31.4. Publishing a Service in an OSGi Container

Overview

When you develop an application that will be deployed into an OSGi container, you need to coordinate the publishing and stopping of your endpoints with the life-cycle of the bundle in which it is packaged. You want your endpoints published when the bundle is started and you want the endpoints stopped when the bundle is stopped.
You tie your endpoints life-cycle to the bundle's life-cycle by implementing an OSGi bundle activator. A bundle activator is used by the OSGi container to create the resource for a bundle when it is started. The container also uses the bundle activator to clean up the bundles resources when it is stopped.

The bundle activator interface

You create a bundle activator for your application by implementing the org.osgi.framework.BundleActivator interface. The BundleActivator interface, shown in Example 31.4, “Bundle Activator Interface”, it has two methods that need to be implemented.

Example 31.4. Bundle Activator Interface

interface BundleActivator
{
  public void start(BundleContext context)
  throws java.lang.Exception;

  public void stop(BundleContext context)
  throws java.lang.Exception;
}
The start() method is called by the container when it starts the bundle. This is where you instantiate and publish the endpoints.
The stop() method is called by the container when it stops the bundle. This is where you would stop the endpoints.

Implementing the start method

The bundle activator's start method is where you publish your endpoints. To publish your endpoints the start method must do the following:
  1. Instantiate an javax.xml.ws.Endpoint object for the service provider.
  2. Create an optional server context to use when publishing the service provider.
  3. Publish the service provider using one of the publish() methods.

Example 31.5. Bundle Activator Start Method for Publishing an Endpoint

package com.widgetvendor.osgi;

import javax.xml.ws.Endpoint;
import org.osgi.framework.BundleActivator;
import org.osgi.framework.BundleContext;

public class widgetActivator implements BundleActivator 
{
  private Endpoint endpt;
  ...

  public void start(BundleContext context)
  { 
1    WidgetOrderImpl impl = new WidgetOrderImpl();
2    endpt = Endpoint.create(impl);
3    endpt.publish("http://localhost:9000/SoapContext/SoapPort");
  }

  ...

}
1
Instantiates a copy of the service's implementation object.
2
Creates an unpublished Endpoint for the service implementation.
3
Publish the service provider at http://localhost:9000/SoapContext/SoapPort.

Implementing the stop method

The bundle activator's stop method is where you clean up the resources used by your application. Its implementation should include logic for stopping all of the endpoint's published by the application.
Example 31.6, “Bundle Activator Stop Method for Stopping an Endpoint” shows a stop method for stopping a published endpoint.

Example 31.6. Bundle Activator Stop Method for Stopping an Endpoint

package com.widgetvendor.osgi;

import javax.xml.ws.Endpoint;
import org.osgi.framework.BundleActivator;
import org.osgi.framework.BundleContext;

public class widgetActivator implements BundleActivator 
{
  private Endpoint endpt;
  ...

  public void stop(BundleContext context)
  { 
    endpt.stop();
  }

  ...

}

Informing the container

You must add inform the container that the application's bundle includes a bundle activator. You do this by adding the Bundle-Activator property to the bundle's manifest. This property tells the container which class in the bundle to use when activating the bundle. Its value is the fully qualified name of the class implementing the bundle activator.
Example 31.7, “Bundle Activator Manifest Entry” shows a manifest entry for a bundle whose activator is implemented by the class com.widgetvendor.osgi.widgetActivator.

Example 31.7. Bundle Activator Manifest Entry

Bundle-Activator: com.widgetvendor.osgi.widgetActivator