Profiling

You can use a profile to perform remote profiling on the Sun Java System Web Server to discover bottlenecks in server-side performance. This section describes how to configure these profilers for use with Sun Java System Web Server:

• The HPROF Profiler

• The Optimizeit Profiler

The HPROF Profiler

HPROF is a simple profiler agent shipped with the Java™ 2 SDK. It is a dynamically linked library that interacts with the JVMPI (Java™ Virtual Machine Profiler Interface) and writes out profiling information either to a file or to a socket in ASCII or binary format. This information can be further processed by a profiler front-end tool such as HAT.

HPROF can present CPU usage, heap allocation statistics, and monitor contention profiles. In addition, it can also report complete heap dumps and states of all of the monitors and threads in the Java virtual machine. For more details on the HPROF profiler, see the JDK documentation at:

http://java.sun.com/j2se/1.4.2/docs/guide/jvmpi/jvmpi.html

Once HPROF is installed using the following instructions, its libraries are loaded into the server process.

To use HPROF profiling on UNIX

1. Configure Sun Java System Web Server in one of these ways:

   • Go to the server instance page in the Administration interface, select the Java tab, click the JVM Profiler link, and edit the following fields before clicking OK:

     • Profiler: Enable

       • Classpath: (leave blank)

       • Native Lib Path: (leave blank)

       • JVM Option: For each of these options, type the option in the JVM Option field, select Add, then check its box in the JVM Options list:

         -Xrunhprof:file=log.txt,options

       Edit the server.xml file:

       <!-- hprof options --> <PROFILER name="hprof" enabled="true"> <JVMOPTIONS> -Xrunhprof:file=log.txt,options </JVMOPTIONS> </PROFILER>

     ---

     Note –

     Do not use the -Xrs flag.

     ---

     Here is an example of options that you can use:

     -Xrunhprof:file=log.txt,thread=y,depth=3

     The file option is important because it determines where the stack dump is written in step 6.

     The syntax of HPROF options is as follows:

     -Xrunhprof[:help]|[:option=value,option2=value2, ...]

     Using help lists options that can be passed to HPROF. The output is as follows:

     Hprof usage: -Xrunhprof[:help]|[:<option>=<value>, ...]
     Option Name and Value Description Default
     --------------------- ----------- -------
     heap=dump|sites|all heap profiling all
     cpu=samples|old CPU usage off
     format=a|b ascii or binary output a
     file=<file> write data to file java.hprof
     (.txt for ascii)
     net=<host>:<port> send data over a socket write to file
     depth=<size> stack trace depth 4
     cutoff=<value> output cutoff point 0.0001
     lineno=y|n line number in traces? y
     thread=y|n thread in traces? n
     doe=y|n dump on exit? y

2. You must also change a line in the Sun Java System Web Server start script. The start script file is instance_dir/start. Change the following line:

   PRODUCT_BIN=webservd-wdog

   to this:

   PRODUCT_BIN=webservd

3. Start the server by running the start script. Since the server runs in the foreground (the change in step 2), the command prompt returns only after the server is stopped.

4. In another window or terminal, find the process ID of the server process.

   % ps -ef | grep webservd

   This command lists two webservd processes. Look at the PPID (parent process ID) column and identify which of the two processes is the parent process and which is the child process. Note the PID (process ID) of the child process ID.

5. Send a SIGQUIT signal (signal 3) to the child process:

   % kill -QUIT child_PID

6. To stop the Web Server, run the stop script from another window.

   % ./stop

   This writes an HPROF stack dump to the file you specified using the file HPROF option in step 1. For general information about using a stack dump, see Generating a Stack Trace for Debugging

7. To return your Web Server to its original configuration, undo the changes in steps 1 and 2.

The Optimizeit Profiler

Information about Optimizeit™ is available at:

http://www.borland.com/us/products/optimizeit/index.html

Once Optimizeit is installed using the following instructions, its libraries are loaded into the server process.

To enable remote profiling with Optimizeit, do one of the following:

• Go to the server instance page in the Administration interface, select the Java tab, click the JVM Profiler link, and edit the following fields before selecting OK:

  • Profiler: Enable

  • Classpath: Optimizeit_dir/lib/optit.jar

  • Native Lib Path: Optimizeit_dir/lib

  • JVM Option: For each of these options, type the option in the JVM Option field, select Add, then check its box in the JVM Options list:

    • -DOPTITHOME=Optimizeit_dir

    • -Xrunoii

    • -Xbootclasspath/a:Optimizeit_dir/lib/oibcp.jar

    Edit the server.xml file:

    <!-- Optimizeit options -->
    <PROFILER classpath="Optimizeit_dir/lib/optit.jar"
        nativelibrarypath="Optimizeit_dir/lib" enabled="true">
        <JVMOPTIONS>
            -DOPTIT_HOME=Optimizeit_dir -Xboundthreads -Xrunoii
            -Xbootclasspath/a:Optimizeit_dir/lib/oibcp.jar
         </JVMOPTIONS>
    </PROFILER>

    In addition, you may need to set the following in your server.policy file:

    grant codeBase "file:Optimizeit_dir/lib/optit.jar" {
    permission java.security.AllPermission;
    };

    For more information about the server.policy file, see The server.policy File

    When the server starts up with this configuration, you can attach the profiler. For further details, see the Optimizeit documentation.

  ---

  Note –

  If any of the configuration options are missing or incorrect, the profiler may experience problems that affect the performance of the Sun Java System Web Server.

  ---