AGGREGATE 模型随着ETL不断进行，查询结果不符合预期

[linlin994395]

1. 建表语句
create table studio_xxx.xxx_dim_server
(
server int COMMENT '服务器',
pt_dt date min NOT NULL COMMENT '开服日期'
) ENGINE=OLAP
AGGREGATE KEY(server) COMMENT 'OLAP'
DISTRIBUTED BY HASH(server) BUCKETS 1
PROPERTIES (
"replication_num" = "1"
);
2. 初始化插入语句
INSERT INTO studio_xxx.xxx_dim_server
select
replace(jsonb_extract(data, '$.server'), '"', ""), min(pt_dt) pt_dt
from studio_xxx.xxx_ods_total
where event_name in ('LogReg', 'LogLogin', 'LogPay')
group by replace(jsonb_extract(data, '$.server'), '"', "")
初始化插入后结果如下

3. ETL语句 每10分钟执行一次
INSERT INTO studio_xxx.xxx_dim_server
select
replace(jsonb_extract(data, '$.server'), '"', "") server,
min(pt_dt) pt_dt
from studio_xxx.xxx_ods_total
where pt_dt >= '{start_time}' and pt_dt <= '{end_time}'
and event_name in ('LogReg', 'LogLogin', 'LogPay')
group by replace(jsonb_extract(data, '$.server'), '"', "")
4. 过一段时间 select结果与预期不符（原有2服，3服数据丢失，1服数据pt_dt发生非预期改变）

分析：

1. 怀疑是多副本原因，每次查询到不同的BE节点，BE节点间同步不及时。但设置为单副本，问题依旧，故排除
2. 怀疑AGGREGATE模型在min，replace等聚合方式下，聚合有问题，求解决

[LemonLiTree]

which version？

[linlin994395]

doris-1.2.2-rc01

[LemonLiTree]

show variables like '%vec%';

[linlin994395]