weblogic performance pbm (production) - URGENT - Weblogic

This is a discussion on weblogic performance pbm (production) - URGENT - Weblogic ; hi all i have one serious pbm n production side. i am using weblogic 81 spc 3, websphere mq 5.3 with latest patch. i have 8 mdb and each mdb Listening jms queues.Database connection pool size is 300, Thread count ...

+ Reply to Thread
Results 1 to 4 of 4

Thread: weblogic performance pbm (production) - URGENT

  1. weblogic performance pbm (production) - URGENT

    hi all

    i have one serious pbm n production side.
    i am using weblogic 81 spc 3, websphere mq 5.3 with latest patch. i have 8 mdb and each mdb Listening jms queues.Database connection pool size is 300, Thread count is 100 and every ejb bean have initial and max bean in pool is 250.

    i am always monitoring server performance thr weblogic admin console. when the thread count (Idle Threads) reach to 80 around, my client branches not able to access my application and database connection also reach to max(300).

    some time i got java OutOfMemory exception.

    kindly give me the solution to solve this problem
    every time i need to restart production server when this pbm arise.

    thanx n advance
    regards
    muruganandam

  2. Re: weblogic performance pbm (production) - URGENT

    Sounds like you have two problems - OutOfMemory and a possible JDBC connection leak.

    What are the JVM arguments you're using?
    What is the hardware and OS configuration?
    Do you have any BMP beans?
    For the connection leak problem, set ConnectionLeakProfiling to true in teh JDBC connection pool configuration. That should spit out stack traces of connection allocation points if there is a connection leak. Note that turning this flag on will degrade performance.
    If you have 80 idle threads, your thread pool size and MDB pool size (I assume the 250 is the MDB pool size?) is way too large. If you store a lot of data in your MDBs and this data is set either in the ejbCreate or setMessageDrivenContext methods, that could explain the OutOfMemory.
    The fact that you eventually run out of memory could also explain why the server becomes unresponsive - the JVM is probably stuck GC'ing.

  3. Re: weblogic performance pbm (production) - URGENT

    hi

    JVM :
    MEM_ARGS="-Xms128m -Xmx256m -XX:MaxPermSize=128m"
    H/w and OS is
    sun v880 -2x900, 4 GB RAM (both Application Server and Database Server). OS is SunSolaris 9.

    i have 17 (stateless session bean managed) out of 81 session bean and 9 MDB.

    for connection leaking , bea support suggested to put one jar file n weblogic classpath (CR189281_81sp3.2.jar).

    and i reduced MDB MAx bean pool size to 150 and thread count is 75.

    still i am getting production server slow down.

    plz help me to solve this problem.

    regards
    muruganandam

  4. Re: weblogic performance pbm (production) - URGENT

    Your heap settings look small to me. I'd suggest increasing both the ms and mx settings to 512 MB, assuming WLS is the only application running on the machine. Turn on verbosegc and check how much gc is occurring. If you need help analyzing the verbosegc output, post the section around the OutOfMemoryError or from around the time the slowdown occurs, here. I assume the DB is running on a different machine? It's not a good idea to run both on the same machine.

    Also, you still seem to have way too many threads for a 2-CPU box. It's likely that you're loading the machine much more than it's capacity would allow. I'd suggest drastically reducing the mdb, slsb pool sizes and the thread count and increase by small increments until you get the best performance.

    Regarding the JDBC connection leak problem, has it gone away after applying the patch? If not, turn on ConnectionLeakProfiling as mentioned in my previous post.

+ Reply to Thread