On Nov 8, 2016, at 6:10 AM, Darran Lofthouse
<darran.lofthouse(a)jboss.com> wrote:
A problem that you quickly run into however is that supporting the
allow-resource-service-restart is that in a number of cases restarting a
service will have the effect of restarting half of the application
server anyway.
As an example if you restart any new security services the chances are
high that you either trigger a restart of all management related
services or all application related services.
So sometimes you may want to go one step further and support real time
updates to the underlying services - however we don't really have a flag
an administrator can use to say if real time updates should be applied.
I would imaging some administrators benefit from being able to apply
management model only changes and coordinate when they actually take effect.
This concern I think would be addressed via a completely separate feature. Get your own
snapshot of the config that is not tied to the runtime at all, manipulate it, and publish
it when ready.
I also think we are more and more moving to a world where the config is set up offline.
That said, when it comes to reducing reload-required and making more things immediately
effective, priority should be given to things where the use case of the admin wanting a
live update is more clear. Things related to diagnostics being a good example.
Regards,
Darran Lofthouse.
On 08/11/16 12:00, Rostislav Svoboda wrote:
>
>> Not all operations handle the allow-service-restart header
>
> Sure, atm it depends a bit on the underlying service.
>
> @statistics-enabled
> Shouldn't be the goal to enable & disable stats without reload or restart
required at all ?
> There are some people/customers which restart or reload their farm of servers only on
planned days.
> For example for WS subsystem we can enable and disable without reload required,
allow-resource-service-restart is to me just helper to avoid full reload.
>
> @Reduce reload-required in general
> I think some long-term effort is needed, less reload-required is better for end-user
experience.
> This probably needs to be more RFE-ish.
>
> Regards.
> Rostislav
>
>>> On 4 Nov 2016, at 15:09, Martin Stefanko <mstefank(a)redhat.com> wrote:
>>>
>>>
>>> Hi,
>>>
>>> I am looking for help with the usage of the allow-resource-service-restart
>>> header. For the following examples each attribute is a boolean with
>>> "restart-required" set to "no-services".
>>>
>>> When I run
>>>
>>>
/subsystem=jmx:write-attribute(name=non-core-mbean-sensitivity,value=true){allow-resource-service-restart=true}
>>>
>>> it works as expected, but
>>>
>>>
/subsystem=undertow:write-attribute(name=statistics-enabled,value=true){allow-resource-service-restart=true}
>>>
>>> for instance, puts the server into the reload-required state.
>>>
>>> This is happening on several other places so I want to ask if it is a
>>> desired behavior.
>>>
>>>
>>> Martin Stefanko
>>>
>>> Associate Software Engineer
>>> JBoss Sustaining Engineering Team
>>> Red Hat Czech s.r.o.
_______________________________________________
wildfly-dev mailing list
wildfly-dev(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/wildfly-dev
--
Brian Stansberry
Manager, Senior Principal Software Engineer
JBoss by Red Hat