|
As of October 27th, please open all new issues in the Red Hat Customer Portal . |
|
[
Permlink
| « Hide
]
Claudio Corsi added a comment - 05/May/11 06:38 AM
This patch contains the fix that can be applied to the 3.3.1 stream.
The patch can basically be translated into two distinct issues to be fixed:
Both issues have been resolved at Apache by now - the fixes still need to get merged into the right branches at FuseSource.
However, in the meanwhile, would it be possible to get back in touch with the customer and check if these two fixes are meeting their requirements. In particular, the proposed fix had a pluggable LateResponseListener but since the main goal there was to avoid throwing exceptions or failing exchanges for timed out HTTP requests and there's really not that much you can do inside the ESB once the request has timed out, I left the extra listener interface out and allowed for configuring the behaviour through a simple 'lateResponseStrategy' flag (cfr. https://issues.apache.org/jira/browse/SMXCOMP-881 Joe,
For SMXCOMP-880, I do agree with Mateusz that the original implementation was broken, but I think the refactoring addresses the original issue as well. For the logging remarks, I'll modify those statements to reuse the existing exchange id variable. With the upgrade to SLF4J a while ago however, we removed the extraneous if (logger.isDebugEnabled()) for all the cases where we can pass the message and a single parameter. Those if-clauses will be there again when this patch gets backported to branches that are using LOG4J though. For SMXCOMP-881, I can see how the exceptions used for a late response can be altered to better reflect what's going on instead of using "HTTP request has timed out..." all over the place, so I'll make that change. I left out the explicit LateResponseListener but allowed for configuring both the behavior in the proposed patch (WARN en end with status DONE) as well as the original behaviour of the component (end the exchange with ERROR). I couldn't really imagine any other way to terminate the MessageExchange, but I could obviously add a callback somewhere for them to implement something else if they have a use case for that? Regards, Gert FIxed the remaining remarks by Mateusz in http://svn.apache.org/viewvc?view=revision&revision=1134568
Backported this fix into the 2011.02-fuse and 2011.01-fuse branches |
||||||||||||||||||||||||||||||||||||||||||