activemq-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Antonio D'Errico (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (AMQ-3830) RAR Transacted Message Delivery (Option B JCA Spec 13-30) fail to manage Timeoutexception
Date Mon, 07 May 2012 10:11:49 GMT

     [ https://issues.apache.org/jira/browse/AMQ-3830?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Antonio D'Errico updated AMQ-3830:
----------------------------------

    Description: 
The MDB receives a message within an XA transaction.
If a timeout occurs, the application server transaction manager move the transaction state
to aborted and after ended, rollbacking changes and resetting the transaction id. 
Everything done after delivery of the message is resetted to its initial state.
When, however, the control returns to ActiveMQSession the transactionContext has been cleaned
up so a new local transaction is started.
At this point is invoked the method ServerSessionImpl.afterDelivery and the the proxy MessageEndpoint
(JBoss instance) starts the afterDelivery stuff. 
The application server transaction manager tries to commit a transaction that is already ROLLED_BACK
and rise an exception. Returning to  ServerSessionImpl.afterDelivery the exception is catched
and is executed the finally block, the transaction context is marked as a local so is done
a local commit that remove the message from the input queue. 
As result the message has been consumed (lost) but not processed.

My solution to avoid this problem is to manage in different way the exception (that contains
server side information about transaction) to understand if we need to commit or not.


  was:
The MDB receives a message within an XA transaction.
If a timeout occurs, the application server transaction manager move the transaction state
to aborted and after ended, rollbacking changes and resetting the transaction id. 
Everything done after delivery of the message is resetted to its initial state.
When, however, the control returns to ActiveMQSession the transactionContext has been cleaned
up so a new local transaction is started.
At this point is invoked the method ServerSessionImpl.afterDelivery and the the proxy MessageEndpoint
(JBoss instance) starts the afterDelivery stuff. 
The application server transaction manager tries to commit a transaction that is already ROLLED_BACK
and rise an exception. Returning to  ServerSessionImpl.afterDelivery the exception is catched
and is executed the finally block, the transaction context is marked as a local so is done
a local commit that remove the message from the input queue. 
As result the message has been consumed (lost) but not processed.

     Patch Info: Patch Available
    
> RAR Transacted Message Delivery (Option B JCA Spec 13-30) fail to manage Timeoutexception
> -----------------------------------------------------------------------------------------
>
>                 Key: AMQ-3830
>                 URL: https://issues.apache.org/jira/browse/AMQ-3830
>             Project: ActiveMQ
>          Issue Type: Bug
>          Components: Connector
>    Affects Versions: 5.5.1, 5.6.0
>         Environment: Linux - JBoss 5.5.1 or JBoss 7.1.1 - ActiveMQ 5.5.1/5.6.0
>            Reporter: Antonio D'Errico
>
> The MDB receives a message within an XA transaction.
> If a timeout occurs, the application server transaction manager move the transaction
state to aborted and after ended, rollbacking changes and resetting the transaction id. 
> Everything done after delivery of the message is resetted to its initial state.
> When, however, the control returns to ActiveMQSession the transactionContext has been
cleaned up so a new local transaction is started.
> At this point is invoked the method ServerSessionImpl.afterDelivery and the the proxy
MessageEndpoint (JBoss instance) starts the afterDelivery stuff. 
> The application server transaction manager tries to commit a transaction that is already
ROLLED_BACK and rise an exception. Returning to  ServerSessionImpl.afterDelivery the exception
is catched and is executed the finally block, the transaction context is marked as a local
so is done a local commit that remove the message from the input queue. 
> As result the message has been consumed (lost) but not processed.
> My solution to avoid this problem is to manage in different way the exception (that contains
server side information about transaction) to understand if we need to commit or not.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira

        

Mime
View raw message