kafka-jira mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "ASF GitHub Bot (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (KAFKA-5435) Produce state lost if no snapshot retained
Date Sat, 17 Jun 2017 00:36:00 GMT

    [ https://issues.apache.org/jira/browse/KAFKA-5435?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16052579#comment-16052579
] 

ASF GitHub Bot commented on KAFKA-5435:
---------------------------------------

GitHub user hachikuji opened a pull request:

    https://github.com/apache/kafka/pull/3361

    KAFKA-5435: Improve producer state loading after failure

    

You can merge this pull request into a Git repository by running:

    $ git pull https://github.com/hachikuji/kafka KAFKA-5435-ALT

Alternatively you can review and apply these changes as the patch at:

    https://github.com/apache/kafka/pull/3361.patch

To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:

    This closes #3361
    
----
commit 5bf7d189589e4986e9552c955d101fe8e352d49c
Author: Jason Gustafson <jason@confluent.io>
Date:   2017-06-17T00:25:12Z

    KAFKA-5435: Improve producer state loading after failure

----


> Produce state lost if no snapshot retained
> ------------------------------------------
>
>                 Key: KAFKA-5435
>                 URL: https://issues.apache.org/jira/browse/KAFKA-5435
>             Project: Kafka
>          Issue Type: Sub-task
>          Components: clients, core, producer 
>            Reporter: Jason Gustafson
>            Assignee: Jason Gustafson
>            Priority: Blocker
>             Fix For: 0.11.0.0
>
>
> We have an optimization in {{Log}} to avoid the need to scan the log to build producer
state during the upgrade path. Basically, if no producer snapshot exists, then we assume that
it's an upgrade and take a new snapshot from the end of the log. Unfortunately, it can happen
that snapshot files are never created or are deleted through truncation. Upon reinitialization,
this can cause the optimization above to kick in and we lose the current state of all producers.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

Mime
View raw message