NAME LIKE '%2018-09-15%' 
      ) 
    LEFT JOIN hive.SDS 
    ON 
      ( 
        PARTITIONS.SD_ID = SDS.SD_ID 
      ) 
    WHERE 
      upper(DBS.name)='RAW_FAT01'