System Crashes

If you are reporting a system crash, try to collect the following information for Support:

• Stack traces. When Java crashes, a stack trace is sometimes printed to Standard Error. If a stack trace is available, forward the trace to Oracle Support.

• Java core files. If the JVM dumps a core file, the file might be useful to Oracle Support or Sustaining. Although core files are large, collect and store them in a safe place so you can deliver them to Support upon request.

• Memory issues. If your system is running out of memory, try to gather the following information for Support before your program crashes:

  • Heap dump. You can automate heap dump generation by adding the -XX:+HeapDumpOnOutOfMemoryError option to the application server's JVM options. You can also force a heap dump on a running process as described in Hung Systems.

  • Histogram. A histogram is a list of class objects that are occupying memory for the Java application.

    You must collect histograms at set intervals over a period of time to track memory growth over time. For example, starting a reconcile consumes all system memory in ten hours, so collecting a histogram every hour results in ten histograms, or ten points of data from which to track memory growth. Save the output from each histogram to a different file, with an appropriate timestamp, and deliver this information to Support.

    To collect histograms for Support, type jmap -histo pid in a terminal window. Where pid is the Process ID of the JVM running Waveset (usually the same as application server).

• Exception tracing. Support can frequently use an exception trace to determine the cause of a fatal error. However, Waveset's exception tracing is disabled by default. To collect a useful exception trace for Support, you must enable exception tracing and then coerce Waveset into crashing again in the same way.

  To enable exception tracing, use the instructions in Tracing Exceptions.