Yeah, I know the rollback doesn't work that way. That wasn't what I was
driving at. I am just trying to deal with a business problem of informing
the investigator why an exchange failed though means other than spelunking
in a verbose log file for clues. Sometimes one investigating a problem may
not have access even to the logs. For example if a business issue such as a
studio's game site went down causing the exchanges to fail. In this
circumstance we would rather push the exchange to a dead letter queue and
then later be able to determine why the problem happened and relish those
exchanges. Again finding the "why" in a log file isn't going to work out.
-- Robert
On Friday, May 23, 2014, Claus Ibsen <claus.ib...@gmail.com> wrote:
> You cannot provide a reason why a rollback happens in the TX API.
> There is no rollback(String message) api for that.
>
> You need to record this somewhere else if you want to use that.
> Or do your own kind of TX rollback.
>
> On Thu, May 22, 2014 at 11:00 PM, kraythe . <kray...@gmail.com<javascript:;>>
> wrote:
> > Greetings, I have many routes that read from AMQ queues and write to
> other
> > AMQ queues. For error handling I have been using the following paradigm:
> >
> > from("activemq:queue:inbox")
> >
> >
> .onException(Exception.class).useOriginalMessage().handled(handled).maximumRedeliveries(0)
> > .setHeader(Exchange.FAILURE_ROUTE_ID,
> property(Exchange.FAILURE_ROUTE_ID))
> > .setHeader(Exchange.EXCEPTION_CAUGHT,
> simple("${exception.stacktrace}"))
> > .to(ExchangePattern.InOnly, "activemq:queue:dead").end();
> > .process(new PreProcessor())
> > .to("activemq:queue:outbox")
> > .process(new PostProcessor());
> >
> > The goal being that if there was any kind of exception, that information
> > would be noted in the headers on the message before it was sent off to
> the
> > DLQ. The problem is this runs into roadblocks when transacted() is added
> to
> > the route.
> >
> > With the code above when the route is transacted and a message fails in
> > post processing I want the message pulled off the outbox, record
> > information indicating what the exception was, and have the message sent
> to
> > the DLQ. Unfortunately I seem to be in a quandary of how to do this.
> >
> > The code above simply wont work because the message being sent to the
> dead
> > letter queue gets rolled back along with the outbox if I mark the
> exchange
> > with rollback() in the exception handler. If I don't mark the message
> with
> > rollback() in the handler then the outbox doesn't get rolled back if the
> > post processor exceptions but the dead letter channel will contain the
> > correct information. On the other hand if I just let activemq handle the
> > transaction, it will retry and then eventually send the message to the
> DLQ
> > but the message in the DLQ will contain no information about why it
> failed.
> >
> > So I want my cake and eat it too. I want to be able to record WHY an
> > exchange failed but still be able to rollback the outbox at the same
> time.
> > I have been plugging away at this a ton and I am out of ideas. What would
> > not be acceptable would be to require the user to troll through log files
> > trying to find a reason why an exchange failed. This would be an
> > operational nightmare. So the message and the reason for the rollback
> need
> > to be somewhere accessible and they need to be together.
> >
> > I would appreciate any suggestions on how I could make this happen.
> >
> > *Robert Simmons Jr. MSc. - Lead Java Architect @ EA*
> > *Author of: Hardcore Java (2003) and Maintainable Java (2012)*
> > *LinkedIn: **http://www.linkedin.com/pub/robert-simmons/40/852/a39
> > <http://www.linkedin.com/pub/robert-simmons/40/852/a39>*
>
>
>
> --
> Claus Ibsen
> -----------------
> Red Hat, Inc.
> Email: cib...@redhat.com <javascript:;>
> Twitter: davsclaus
> Blog: http://davsclaus.com
> Author of Camel in Action: http://www.manning.com/ibsen
> hawtio: http://hawt.io/
> fabric8: http://fabric8.io/
>
--
*Robert Simmons Jr. MSc. - Lead Java Architect @ EA*
*Author of: Hardcore Java (2003) and Maintainable Java (2012)*
*LinkedIn: **http://www.linkedin.com/pub/robert-simmons/40/852/a39
<http://www.linkedin.com/pub/robert-simmons/40/852/a39>*
