Sr的Hive外表查询结果和presto 结果不一样

【详述】问题详细描述
【背景】做过哪些操作?
【业务影响】
【StarRocks版本】例如:2.5.1
【集群规模】例如:3fe(1 follower+2observer)+4be(fe与be混部)
【sql】
·SELECT adm_art_flt_analyseorderpayfail_statistics_topfail.dt, adm_art_flt_analyseorderpayfail_statistics_topfail.vendor, SUM(adm_art_flt_analyseorderpayfail_statistics_topfail.fullfail_rate) AS m_a0
, SUM(adm_art_flt_analyseorderpayfail_statistics_topfail.fullfail_rate_diff) AS m_b1, SUM(adm_art_flt_analyseorderpayfail_statistics_topfail.fullfail_effect_rate) AS m_c2
FROM hive_catalog.ads_traindb.adm_art_flt_analyseorderpayfail_statistics_topfail
WHERE adm_art_flt_analyseorderpayfail_statistics_topfail.dt = ‘2023-03-26’
GROUP BY adm_art_flt_analyseorderpayfail_statistics_topfail.dt, adm_art_flt_analyseorderpayfail_statistics_topfail.vendor
HAVING SUM(adm_art_flt_analyseorderpayfail_statistics_topfail.fullfail_rn) <= 3
OR SUM(adm_art_flt_analyseorderpayfail_statistics_topfail.fullfail_rn) IS NULL
ORDER BY adm_art_flt_analyseorderpayfail_statistics_topfail.dt DESC, adm_art_flt_analyseorderpayfail_statistics_topfail.vendor DESC
LIMIT 1000000

explain:新建文本文档 (3).txt (6.9 KB)
explain costs:costs.txt (55.5 KB)

sr hive外表 result:

presto result:

请问使用hive sql的实际执行结果是多少行呢?

实际返回4行,
我仔细看了下,hive外表的查询多了很多行的原因是where 过滤条件没生效,如下:(adm_art_flt_analyseorderpayfail_statistics_topfail.dt = ‘2023-03-26’)

文件格式是orc的吗?

对的,文件是orc 格式

哦,这个问题已经fix了,这两天2.5.4会带着这个fix。

1赞