[jira] [Commented] (ARTEMIS-769) LargeMessage lost when server shutdown while browsing queue

["Damien Hollis (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/ARTEMIS-769?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15537967#comment-15537967 ] 

Damien Hollis commented on ARTEMIS-769:
---------------------------------------

After raising this bug, I've had difficulty replicating the issue with that example - I think the example might actually have a bug in it.  I was going to close the bug but I thought I might just leave it open in case the example does actually show a problem - I do get some strange counts.

Also, I realize now that I think the lost message was caused by something else.  We were just finishing processing a message when we shutdown the server and the affect of that message did not get committed to the DB.  When we restarted, the message whose affect failed to commit to the DB should have been redelivered so it could be processed again but it wasn't.  I guess because we aren't using XA, this is possible - I was hoping it would be the other way around and the DB would commit but the queue would still have the message and we would need to deal with duplicates, which is a lot easier problem.

> LargeMessage lost when server shutdown while browsing queue
> -----------------------------------------------------------
>
>                 Key: ARTEMIS-769
>                 URL: https://issues.apache.org/jira/browse/ARTEMIS-769
>             Project: ActiveMQ Artemis
>          Issue Type: Bug
>          Components: Broker
>    Affects Versions: 1.4.0
>         Environment: Windows and Linux
>            Reporter: Damien Hollis
>            Priority: Critical
>         Attachments: EmbeddedExample.java
>
>
> We have come across an issue that seems to be caused by server being shutdown while browsing a queue with large messages.  In our application scenario a message gets lost from the queue.
> Attached is an example program that shows the issue.  The only problem is that in the test I did, the queue actually has one extra message rather than one less message.  Still, hopefully it helps to point at where this could be happening.
> The example program runs in two modes, first run the program with no arguments and it should cause the problem to occur.  Next re-run the program with an argument of "count" and you should see that the number of messages on the queue is different from the expected number.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)