I'm working on adding a remote JMXConnector to AS trunk via a new jmx-jboss-beans.xml (see http://pastebin.com/JjsiRrh9). I would like to bind a string representation of the JMX service URL to JNDI. Any pointers to doing the bind (from a jboss-beans.xml) are welcome. If not, I will hack through it but wanted to check first. :)

jboss-as-testsuite/src/main/org/jboss/test/naming/restart/ObjectBinder.java binds a particular kind of object into jndi. See jboss-as-testsuite/src/resources/naming/restart/naming-restart-jboss-beans.xml: <bean name="RestartObjectBinder" class="org.jboss.test.naming.restart.ObjectBinder"> <property name="namingService"><inject bean="RestartNamingService"/></property> </bean> It could be generalized. -Ron Scott Marlow wrote: > I'm working on adding a remote JMXConnector to AS trunk via a new > jmx-jboss-beans.xml (see http://pastebin.com/JjsiRrh9). I would like to > bind a string representation of the JMX service URL to JNDI. Any > pointers to doing the bind (from a jboss-beans.xml) are welcome. > > If not, I will hack through it but wanted to check first. :) > > Thanks, > Scott > > _______________________________________________ > jboss-development mailing list > jboss-development(a)lists.jboss.org > https://lists.jboss.org/mailman/listinfo/jboss-development > _______________________________________________ jboss-development mailing list jboss-development(a)lists.jboss.org https://lists.jboss.org/mailman/listinfo/jboss-development

Or you could just create a simple bean to represent the remote JMX service, and have it construct the remote JMX objects and jndi entries as part of start/stop. Although this makes me wonder what cases we have where you need to pull a remote jmx reference from jndi? If this is locally running code the current mbeanserver is all you need. So I am guessing there is some ha-jndi case you are trying to support? David M. Lloyd wrote: > Also maybe the @org.jboss.aop.microcontainer.aspects.jndi.JndiBinding > annotation might do this another way? > > Either way it would be nicer if you could do: > > <jndi-binding bean="TheBean" name="the/name/in/JNDI"/> > > or something like that. > > On 03/25/2010 01:12 PM, Ron Sigal wrote: >> jboss-as-testsuite/src/main/org/jboss/test/naming/restart/ObjectBinder.java >> binds a particular kind of object into jndi. See >> jboss-as-testsuite/src/resources/naming/restart/naming-restart-jboss-beans.xml: >> >> >> <bean name="RestartObjectBinder" >> class="org.jboss.test.naming.restart.ObjectBinder"> >> <property name="namingService"><inject >> bean="RestartNamingService"/></property> >> </bean> >> >> It could be generalized. >> >> -Ron >> >> Scott Marlow wrote: >>> I'm working on adding a remote JMXConnector to AS trunk via a new >>> jmx-jboss-beans.xml (see http://pastebin.com/JjsiRrh9). I would like to >>> bind a string representation of the JMX service URL to JNDI. Any >>> pointers to doing the bind (from a jboss-beans.xml) are welcome. >>> >>> If not, I will hack through it but wanted to check first. :) >>> >>> Thanks, >>> Scott >>> >>> _______________________________________________ >>> jboss-development mailing list >>> jboss-development(a)lists.jboss.org >>> https://lists.jboss.org/mailman/listinfo/jboss-development >>> >> _______________________________________________ >> jboss-development mailing list >> jboss-development(a)lists.jboss.org >> https://lists.jboss.org/mailman/listinfo/jboss-development >

