2 Replies Latest reply on Aug 7, 2004 4:12 AM by josefine

    Problems with ClientMonitorInterceptor?

    josefine Newbie

      Hi,

      we have a problem with client subscriptions to topic.

      sometimes a client get an out of memory or something like that, so he can't fetch messages anymore; seems that he get "stuck". this again will leads to remaining topic queue and constrains the other clients as well.

      we have tried the ClientMonitorInterceptor and set the timeout to different values. we hope that this will kill the "stuck" subsciber and cleans up the topic queue.

      unfortunatly this doesn't work steadily. sometimes it seems that it works the way we excpected. but most of the time it doesn't.

      it seems that our "stuck" client sends further pings so the clientmonitorinterceptor doesn't realise that the client get "stuck".

      does anybody can give us some hints?

      best regards

      josefine

        • 1. Re: Problems with ClientMonitorInterceptor?
          Elias Ross Master


          If you can identify the client by thread ID, and get a stack trace, post the stack trace of the stuck client.

          Also, always post your JBoss version, OS, and if you're using a client, the client factory you're using.

          • 2. Re: Problems with ClientMonitorInterceptor?
            josefine Newbie

            Hi,

            sorry for forgetting these important informations!

            we are having jboss 3.0.8 and windows 2000 server on server side and windows2000 workstations on client side. we are using the oil invocation layer and a temporary topic with non durable subscription. for the java parts we are using j2sdk-1.3.1_10.

            simplified our application is getting data from a oracle9i database using jboss and then sending these datas (also jboss) to the java clients. the amount of connected clients can reach up to 80.

            so if one client get stuck it can occurs that this client will affect all other clients in performance etc. at the end the jboss must be restarted.

            at the moment we can't provide a stack trace. but we will post it, as soon as we get it.

            any help will be very, very appreciated!

            best regards.

            josefine