Hi Benjamin, The Thread.sleep() was an attempt to create a blacklisted service to see what else happens in the container when blacklisting occurs.
The service that was blacklisted is the head of a chain of events and is fired on a timer. It's unlikely but possible that the chain of events took longer than 5 seconds, in which case, we do want to fail fast---but we would also like to recover. It seems like the only option is prevention (wrap it in a thread or disable/increase the timeout). Thanks, Adam From: Benjamin Debeerst <[email protected]> To: "[email protected]" <[email protected]>, Date: 10/18/2013 12:46 PM Subject: RE: EventAdmin Blacklist Hi Adam, I'm not exactly sure which circumstances lead to a blacklisting, but we have also experienced blacklisting in case of Exceptions and long-running event processing. You should keep in mind that the EventAdmin may process the event delivery for each event synchronously to the different listeners, so when doing stuff like Thread.sleep(...) you might actually prevent other listeners to be notified in time. We solved the problem by unpacking the relevant information from the delivered Event object and putting that into a Queue, which in turn is processed asynchronously. This allows to minimize the processing time and the number of exceptions that can rise. (Also see OSGi r4 enterprise specification, chapter 113.8.2: Dealing with Stalled Handlers. "Event handlers should not spend too long in the handleEventmethod.") Regards, Benjamin -----Original Message----- From: Adam Wilson [mailto:[email protected]] Sent: Freitag, 18. Oktober 2013 18:30 To: [email protected] Subject: EventAdmin Blacklist Hi all, We're experiencing a blacklist of a crucial service on a remote system and have not been able to replicate it locally. Once a service has been blacklisted, what can we do about it? From the EventAdmin code, it emits a log message and releases the service. There doesn't appear to be any way to detect the blacklisting and mitigate it. Ideally, it seems like a BLACKLIST event should be emitted so that you can set up handlers to manage these situations---but, obviously, that could get mess.. We're also having a hard time creating a circumstance that creates a blacklisted service to inspect what happens. Is there an easier way to simulate blacklisting? Below is the code that I tried to get something blacklisted, but it did not work. Thanks! Adam package osgi.event; import aQute.bnd.annotation.component.Activate; import aQute.bnd.annotation.component.Component; import aQute.bnd.annotation.component.Deactivate; import aQute.bnd.annotation.component.Reference; import org.osgi.framework.BundleContext; import org.osgi.service.event.Event; import org.osgi.service.event.EventAdmin; import org.osgi.service.event.EventConstants; import org.osgi.service.event.EventHandler; import java.util.HashMap; @Component(immediate=true, name="tester", properties=EventConstants.EVENT_TOPIC+"=testingevent") public class Tester implements EventHandler { public static EventAdmin event; @Reference public void setEvent(EventAdmin e) { event = e; } @Activate public void start(BundleContext ctx) throws Exception { event.postEvent(new Event("testingevent",new HashMap<Object,Object> ())); } @Override public void handleEvent(final Event event) { System.out.println("Event received, sleeping."); Thread.sleep(10*1000); System.out.println("Awake."); } } --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
