在最新的3.2.4版本下，底层表是一张有7天数据，按照时间做小时分区的iceberg表，存储是hdfs。
相同的sql,直接执行需要只需要5s，如果换成异步物化视图,cpu都要吃满，而且查询性能好好几分钟还出不来结果。这个是什么情况？
普通sql：
select
date_trunc(“HOUR”,mtimestamp) as ts,
time_slice(mtimestamp, INTERVAL 5 second) as second_ts,
id,
HLL_UNION_AGG(hll_hash(aid)) as num
from
iceberg.iceberg_db.test
where
mtimestamp >'2024-04-18 00:00:00’and
mtimestamp <‘2024-04-18 01:00:00’
group by
time_slice(mtimestamp, INTERVAL 5 second),
date_trunc(“HOUR”,mtimestamp),
id;
异步物化视图sql：
CREATE MATERIALIZED VIEW test_view
REFRESH DEFERRED MANUAL
PARTITION BY ts
PROPERTIES
(
“partition_refresh_number”=“1”,
“replication_num”=“3”,
“query_rewrite_consistency”=“disable”,
“session.enable_spill”=“true”
)
AS
select
date_trunc(“HOUR”,mtimestamp) as ts,
time_slice(mtimestamp, INTERVAL 5 second) as second_ts,
id,
HLL_UNION_AGG(hll_hash(aid)) as num
from
iceberg.iceberg_db.test
group by
time_slice(mtimestamp, INTERVAL 5 second),
date_trunc(“HOUR”,mtimestamp),
id;
REFRESH MATERIALIZED VIEW test_view PARTITION START (“2024-04-18 00:00:00”) END (“2024-04-18 01:00:00”) ;