You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@camel.apache.org by "Claus Ibsen (Jira)" <ji...@apache.org> on 2025/07/23 08:45:00 UTC

[jira] [Resolved] (CAMEL-22259) camel-core - Splitter with shareUnitOfWork should use single uow

     [ https://issues.apache.org/jira/browse/CAMEL-22259?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Claus Ibsen resolved CAMEL-22259.
---------------------------------
    Resolution: Fixed

> camel-core - Splitter with shareUnitOfWork should use single uow
> ----------------------------------------------------------------
>
>                 Key: CAMEL-22259
>                 URL: https://issues.apache.org/jira/browse/CAMEL-22259
>             Project: Camel
>          Issue Type: Improvement
>          Components: camel-core
>    Affects Versions: 4.13.0
>            Reporter: Claus Ibsen
>            Assignee: Claus Ibsen
>            Priority: Major
>             Fix For: 4.14.0
>
>
> A bit related to
> [https://camel.zulipchat.com/#narrow/channel/257298-camel/topic/Camel.20Kafka.20Producer.20throws.20Exception.20in.20loop.2Fsplit/with/530123024]
>  
> The shareUnitOfWork feature is unfortunately not a good solution but was needed at the time to deal with failures after a split and access to the caused exception from the splitting. Using UoW in this solution was maybe in hindsight not as good, and today its a bit of problem.
> The splitter uses child UoW that has the shared uow as parent, but each child is independent despite the name of the option would imply its all a single shared UoW.
> We can refactor the code to make it truly shared as the documentation and name of the option suggests. 
> The existing functionality keeps working. However this change means the done work for each split UoW will be triggered when the parent ends at the end, so if 1000 sub messages in split has some UoW then those 1000 tasks are executed at the very end.
> According to the doc this is what you would expect. However it may have some implications as when those 1000 sub tasks are executed then they may no longer have access to a local resource such as a network connection. For example a FTP client that needs to delete a file after processing. So share unit of work is not "thread safe" to be used in 100% situations. 
> We may consider also deprecating this feature and come up with another solution for the original problem.



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