On 18/06/2010, João Tiago Ferreira <[email protected]> wrote: > Hi. I dont think there are so many use cases. > > First point is the sender of the message can't set the messageId. It is an > identifier of the message generated by the JMS provider.
I know. > Moreover two different messages can't have the same messageID, so I would say > comparing "request:messageId == reply:messageId" makes no sense. It *can* make sense in "echo" mode: The sampler can send to a queue and then receive the original message from the same queue. Could be useful for checking queue throughput. In that case, *if* the correlationID is present of course one could use that instead. > Here is a nice explanation of the different JMS request response patterns > [1]. Does not work for me - I get an empty page. > So when you check the option "Use Request Message Id As Correlation Id" we > are using the "JMS Message ID Pattern" where no correlationId is necessary to > set in the request and in the receiver we should check for request:messageId > == reply:correlationId instead of "request:messageId == reply:messageId" That relies on the response correlation ID being set to the request message ID - which does not happen automatically, as far as I can tell. > When you uncheck the "Use Request Message Id As Correlation Id" we are using > the "JMS Correlation ID Pattern" and is necessary to set the correlatioId in > the request and in the receiver we should check for "request:correlationId == > reply:correlationId" as is implemented now. The service has to copy the correlationId from request to response. Your use cases both rely on checking the response correlationId which it is assumed will be set from either the request correlationId or the requestMessageId. The person who raised the original Bug 46142 reported that the correlationId was null for JBoss, and the cited message thread says that the same problem happens with Weblogic. If the JMeter receiver only looks at the response correlationId, it seems this will cause problems for JBoss and Weblogic. So I think it is necessary to make the request and response fields independently selectable. I don't think on can satisfy everyone without. I've already done much of the work for this; it seems to work OK for ActiveMQ in "echo" mode at least. > Note: when I say "receiver" I mean the client side receiver, i.e the Jmeter > receiver... > > Thanks for support > > João Ferreira > > > [1] > http://download.oracle.com/docs/cd/E13159_01/osb/docs10gr3/interopjms/MsgIDPatternforJMS.html > --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
