Thanks for your answer. That's what is suspected. So my options are to use
tcp between modules that run in the same JVM but in diffrent classloaders,
or put activemq + deps in a shared classloader. Hmm..
Anyway i tried this out a bit and it works when sending strings. When i try
to send a Serializable object it also works, but as soon as i do parallel
requests i get this exception:
ALLVARLIG: error dispatching message:
java.lang.RuntimeException: Invalid message will be discarded:
ActiveMQObjectMessage {commandId = 7, responseRequired = true, messageId =
ID:artimus-3
897-1161358090985-3:11:1:1:1, originalDestination = null,
originalTransactionId = null, producerId =
ID:artimus-3897-1161358090985-3:11:1:1, destinati
on = topic://ericsson.ConsumerTopic, transactionId = null, expiration =
1161358135891, timestamp = 1161358105891, arrival = 0, correlationId = null,
r
eplyTo = null, persistent = true, type = null, priority = 4, groupID = null,
groupSequence = 0, targetConsumerId = null, compressed = false, userID =
null, content = [EMAIL PROTECTED],
marshalledProperties = [EMAIL PROTECTED],
dataStructure =
null, redeliveryCounter = 0, size = 0, properties = null,
readOnlyProperties = true, readOnlyBody = true}
at
org.logicblaze.lingo.jms.marshall.DefaultMarshaller.onInvalidMessage(DefaultMarshaller.java:148)
at
org.logicblaze.lingo.jms.marshall.DefaultMarshaller.readRemoteInvocation(DefaultMarshaller.java:92)
at
org.logicblaze.lingo.jms.JmsServiceExporterMessageListener.onMessage(JmsServiceExporterMessageListener.java:99)
at
org.jencks.LocalTransactionEndpoint.onMessage(LocalTransactionEndpoint.java:68)
at
org.apache.activemq.ra.MessageEndpointProxy$MessageEndpointAlive.onMessage(MessageEndpointProxy.java:121)
at
org.apache.activemq.ra.MessageEndpointProxy.onMessage(MessageEndpointProxy.java:60)
at org.apache.activemq.ActiveMQSession.run(ActiveMQSession.java:693)
at
org.apache.activemq.ra.ServerSessionImpl.run(ServerSessionImpl.java:163)
at
org.apache.geronimo.connector.work.WorkerContext.run(WorkerContext.java:291)
at
EDU.oswego.cs.dl.util.concurrent.PooledExecutor$Worker.run(Unknown Source)
at java.lang.Thread.run(Thread.java:595)
I browsed the lingo source and found this:
public RemoteInvocation readRemoteInvocation(Message message) throws
JMSException {
handleInvocationHeaders(message);
if (message instanceof ObjectMessage) {
ObjectMessage objectMessage = (ObjectMessage) message;
Object body = objectMessage.getObject();
if (body instanceof RemoteInvocation) {
return (RemoteInvocation) body;
}
}
return onInvalidMessage(message);
}
So it seems that the message i send is either not a ObjectMessage or the
body is not a RemoteInvocation object. But it is a bit strange, since this
works when doing single shot requests.
I connect using on the client using:
vm://localhost?marshal=true&broker.useJmx=false&broker.persistent=false
Any ideas why this is happening?
cheers,
-Kristoffer
gnodet wrote:
>
> The vm transport works only inside one classloader,
> so you need to put these in a shared classloader
> for you scenario to work.
>
> On 10/20/06, kristoffer <[EMAIL PROTECTED]> wrote:
>>
>> I did some test for this but my initial idea did not work (maybe for
>> obvious
>> reasons).
>>
>> This is how i understand what is happening:
>> Seems as if the service modules are not able to connect to the broker
>> because of classloader visibility since they are in separate WARs (this
>> is
>> how vm:// works?). These service modules then instead try to create their
>> own embeded brokers. The brokers will throw exceptions all over the place
>> with conflicting journals.
>>
>> However, i managed to get this to work, by putting activemq (4.0.2) and
>> most
>> of its dependencies (spring,xbeans etc) in system classpath. All service
>> modules seem to connect to _the_ broker and are able to communicate
>> between
>> eachother properly.
>>
>> Do i really have to restrict myself and put these jars in system
>> classpath
>> to get this to work? What am I doing wrong? (Is vm:// is infact a
>> classloader:// transport?)
>>
>> One idea I have is to make the (unwanted) creation of brokers (in the
>> service modules) to be non-persistent by providing some configuration in
>> order to avoid the journal conflicts. Would this work?
>>
>> I dont want to have a monolithic design where i put all my modules in a
>> single WAR. I also want to avoid putting things in system classpath when
>> rolling out the product in production.
>>
>> cheers,
>> -Kristoffer
>>
>>
>> kristoffer wrote:
>> >
>> > Hi,
>> >
>> > Im designing a modular system inside a servlet container were i want
>> > modules (deployed as WARs, thus separte classloaders) to communicate
>> with
>> > eachother through JMS. In order to do this effeciently I was thinking
>> of
>> > using Jencks to utilize JCA for this nice pooling, transaction and
>> > workmanager support. I dont want embedded brokers in these service
>> > modules, if it is possible?
>> >
>> > Can I deploy a broker as a separate WAR module (and classloader), which
>> > exposes a vm:// transport to the service modules? Reading from the
>> > documentation it seems as if this would work, as long as i send
>> > ObjectMessage messages between the service modules. Is this correct?
>> >
>> > I was also thinking of using Lingo so that i can have message driven
>> pojos
>> > interfaces between the service modules. Does lingo always use
>> > ObjectMessage to communicate with the broker, or is there some way i
>> can
>> > force this behaviour?
>> >
>> > One "problem" i realized (not really related to this) is that since the
>> > broker WAR need to be deployed before the service modules i need some
>> > mechanism of setting deployment dependencies between the broker and
>> > service modules. Possibly between service modules aswell. I know
>> weblogic
>> > and Jboss have a loadorder feature, but have yet to find one in for
>> > example jetty or tomcat? Anyone have a tip?
>> >
>> > As a future requirement i want to be able to scale up with multiple
>> > servlet containers that should share load between eachother, both on
>> HTTP
>> > request level (with a loadbalancer), but also between brokers.
>> >
>> > Please correct me if i missunderstood how to use activemq in the best
>> way.
>> >
>> > keep up the good work!
>> >
>> > cheers,
>> > -Kristoffer
>> >
>>
>> --
>> View this message in context:
>> http://www.nabble.com/Jecks%2C-Lingo-and-vm%3A---tf2479865.html#a6917416
>> Sent from the ActiveMQ - User mailing list archive at Nabble.com.
>>
>>
>
>
> --
> Cheers,
> Guillaume Nodet
>
>
--
View this message in context:
http://www.nabble.com/Jecks%2C-Lingo-and-vm%3A---tf2479865.html#a6919304
Sent from the ActiveMQ - User mailing list archive at Nabble.com.
