[
https://issues.apache.org/jira/browse/FLINK-38998?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Sergey Kononov updated FLINK-38998:
-----------------------------------
Description:
Hi!
The following SQL query with a cascaded tumbling window aggregate in the
streaming mode doesn't produce data if offset parameter of window TVF is
specified:
{noformat}
CREATE TABLE source_table (
dim STRING,
user_id BIGINT,
price BIGINT,
row_time TIMESTAMP(3),
WATERMARK FOR row_time AS row_time-INTERVAL '1' SECONDS
) WITH (
'connector' = 'datagen',
'rows-per-second' = '5'
);
CREATE TABLE sink_table (
dim STRING,
uv BIGINT,
window_start TIMESTAMP(3),
window_end TIMESTAMP(3)
) WITH ('connector' = 'print');
INSERT INTO
sink_table
SELECT
dim,
COUNT(*) as uv,
window_start,
window_end
FROM
(
SELECT
dim,
user_id,
window_start,
window_end
FROM
TABLE(
TUMBLE(
TABLE source_table,
DESCRIPTOR(row_time),
INTERVAL '10' SECONDS,
INTERVAL '5' SECONDS
)
)
GROUP BY
dim,
user_id,
window_start,
window_end
)
GROUP BY
dim,
window_start,
window_end;{noformat}
!cascaded_window_agg_job_webui.png|width=882,height=428!
The behavior is the same for hop and cumulate windows with offsets.
The bug is tracked down to FLINK-21305 and Flink version 1.13.
I have a fix for the issue. Please, assign it to me.
Best regards,
Sergey
was:
Hi!
The following SQL query with a cascaded tumbling window aggregate in the
streaming mode doesn't produce data if offset parameter of window TVF is
specified:
{noformat}
CREATE TABLE source_table (
dim STRING,
user_id BIGINT,
price BIGINT,
row_time TIMESTAMP(3),
WATERMARK FOR row_time AS row_time-INTERVAL '1' SECONDS
) WITH (
'connector' = 'datagen',
'rows-per-second' = '5'
);
CREATE TABLE sink_table (
dim STRING,
uv BIGINT,
window_start TIMESTAMP(3),
window_end TIMESTAMP(3)
) WITH ('connector' = 'print');
INSERT INTO
sink_table
SELECT
dim,
COUNT(*) as uv,
window_start,
window_end
FROM
(
SELECT
dim,
user_id,
window_start,
window_end
FROM
TABLE(
TUMBLE(
TABLE source_table,
DESCRIPTOR(row_time),
INTERVAL '10' SECONDS,
INTERVAL '5' SECONDS
)
)
GROUP BY
dim,
user_id,
window_start,
window_end
)
GROUP BY
dim,
window_start,
window_end;{noformat}
!cascaded_window_agg_job_webui.png|width=721,height=350!
The behavior is the same for hop and cumulate windows with offsets.
The bug is tracked down to FLINK-21305 and Flink version 1.13.
I have a fix for the issue. Please, assign it to me.
Best regards,
Sergey
> Cascaded event time Window Aggregate does not produce data when offset is
> specified
> -----------------------------------------------------------------------------------
>
> Key: FLINK-38998
> URL: https://issues.apache.org/jira/browse/FLINK-38998
> Project: Flink
> Issue Type: Bug
> Components: Table SQL / Runtime
> Affects Versions: 2.2.0
> Reporter: Sergey Kononov
> Priority: Major
> Attachments: cascaded_window_agg_job_webui.png
>
> Original Estimate: 24h
> Remaining Estimate: 24h
>
> Hi!
> The following SQL query with a cascaded tumbling window aggregate in the
> streaming mode doesn't produce data if offset parameter of window TVF is
> specified:
> {noformat}
> CREATE TABLE source_table (
> dim STRING,
> user_id BIGINT,
> price BIGINT,
> row_time TIMESTAMP(3),
> WATERMARK FOR row_time AS row_time-INTERVAL '1' SECONDS
> ) WITH (
> 'connector' = 'datagen',
> 'rows-per-second' = '5'
> );
> CREATE TABLE sink_table (
> dim STRING,
> uv BIGINT,
> window_start TIMESTAMP(3),
> window_end TIMESTAMP(3)
> ) WITH ('connector' = 'print');
> INSERT INTO
> sink_table
> SELECT
> dim,
> COUNT(*) as uv,
> window_start,
> window_end
> FROM
> (
> SELECT
> dim,
> user_id,
> window_start,
> window_end
> FROM
> TABLE(
> TUMBLE(
> TABLE source_table,
> DESCRIPTOR(row_time),
> INTERVAL '10' SECONDS,
> INTERVAL '5' SECONDS
> )
> )
> GROUP BY
> dim,
> user_id,
> window_start,
> window_end
> )
> GROUP BY
> dim,
> window_start,
> window_end;{noformat}
> !cascaded_window_agg_job_webui.png|width=882,height=428!
> The behavior is the same for hop and cumulate windows with offsets.
> The bug is tracked down to FLINK-21305 and Flink version 1.13.
> I have a fix for the issue. Please, assign it to me.
>
> Best regards,
> Sergey
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
