You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@flink.apache.org by "Valentina Predtechenskaya (Jira)" <ji...@apache.org> on 2023/05/13 18:26:00 UTC

[jira] [Commented] (FLINK-28842) Add client.id.prefix for the KafkaSink

    [ https://issues.apache.org/jira/browse/FLINK-28842?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17722386#comment-17722386 ] 

Valentina Predtechenskaya commented on FLINK-28842:
---------------------------------------------------

Is there any progress on this issue ? As I understood, PR was not accepted because realizaion doesn't guarantee uniqueness of producer's client.id in any case ?

I would be happy to offer another aproach: 
 * client.id can be built exactly in the constructor of [FlinkKafkaInternalProducer|https://github.com/apache/flink/blob/master/flink-connectors/flink-connector-kafka/src/main/java/org/apache/flink/connector/kafka/sink/FlinkKafkaInternalProducer.java]
 * to support uniqueness of client.id it can be built with static counter (not subtaskId) - same as it was [previously done|https://github.com/apache/kafka/blob/2.4/clients/src/main/java/org/apache/kafka/clients/producer/KafkaProducer.java#L443] in kafka-clients with default "producer-n" value 

I have already checked this approach in the local patch of the connector, I suppose it works well.

 

> Add client.id.prefix for the KafkaSink
> --------------------------------------
>
>                 Key: FLINK-28842
>                 URL: https://issues.apache.org/jira/browse/FLINK-28842
>             Project: Flink
>          Issue Type: New Feature
>          Components: Connectors / Kafka
>    Affects Versions: 1.15.1
>            Reporter: Yaroslav Tkachenko
>            Assignee: Yaroslav Tkachenko
>            Priority: Major
>              Labels: pull-request-available
>
> Currently, KafkaSink doesn't provide a way to configure a client.id.prefix like KafkaSource does. client.id is as important for Kafka Producers, so it makes sense to implement the missing logic for the KafkaSink. 
> A similar implementation that leverages subtaskId for uniqueness can be used here.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)