You are viewing a plain text version of this content. The canonical link for it is here.

Posted to jira@kafka.apache.org by "feyman (Jira)" <ji...@apache.org> on 2020/06/04 07:56:00 UTC

[jira] [Commented] (KAFKA-10097) Avoid getting null map for task checkpoint

    [ https://issues.apache.org/jira/browse/KAFKA-10097?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17125652#comment-17125652 ] 

feyman commented on KAFKA-10097:
--------------------------------

Hi, [~bchen225242], I'm interested in kafka streams and would like to take some tasks about it to get some further understanding, wondering if this one is appropriate as a starting point ? Or could you recommend some alternatives if not? Thanks!

> Avoid getting null map for task checkpoint
> ------------------------------------------
>
>                 Key: KAFKA-10097
>                 URL: https://issues.apache.org/jira/browse/KAFKA-10097
>             Project: Kafka
>          Issue Type: Improvement
>          Components: streams
>            Reporter: Boyang Chen
>            Priority: Major
>
> In StreamTask, we have the logic to generate a checkpoint offset map to be materialized through StateManager#checkpoint. This map could be either empty map or null, which the former case indicates to only pull down existing state store checkpoint data, while the latter indicates no need to do a checkpoint in the case such as we are suspending a task.
> Having two similar special logics for checkpointing could lead to unexpected bugs, also we should think about separating the empty checkpoint case vs passed-in checkpoint case.



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