[ 
https://issues.apache.org/jira/browse/ARTEMIS-5753?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18037334#comment-18037334
 ] 

ASF subversion and git services commented on ARTEMIS-5753:
----------------------------------------------------------

Commit d84c53ca78f4f14922723eb51af4cfdde9aa35aa in activemq-artemis's branch 
refs/heads/main from Clebert Suconic
[ https://gitbox.apache.org/repos/asf?p=activemq-artemis.git;h=d84c53ca78 ]

ARTEMIS-5753 Fixing Races between Page Oscilation and Replication

this is dealing with possible races between Page oscilation (page / not
paging) and replication start / stop / failback.

Older versions could have issues with PageComplete on future pages.

This is also dealing with a possible deadlock that would happen during
this scenario (replication start and page oscillation)


> Paging has the possibility on losing messages during failback in older 
> versions (before 2.33)
> ---------------------------------------------------------------------------------------------
>
>                 Key: ARTEMIS-5753
>                 URL: https://issues.apache.org/jira/browse/ARTEMIS-5753
>             Project: ActiveMQ Artemis
>          Issue Type: Bug
>    Affects Versions: 2.33.0
>            Reporter: Clebert Suconic
>            Priority: Major
>              Labels: pull-request-available
>             Fix For: 2.45.0
>
>          Time Spent: 1h 40m
>  Remaining Estimate: 0h
>
> We recently changed the paging thread model to be single threaded, so recent 
> versions are not affected by this.
> 2.33 and before could get into a race where Pages are marked as complete and 
> the entire set of files removed, future written pages would then be marked as 
> complete and these messages wouldn't be delivered.
> This happened during failback on the reports I had seen, and on my tests.
> I couldn't reproduce this on latest versions, and that's consistent with the 
> new threading model.
> I am adding a UnitTest introducing the exact conditon that would cause the 
> loss on the latest version and still applying a defense in case it happened, 
> just to be safe.
> I am also applying some extra cleanup on Acks in case the page files are not 
> present, to prevent any possible loss.



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

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
For further information, visit: https://activemq.apache.org/contact


Reply via email to