Details

    • Type: Bug Bug
    • Status: Closed Closed (View Workflow)
    • Priority: Major Major
    • Resolution: Done
    • Affects Version/s: JBossAS-3.2.7 Final
    • Fix Version/s: JBossAS-3.2.7 Final
    • Component/s: Test Suite
    • Security Level: Public (Everyone can see)
    • Labels:
      None
    • Environment:
    • Similar Issues:
      Show 10 results 

      Description

      When running the jboss-all-config-tests I have been seeing OutOfMemoryErrors well into the tests, but -verbose:gc clearly shows that there is plenty of memory. The process views also show that the handle/thread count is fine. That pretty much only left the MaxPermSize as becoming full, and I was able to run the tests using:

      JAVA_OPTS="-verbose:gc -XX:MaxPermSize=70m -Xmx128m"

      This seems like an excessive amount of perm heap memory so I would like to look into this further. This does not happen when running JDK 1.4.2.

      1. CorrectAnalysis-jvm14.permanentCapacity.txt
        54 kB
        Clebert Suconic
      2. jboss-all-config-tests_gc.log
        23 kB
        Scott Stark
      3. jboss-testsuite-analysis.txt
        25 kB
        Clebert Suconic

        Activity

        Hide
        Scott Stark
        added a comment -

        I ran the jboss-all-config-tests using jdk1.5.0_01:

        [starksm@banshee9100 testsuite]$ ant -Dnojars=t jboss-all-config-tests
        ...

        and tracked its permanent space using jstat. The jboss-all-config-tests_gc.log attachment is the result,
        and I see the PU go from 19782.2 at the start to 43921.7 at the end. So the permanent space is definitely growing, but its not approaching the levels apparently needed by jdk1.3.1.

        Show
        Scott Stark
        added a comment - I ran the jboss-all-config-tests using jdk1.5.0_01: [starksm@banshee9100 testsuite] $ ant -Dnojars=t jboss-all-config-tests ... and tracked its permanent space using jstat. The jboss-all-config-tests_gc.log attachment is the result, and I see the PU go from 19782.2 at the start to 43921.7 at the end. So the permanent space is definitely growing, but its not approaching the levels apparently needed by jdk1.3.1.
        Hide
        Clebert Suconic
        added a comment -

        Due to a problem during the execution of the testcases I was getting wrong results.

        Here is a more accurate analysis for JVM1.4.2

        Note:

        I marked where the PC (Permanent capacity) had the most significant changes I could notice. Here is a synchronization between the JUnit test which was running and the PC measure:

        Running org.jboss.test.jmsra.test.RaSyncRecUnitTestCase First hit to 37M
        Running org.jboss.test.jmx.test.JMXInvokerProxyUnitTestCase 39M

        (decreased to 35M after a while)

        Running org.jboss.test.lock.test.SpinUnitTestCase - 36M
        Running org.jboss.test.mdb.test.MDBUnitTestCase - 37M
        Running org.objectweb.jtests.jms.conform.topic.TemporaryTopicTest = 39M
        Running org.jboss.test.security.test.EJBSpecUnitTestCase = 40M
        Running org.jboss.test.security.test.LoginModulesUnitTestCase = 41M

        (Decreased to 38912 after a while)

        Running org.jboss.test.jbossmx.compliance.standard.InfoTortureTestCa = 40448

        Show
        Clebert Suconic
        added a comment - Due to a problem during the execution of the testcases I was getting wrong results. Here is a more accurate analysis for JVM1.4.2 Note: I marked where the PC (Permanent capacity) had the most significant changes I could notice. Here is a synchronization between the JUnit test which was running and the PC measure: Running org.jboss.test.jmsra.test.RaSyncRecUnitTestCase First hit to 37M Running org.jboss.test.jmx.test.JMXInvokerProxyUnitTestCase 39M (decreased to 35M after a while) Running org.jboss.test.lock.test.SpinUnitTestCase - 36M Running org.jboss.test.mdb.test.MDBUnitTestCase - 37M Running org.objectweb.jtests.jms.conform.topic.TemporaryTopicTest = 39M Running org.jboss.test.security.test.EJBSpecUnitTestCase = 40M Running org.jboss.test.security.test.LoginModulesUnitTestCase = 41M (Decreased to 38912 after a while) Running org.jboss.test.jbossmx.compliance.standard.InfoTortureTestCa = 40448
        Hide
        Clebert Suconic
        added a comment -

        For generating CorrectAnalysis-jvm14.permanentCapacity.txt I've used this line command:

        jstat -gcpermcapacity 3224 5s > CorrectAnalysis-jvm14.permanentCapacity.txt

        Note that JBoss was running into JVM 1.4.2 with the Hotspot instrumentation active (-XX:+UsePerfData according to http://java.sun.com/performance/jvmstat/)

        I used jstat provided by JDK 5.0 against the process without any problems.

        Show
        Clebert Suconic
        added a comment - For generating CorrectAnalysis-jvm14.permanentCapacity.txt I've used this line command: jstat -gcpermcapacity 3224 5s > CorrectAnalysis-jvm14.permanentCapacity.txt Note that JBoss was running into JVM 1.4.2 with the Hotspot instrumentation active (-XX:+UsePerfData according to http://java.sun.com/performance/jvmstat/ ) I used jstat provided by JDK 5.0 against the process without any problems.
        Hide
        Clebert Suconic
        added a comment -

        For generating CorrectAnalysis-jvm14.permanentCapacity.txt I've used this line command:

        jstat -gcpermcapacity 3224 5s > CorrectAnalysis-jvm14.permanentCapacity.txt

        Note that JBoss was running into JVM 1.4.2 with the Hotspot instrumentation active (-XX:+UsePerfData according to http://java.sun.com/performance/jvmstat/)

        I used jstat provided by JDK 5.0 against the process without any problems.

        Show
        Clebert Suconic
        added a comment - For generating CorrectAnalysis-jvm14.permanentCapacity.txt I've used this line command: jstat -gcpermcapacity 3224 5s > CorrectAnalysis-jvm14.permanentCapacity.txt Note that JBoss was running into JVM 1.4.2 with the Hotspot instrumentation active (-XX:+UsePerfData according to http://java.sun.com/performance/jvmstat/ ) I used jstat provided by JDK 5.0 against the process without any problems.
        Hide
        Clebert Suconic
        added a comment -

        For JDK 1.3.1 we will have to add the parameter -XX:MaxPermSize=70m as it requires more PermSize than the 1.4 versions.

        Show
        Clebert Suconic
        added a comment - For JDK 1.3.1 we will have to add the parameter -XX:MaxPermSize=70m as it requires more PermSize than the 1.4 versions.
        Hide
        Dimitris Andreadis
        added a comment -

        Closing resolved issues.

        Show
        Dimitris Andreadis
        added a comment - Closing resolved issues.

          People

          • Assignee:
            Clebert Suconic
            Reporter:
            Scott Stark
          • Votes:
            0 Vote for this issue
            Watchers:
            0 Start watching this issue

            Dates

            • Created:
              Updated:
              Resolved:

              Time Tracking

              Estimated:
              Original Estimate - Not Specified
              Not Specified
              Remaining:
              Remaining Estimate - 0 minutes
              0m
              Logged:
              Time Spent - 1 day, 3 hours
              1d 3h