Ah ok. I had it in my head that jconsole used some kind of fake representation of jndi. David M. Lloyd wrote: > The problem is that in order to use the JSR160 RMI transport, you have to > get the stub to the client. You can either do this by URL (which means > that the client has to have the whole insanely long URL > ["service:jmx:rmi://[host[:port]]/stub/really-long-base64-encoded-stub"] in > order to connect, thus making automated connection impossible), or by > binding the stub into JNDI and using a JNDI URL > ["service:jmx:rmi://[host[:port]]/jndi/jndi-name"]. The latter is roughly > what we do today with RMIAdaptor (sic), so it seems the least invasive to > me since one would think that things like Twiddle would then only have to > change the interface of what they pull out of JNDI. > > On 03/25/2010 01:26 PM, Jason T. Greene wrote: >> Or you could just create a simple bean to represent the remote JMX >> service, and have it construct the remote JMX objects and jndi entries >> as part of start/stop. >> >> Although this makes me wonder what cases we have where you need to pull >> a remote jmx reference from jndi? If this is locally running code the >> current mbeanserver is all you need. So I am guessing there is some >> ha-jndi case you are trying to support? >> >> David M. Lloyd wrote: >>> Also maybe the @org.jboss.aop.microcontainer.aspects.jndi.JndiBinding >>> annotation might do this another way? >>> >>> Either way it would be nicer if you could do: >>> >>> <jndi-binding bean="TheBean" name="the/name/in/JNDI"/> >>> >>> or something like that. >>> >>> On 03/25/2010 01:12 PM, Ron Sigal wrote: >>>> jboss-as-testsuite/src/main/org/jboss/test/naming/restart/ObjectBinder.java >>>> binds a particular kind of object into jndi. See >>>> jboss-as-testsuite/src/resources/naming/restart/naming-restart-jboss-beans.xml: >>>> >>>> >>>> <bean name="RestartObjectBinder" >>>> class="org.jboss.test.naming.restart.ObjectBinder"> >>>> <property name="namingService"><inject >>>> bean="RestartNamingService"/></property> >>>> </bean> >>>> >>>> It could be generalized. >>>> >>>> -Ron >>>> >>>> Scott Marlow wrote: >>>>> I'm working on adding a remote JMXConnector to AS trunk via a new >>>>> jmx-jboss-beans.xml (see http://pastebin.com/JjsiRrh9). I would like to >>>>> bind a string representation of the JMX service URL to JNDI. Any >>>>> pointers to doing the bind (from a jboss-beans.xml) are welcome. >>>>> >>>>> If not, I will hack through it but wanted to check first. :) >>>>> >>>>> Thanks, >>>>> Scott >>>>> >>>>> _______________________________________________ >>>>> jboss-development mailing list >>>>> jboss-development(a)lists.jboss.org >>>>> https://lists.jboss.org/mailman/listinfo/jboss-development >>>>> >>>> _______________________________________________ >>>> jboss-development mailing list >>>> jboss-development(a)lists.jboss.org >>>> https://lists.jboss.org/mailman/listinfo/jboss-development

On Thu, 2010-03-25 at 13:35 -0500, David M. Lloyd wrote: > The latter is roughly > what we do today with RMIAdaptor (sic), so it seems the least invasive to > me since one would think that things like Twiddle would then only have to > change the interface of what they pull out of JNDI. Not really because RMIAdaptor extends javax.management.MBeanServerConnection the later is what people should be using rather a jboss api.

P.S. I don't really understand this change. It breaks a number of "features". By using the JMX connector in the JDK you loose the ability to do remote JMX invocations that are involved in user transactions.

You're also missing a way to integrate the security with the java:/jaas/jmx-console policy. You can't add load balancing/failover which the old jmx-invoker could do if you configured it.

Finally, you loose the serialization workarounds, e.g. <!-- Interceptor that deals with non-serializable results --> <interceptor code="org.jboss.jmx.connector.invoker.SerializableInterceptor" policyClass="StripModelMBeanInfoPolicy"/> and the swapping of the MBeanServer to an MBeanServerConnection when passing MBeanProxys over the network.

Also, it means you pretty much have to use RMI (that's all that is implemented in the JDK) which a lot of people don't want to do, its painful through a proxy.

While the old jmx-invoker can be swapped out to use the HTTPInvoker. There was a customer on the support list asking about that only last week.

On Thu, 2010-03-25 at 14:57 -0500, David M. Lloyd wrote: > On 03/25/2010 02:05 PM, Adrian Brock wrote: > > By using the JMX connector in the JDK you loose the ability to do > > remote JMX invocations that are involved in user transactions. > > How is this a bad thing? That there's even an expectation that > transactions will "work" over JMX seems crazy to me. It's ok to > discontinue features which don't make sense. > It's a bad thing because you've deleted a feature.

The requirement is not completely off the wall, since it came up in the JMX 2.0 spec committee. We decided there that we wouldn't address it, since the mechanism depends upon the transport, e.g. UserTransaction in JavaEE or "current" in CORBA, etc. So its really an implementation detail (value add) of the protocol.

I'm working on adding a remote JMXConnector to AS trunk via a new jmx-jboss-beans.xml (see http://pastebin.com/JjsiRrh9). I would like to bind a string representation of the JMX service URL to JNDI. Any pointers to doing the bind (from a jboss-beans.xml) are welcome. If not, I will hack through it but wanted to check first. :)

Thanks, Scott _______________________________________________ jboss-development mailing list jboss-development(a)lists.jboss.org https://lists.jboss.org/mailman/listinfo/jboss-development