需求：对于xxx项目入库到hive数仓的小时数据，用户希望准实时了解这部分入库数据的hdfs大小

方案1：直接du

缺点：高峰期如果用户du的path不合理，或者频率过高，可能会导致namenode 出现hung，造成集群响应慢

方案2：通过metastore的parameters.totalSize拿到表或者分区的大小

也可以直接查 metastore 数据库，数据源和上面使用desc查询一样：

select
    dbs.DB_ID,
    dbs.NAME,
    tbls.TBL_NAME,
    tbls.TBL_ID,
    partition_params.PARAM_KEY,
    sum(partition_params.PARAM_VALUE) as totalSize
from
    DBS As dbs
    left join TBLS as tbls
    on tbls.DB_ID = dbs.DB_ID
    left JOIN partitions AS partitions
    on tbls.TBL_ID = partitions.TBL_ID
    left JOIN partition_params
    on partition_params.part_ID = partitions.part_ID
where
    dbs.NAME in ('pods','yfzx','pdwd')
    AND tbls.TBL_NAME in ('x','y','z','u')
    AND partition_params.PARAM_KEY in ('totalSize')
    group by dbs.DB_ID,dbs.NAME,tbls.TBL_NAME,partition_params.PARAM_KEY,tbls.TBL_ID;