4 Replies Latest reply on Jul 22, 2010 3:45 AM by black_ice

    *Edited* Maybe a bug with filtered queues?

    black_ice

      Hello.

       

       

       

      Im using HornetQ 2.1.1 integrated on AS 5.1. I have 7queues. But i've got a problem with my two last value queues.

      They are setted up to block and rest of values are on default (except connectionfactory timeouts increased, its default*3).

      There is no paging involved, and jboss has -Xmx property on 1.5gb of memory. (uses around 800mb according to windows task manager).

      Also there is no heavy cpu/memory cpu usage on server. Its just that my process becomes slow when i use last value queues (work fine on other).

       

       

      Ther issue, but a normal single queue, but filter messages via Type='0' and Type='1'. Cpu usage goes 99% on the server and it becomes unresponsive (old jboss messaging wont go over 10%). It keeps 99% usage even after empying queues and killing all clients (bug?). I can avoid it using two queues instead of filtering because of two types.

       

       

       

       

       

      PC Information:

       

      Virtual machine with one Xeon's 2'4ghz "single core" and 2.5gb reserved ram. on SCSI disk, i dont know much more about hardware. Using remote desktop. Network... its a corporate network, but all the virtual-machines (database, clients and queue's server) are in the same real machine.

       

       

       

       

       

      Thanks for your help, and excuse my "name", thought nickname was showing.

        • 1. Re: Help with last value queue in 2.1.1 delaying a lot.
          ataylor

          Could you provide a test program so we can recreate the issue.

          1 of 1 people found this helpful
          • 2. Re: Help with last value queue in 2.1.1 delaying a lot.
            black_ice

            I think that the last value queue problem got solved, i think it was because of excess paging on other queue, ill delete that part for now.

             

            Problem with 99% cpu usage filtered queue its, while having activity on the other 3-4 queues, if one producer sends ObjectMessage containing objects with around ~40 integers/long and 2-3 dates per object. With a StringProperty (or integer) which its Type=0 and Type=1. You setup two consumers one for type=0 and other for type=1. That makes the server go crazy and 99% cpu usage. Even after finishing all the producers/consumers on the client (and they close sessions and stop consuming). Its still consuming 99% cpu until i restart jboss AS 5.1.

             

             

             

            Exact sample code cant be given, its really big and im not allowed to.

            • 3. Re: Help with last value queue in 2.1.1 delaying a lot.
              timfox

              asdsad qweqwewq wrote:


              Problem with 99% cpu usage filtered queue its, while having activity on the other 3-4 queues, if one producer sends ObjectMessage containing objects with around ~40 integers/long and 2-3 dates per object. With a StringProperty (or integer) which its Type=0 and Type=1. You setup two consumers one for type=0 and other for type=1. That makes the server go crazy and 99% cpu usage. Even after finishing all the producers/consumers on the client (and they close sessions and stop consuming). Its still consuming 99% cpu until i restart jboss AS 5.1.

               


              You should read the last post I made on this forum.

              • 4. Re: Help with last value queue in 2.1.1 delaying a lot.
                black_ice

                Oh ok, thank you both for reading and helping =).

                 

                Will wait for that fix.