1. Axis2
  2. AXIS2-4629

Poor performance in 1.5.1 due to threads waiting on monitor for synchronized jre method


    • Type: Bug Bug
    • Status: Resolved
    • Priority: Critical Critical
    • Resolution: Duplicate
    • Affects Version/s: 1.5.1
    • Fix Version/s: None
    • Component/s: adb
    • Labels:
    • Environment:
      RedHat 5.4, Solaris 10


      We recently upgraded from axis2 1.4.1to 1.5.1 and with no changes to our java methods which are exposed using axis we have seen a significant drop in performance from around 500-600 TPS to 50 TPS at most. We took a jstack while load is on our system and can see alot of threads blocked waiting for an object monitor to enter the private method java.beans.Introspector.getPublicDeclaredMethods in the JRE which is called by axis - stack trace is below.

      "http-9080-7" daemon prio=10 tid=0x0a95d800 nid=0x2f1d waiting for monitor entry [0xdce76000]
      java.lang.Thread.State: BLOCKED (on object monitor)
      at java.beans.Introspector.getPublicDeclaredMethods(Unknown Source)

      • waiting to lock <0xf0882fa0> (a java.lang.Class for java.beans.Introspector)
        at java.beans.Introspector.getTargetPropertyInfo(Unknown Source)
        at java.beans.Introspector.getBeanInfo(Unknown Source)
        at java.beans.Introspector.getBeanInfo(Unknown Source)
        at java.beans.Introspector.getBeanInfo(Unknown Source)
        at org.apache.axis2.databinding.utils.BeanUtil.getPropertyQnameList(
        at org.apache.axis2.databinding.utils.BeanUtil.getPullParser(
        at org.apache.axis2.rpc.receivers.RPCUtil.processResponse(
        at org.apache.axis2.rpc.receivers.RPCUtil.processResponseAsDocLitWrapped(
        at org.apache.axis2.rpc.receivers.RPCMessageReceiver.invokeBusinessLogic(
        at org.apache.axis2.receivers.AbstractInOutMessageReceiver.invokeBusinessLogic(
        at org.apache.axis2.receivers.AbstractMessageReceiver.receive(
        at org.apache.axis2.engine.AxisEngine.receive(
        at org.apache.axis2.transport.http.HTTPTransportUtils.processHTTPPostRequest(
        at org.apache.axis2.transport.http.AxisServlet.doPost(
        at javax.servlet.http.HttpServlet.service(
        at javax.servlet.http.HttpServlet.service(
        at org.apache.catalina.core.ApplicationFilterChain.internalDoFilter(
        at org.apache.catalina.core.ApplicationFilterChain.doFilter(
        at org.apache.catalina.core.StandardWrapperValve.invoke(
        at org.apache.catalina.core.StandardContextValve.invoke(
        at org.apache.catalina.core.StandardHostValve.invoke(
        at org.apache.catalina.valves.ErrorReportValve.invoke(
        at org.apache.catalina.core.StandardEngineValve.invoke(
        at org.apache.catalina.connector.CoyoteAdapter.service(
        at org.apache.coyote.http11.Http11Processor.process(
        at org.apache.coyote.http11.Http11Protocol$Http11ConnectionHandler.process(
        at Source)

      From looking at the stack trace it seems like this is called when generating every response hence the performance hit and having compared the 1.4.1 and 1.5.1 source we can see that the call to the synchronized JRE method is not made in 1.4.1. This issue is critical to our release as we expect this interface to have a high throughput.

        Issue Links


          No work has yet been logged on this issue.


            • Assignee:
              Chris van Es
            • Votes:
              1 Vote for this issue
              3 Start watching this issue


              • Created: