[Bug] Select count(*) return 2 results during stream load 2w rows/per seconds

[wangwq10]

Search before asking

• I had searched in the issues and found no similar issues.

Version

2.1.5 rc02

What's Wrong?

back ground: stream load within 2w rows/per seconds.
case:
1 select count() return 2 results and select(dt) get one result
2. sum results in select count() which equals to select(dt)
3. all stream load label has commit and visible matched, suppose data is all really visible.
4. after stop stream for about 20min, select count(*) return one result.

Transation like :

ddl like:

CREATE TABLE IF NOT EXISTS xxxx
(
__time__ DATETIME default CURRENT_TIMESTAMP,
trace_id VARCHAR(200),
instance VARCHAR(500),
span_id STRING,
service_name VARCHAR(200),
duration BIGINT,
__dt__ DATE ,
resource_attributes MAP<STRING, STRING>,
events ARRAY<STRUCT<time:DATETIME, name:STRING, attributes:MAP<STRING, STRING>>>
INDEX idx_service_name(service_name) USING INVERTED,
INDEX idx_timestamp(time) USING INVERTED,
INDEX idx_trace_id(trace_id) USING INVERTED,
INDEX idx_instance(instance) USING INVERTED,
INDEX idx_span_id(span_id) USING INVERTED
) ENGINE = OLAP
DUPLICATE KEY(time, trace_id)
PARTITION BY RANGE(dt) ()
DISTRIBUTED BY HASH(trace_id) BUCKETS AUTO

What You Expected?

select count(*) got 1 result

How to Reproduce?

No response

Anything Else?

No response

Are you willing to submit PR?

• Yes I am willing to submit a PR!

Code of Conduct

• I agree to follow this project's Code of Conduct

[ixzc]

we will fix this

[mrhhsg]

@wangwq10
Hi,
If this issue can be reproduced, please provide the profile file.