I start to integrate Activiti in my application, and I faced architecture problems.My bussiness logic looks like:
//this handler proceed ALL logic, and call needed activiti to do the final actions
public void handle(org.eclipse.jetty.websocket.api.Session clientSession) {
//I. Create and Save to database
Session session = openHibernateSession();
MyObject obj = new MyObject();
session.save(obj);
….
//II. Notify client
clientSession.getRemote().sendString(Utils.objectToJson(obj));
//III. Start activiti
runtimeService.signalEventReceived("signal", executionId); //only proceed some minor logic, like saving history!!!
}
Problem:I use WebSockets in my application, so I can't serialize clientSession and pass it to activiti's tasks. As a result, I can't proceed II-section of code (Notify Client) in activiti scope. As a result, I can't proceed in activiti scope all logic that goes before II-section of code. As I understand, the problem is that Activiti should wrap the whole application logic, and provide more flexibility and clearance to my workflow!What I want:Something like next architecture:
//this handler only start the needed activity, and nothing else
public void handle(org.eclipse.jetty.websocket.api.Session clientSession) {
//I. Kick needed Activiti
runtimeService.signalEventReceived("signal", executionId); //proceeed ALL logic like I, II & III-code sections in example above
}
Bad way to solve the problem:I can't pass websocket session to the activiti's tasks, but I can ask it from task (for example save all sessions in static collection).Disadvantage:If I move Activiti to the separate machine, I can't pull static session list!Question: What is the right architecture for this case?
