Thanks. I can get the domain manger running with your changes (should be
1.6-SNAPSHOT instead of 1.5-SNAPSHOT).
I also have to fix the following issue to get the node cleanly stopped:
https://issues.apache.org/jira/browse/TUSCANY-3077
Thanks,
Raymond
--------------------------------------------------
From: "Simon Laws" <simonsl...@googlemail.com>
Sent: Tuesday, June 02, 2009 3:32 PM
To: "tuscany-dev" <dev@tuscany.apache.org>
Subject: [1.x] distributed travel sample pain
I've been changing the travel sample that some of us are working on to
work with the domain manager. It has been a painful experience. The
problems I came across.
Activators are read twice TUSCANY-3070
Checked work around to symptom into 1.x
Dependencies are not managed correctly TUSCANY-3071
Set up imports to make it work
Lack of error reporting in domain manager TUSCANY-3072
Used debugger
Callbacks over binding.sca are not constructed correctly with domain
manager TUSCANY-3073
Changed to binding.ws in sample
Non QName generated for binding name TUSCANY-3074
Just a validation warning when configured composite it read
Composite processor shouldn't write out callback generated
services/references TUSCANY-3075
Checked fix in 1.x
Deadlock in binding.ws with more than 2 concurrent callbacks TUSCANY-3076
Changed the TravelCatalog to only do two at a time
Still problems with composite resolution in domain manager TUSCANY-3062
Checked in work around but needs further consideration
I've checked in changes to the travel sample (r781189) that should
make it work in the domain manager if you also take the fixes for
3070/3075/3062 (I was actually running on top of 1.5 with these
changes, hence you will see some 1.5-SNAPSHOT dependencies in the
fullapp-launcher pom, but hopefully 1.x is not that different)
I''ve added the distributed configuration to fullapp-launcher. Use
LaunchFullAppDomain to start the domain manager
LaunchFullAppDomainNodes to start all the nodes inside one JVM. I
haven't got to having the nodes start in separate JVMs yet.
The original LaunchFullAppNode that loads all the contributions into
one node is still there and still works for me. It ignores all the
distributed configuration.
Regards
Simon