flink-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Hang Ruan (Jira)" <j...@apache.org>
Subject [jira] [Commented] (FLINK-23814) Test FLIP-143 KafkaSink
Date Wed, 18 Aug 2021 09:30:00 GMT

    [ https://issues.apache.org/jira/browse/FLINK-23814?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17400938#comment-17400938
] 

Hang Ruan commented on FLINK-23814:
-----------------------------------

Hi, [~fabian.paul], thanks for your work,

I have applied some tests on the KafkaSink. 
 # run a job with the none delivery guarantee (/)
 # run a job with the at-least once delivery guarantee(/)
 # run a job with the exactly-once delivery guarantee(/)
 # kill a taskmanager while writing in exactly-once mode and restart the taskmanager(/)
 # restart a job with the exactly-once guarantee by some checkpoint(/)
 # restart a job with the exactly-once guarantee by some checkpoint(scale-up)(/)
 # restart a job with the exactly-once guarantee by some checkpoint(scale-down)(x)

The KafkaSink works well in all situations, besides the scale-down situation.

I have proposed the issue here: https://issues.apache.org/jira/browse/FLINK-23854.

Don't know where is wrong.

 

> Test FLIP-143 KafkaSink
> -----------------------
>
>                 Key: FLINK-23814
>                 URL: https://issues.apache.org/jira/browse/FLINK-23814
>             Project: Flink
>          Issue Type: Sub-task
>          Components: Connectors / Kafka
>            Reporter: Fabian Paul
>            Priority: Blocker
>              Labels: release-testing
>             Fix For: 1.14.0
>
>
> The following scenarios are worthwhile to test
>  * Start simple job with None/At-least once delivery guarantee and write records to kafka
topic
>  * Start simple job with exactly-once delivery guarantee and write records to kafka topic.
The records should only be visible with a `read-committed` consumer
>  * Stop a job with exactly-once delivery guarantee and restart it with different parallelism
(scale-down, scale-up)
>  * Restart/kill a taskmanager while writing in exactly-once mode



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

Mime
View raw message