kafka-jira mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Marcel "childNo͡.de" Trautwein (JIRA) <j...@apache.org>
Subject [jira] [Commented] (KAFKA-5157) Options for handling corrupt data during deserialization
Date Thu, 22 Feb 2018 09:04:00 GMT

    [ https://issues.apache.org/jira/browse/KAFKA-5157?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16372564#comment-16372564
] 

Marcel "childNo͡.de" Trautwein commented on KAFKA-5157:
-------------------------------------------------------

Just a question ... What I "dislike" here from a native point of view, is that if you want
to configure your implementing {{DeserializationExceptionHandler}} you have that well defined
{{Configurable}} interface, but what is injected is the {{StreamsConfig}} so you have to put
configurations into the streamsConfig map to access them later in the implementation ... is
this really the desired / designed procedure and / or best practice?

> Options for handling corrupt data during deserialization
> --------------------------------------------------------
>
>                 Key: KAFKA-5157
>                 URL: https://issues.apache.org/jira/browse/KAFKA-5157
>             Project: Kafka
>          Issue Type: Sub-task
>          Components: streams
>            Reporter: Eno Thereska
>            Assignee: Eno Thereska
>            Priority: Major
>              Labels: user-experience
>             Fix For: 1.0.0
>
>
> When there is a bad formatted data in the source topics, deserialization will throw a
runtime exception all the way to the users. And since deserialization happens before it was
ever processed at the beginning of the topology, today there is no ways to handle such errors
on the user-app level.
> We should consider allowing users to handle such "poison pills" in a customizable way.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Mime
View raw message