Debugging High Latency Issues

The latency of particular service depends on the complexity of service implementation, payload, workload, the number of services deployed on an AppNode, and CPU or memory resources made available to the AppNode.

1. Capture the process execution statistics for the test run. This would help analyze the time spent in individual processes and activities. For more information about enabling statistics on deployed service, see the TIBCO ActiveMatrix BusinessWorks™ Administration guide.
2. Capture thread dumps for analyzing the thread state and calls. Capture 5 thread dumps at an interval of 5 seconds each. Redirect all the collected threads dumps to separate files.
   • The thread dumps can be captured using jstack utility shipped with JDK.

     ./jstack <PID of appnode> > ThreadDump_n.txt

   • The thread dumps can be captured through JConsole or JVisualVM utilities.
3. Capture top CPU consuming threads data for 5 minutes run by using jvmtop utility.
   For more information about jvmtop utility, see support article KB000034702.

   ./jvmtop <PID of appnode> > JVM_topthreads.txt
   

4. Capture method level CPU profiling data for 5 minutes run by using the jvmtop utility.

   ./jvmtop --profile <PID of appnode> > JVM_CPUProfile.txt

5. Capture AppNode logs for the test run duration.
6. Capture the CPU and the memory utilization data for 5 runs on the AppNode.
   • The data can be captured through the top utility on Unix.

     top -p <PID of appnode> > top_appnode.txt
     

   • The data can also be captured through JConsole or JVisualVM utilities.
7. Capture the AppNode config.ini and the TRA file of the AppNode. It helps analyze the JVM parameters, and other engine tuning parameters.
8. Capture system configurations of servers such as CPU details, RAM, and number of cores where ActiveMatrix BusinessWorks™ AppNode, external services, and load generator are running. Capture details of /proc/meminfo and /proc/cpuinfo files.

   cat /proc/meminfo and cat /proc/cpuinfo