JBoss 3.2.1 Linux 2.4.20/Mandrake 9.1 Sun J2SDK 1.4.1_01 600Mhz Pentium 3 (gateway) Oracle 8.1 (for Entity beans, JMS and Quartz scheduler)
This is perhaps a bit premature as I have not managed to isolate my problem yet but I'd like to ask if anyone has seen anything like this before. We have a J2EE application that is primarily message driven. It consists of an MDB that listens for external events, on receipt it creates a new Entity bean instance and then publishes another event that is received by a second MDB. This bean uses a quartz scheduler to publish yet another event at some time now or in the future. And it goes on, there are five MDBs in all and the events sort of travel in a loop around them until reaching the last bean that publishes a result to an external topic. It's not all that complex really and in our functional testing it has been working quite well. Then I got into some load testing. It takes about 2.5 minutes for a round trip and we get a through put of about 3 events/minute, which is most encouraging as there is quite a bit of database work done inbetween. Then I noticed that out of 25 incoming external events I was getting one or two lost on input and about another 3-4 would not complete the round trip, being list somewhere between the input MDB and the output MDB. The way they are lost is disconcerting. The first MDB does a JAAS login as the first operation in the onMessage() method and then processes the event and creates a new entity instance. in about 2 in 25 times this fails due the the thread not being logged in! How? It just logged in and the login did not fail (and it worked for the other 23 events) Further on, events just seem to vanish. Detailed tracing shows two events arriving at an MDB at about the same time, the MDB container tracing shows both being read by the container apparently before entry to the bean and subsequent tracing in our MDB code shows processing of the second event. Now the really worrying bit is that I retested on a Solaris 8 machine (4*400MHz processors, sun j2sdk 1.4.1_01) and the events are not lost, at least not so far, I haven't tested it enough to be sure. This may be due to having 4 processors or maybe there is a nasty bug in the Linux JVM? So, any suggestions would be most welcome. I'm currently writing a much simpler test case to see If I can reproduce the behaviour. thanks, brian wallis... ------------------------------------------------------- This SF.Net email sponsored by: ApacheCon 2003, 16-19 November in Las Vegas. Learn firsthand the latest developments in Apache, PHP, Perl, XML, Java, MySQL, WebDAV, and more! http://www.apachecon.com/ _______________________________________________ JBoss-user mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/jboss-user
