Kedar Mhaswade wrote:
>
>> The patch will regress because dynamic reconfig was written based on
>> http-service CHANGE events for handling http-service
>> properties\attributes changes rather than registering a
>> ConfigListener for each
>> com.sun.enterprise.config.serverbeans.Property itself. This is why
>> Jerome added parent notification. Again, we could fine-tune each
>> Property notification to get handled separately and that'll take some
>> time.
>>
> Does this mean that you do in fact rely on http-service CHANGE to be
> delivered upon creation/deletion of a property or any other child?
> This contradicts with what Jan said earlier -- CHANGE event on
> http-service
> should not have been delivered on creation of an http-listener.
I agree with Jan that CHANGE event on http-service delivered on creation
of an http-listener is unnecessary. CHANGE event on http-service should
be delivered when, in fact, http-service property/attribute has been
changed. Property dynamic reconfig rely on parent's CHANGE notification.
> So, for Prelude, how should 6001 be fixed? (Note -- I want to minimize
> the changes and I agree that this has been confusing for a long time).
As Jan suggested earlier, if we could somehow limit/avoid CHANGE event
on http-service when http-listener is created, we can avoid restarting
all http-listeners. Within ConfigListener implementation, AFAIK, cannot
tell whether CHANGE event on http-service is from http-listener creation
or Property change on http-service.
Amy
>
> Thanks,
> Kedar
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe_at_glassfish.dev.java.net
> For additional commands, e-mail: dev-help_at_glassfish.dev.java.net
>