Are you running glassfish in a separate JVM/box and you want to deploy a Camel route on another JVM/box which runs JBoss Fuse?
Or how is your setup?
Yes, glassfish is running in a separate VM. I want to make that service available to agents on the Fabric in JBoss Fuse with the minimum number of hops. The service running on glassfish running in a separate VM should be bound as directly as possible to the Fabric with minimal change. It has been suggested to me that the way to do this is to use a ServletContextListner in the service running on glassfish and have it (using the ZooKeeper client libs) register the endpoints into the Fabric registry.
More on the set-up.
Before I had gotten to this point I had my POC working by installing JBoss Fuse on the same VM as the Glassfish container - it took a little port manipulation. Then I joined the Fabric from the Fabric Container. Once the Fabric Container had joined the Fabric (or perhaps I should write, once that JBoss Fuse instance had joined the Fabric and become a Fabric Container) I added my profile, with a Camel Http Proxy route, to that instance of the container. I was then able to call my service through the ESB using a consuming route on another Fabric Container.
So what I am trying to do is get that same route, without having to install JBoss Fuse on every node with a Glassfish service we want on the Bus.
Yeah when your glasfish services becomes available you can use the ZooKeeper API to register the services in the ZK Registry in Fabric. Then fabric can _see_ this service and leverage it.
1 of 1 people found this helpful
I logged a ticket in the fabric8 project about having a simpler API you can use in these kind of use-cases
Thank you for looking into this. I agree that this would be a great feature for Fabric8. During this entire adventure I have been thinking this has to be a common scenario.
I guess the thing to figure out now is the ZooKeeper API. If you have a sample lying around send it my way.
If I figure it out I'll post the code here. In the meantime the Fuse support guys at RedHat actually got behind the "parallel container" solution as a supportable solution which I think, for the forum, is considered the answer to my question.