NAME

Java::SJ - Highly configurable Java program startup system

SYNOPSIS

  sj myprogram.sj

DESCRIPTION

This module allows you to very easily run Java services that rely on complex configuration at the VM and library level. It also provides an easy way of specifying a sensible 'default' configuration that can be overridden by specific applications should they need to.

The system is configured on a machine and application level. The system looks for configuration files in a set of well-known locations, currently these are:

  • /etc/sj.conf

  • .sj.conf in users HOME directory

  • .sj.conf in current working directory

Every application is defined in terms of a similar configuration file. The configuration system has been designed so that it is easy to write a simple and minimal configuration file for a program.

Provided the system has a fairly complete configuration associated with it then an application configuration file need only have the class name to be executed.

FEATURES

Some of the Goodness(tm) that you get with SJ is as follows:

Easy co-existence of multiple Virtual Machines

Any number of VMs can be supported and used concurrently on the same machine. Developers don't need to know where the JDK/JRE resides, just a symbolic name for it.

Easy co-existence of multiple versions of JAR files

Any number of different versions of the same JAR file may co-exist. SJ sorts out which ones to use and only places those JAR files that are required in an application's CLASSPATH

Control over BOOTCLASSPATH variables

All three flavours of the bootclasspath can be configured on a system-wide and application specific basis.

Process control

PID files can be automatically generated and placed wherever you wish.

Cache of executable scripts

Application configuration files are cached as executable scripts that can be directly invoked.

This will probably make more sense as a set of examples, so here goes.

EXAMPLES

Simple Configuration

        <?xml version="1.0"?>
        <sj>
          <!-- define a virtual machine -->
          <vm name="ibm" 
              home="/usr/local/IBMJava-1.3.1" 
              default="true"/>
        </sj>

The above minimal system configuration simply tells the system where to find a virtual machine called 'ibm' and that it should be used as the default VM unless an application specifically requests another one.

        <?xml version="1.0"?>
        <sj>
          <!-- define the class file to run -->
          <class>myclass</class>
        </sj>

The above minimal application configuration simply provides a class file name to run. If both of the above simple configuration files were used then the class 'myclass' would have to be in the system's CLASSPATH environment variable already.

Useful Configuration

The above simple configuration is only really useful to test that the system is working. Running a simple class on the command line isn't normally very difficult and so sj doesn't actually add very much to the above case.

If we had a system where we wished to test that the same code was compatible with multiple virtual machines and multiple library versions then the following configuration files would enable us to run these programs with different parameters easily.

        <?xml version="1.0"?>
        <sj>
          <!-- important locations -->
          <var name="dir.base" value="/usr/local"/>
          <var name="dir.pid" value="${dir.base}/var/run"/>
          <var name="dir.log" value="${dir.base}/var/log/sj"/>
          <var name="dir.tmp" value="/tmp"/>
          <var name="dir.lib" value="${dir.base}/lib/sj"/>
          <var name="dir.script" value="${dir.base}/var/sj/script"/>

          <!-- write out a PID file for every program -->
          <pid/>        

          <!-- use this as our default CLASSPATH -->
          <classpath>
            <jar name="xalan"/>
            <jar name="xerces"/>
            <jar name="xml-apis"/>
            <jar name="commons-cli" version="1.0.3"/>
          </classpath>

          <vm name="ibm118" 
              vendor="IBM" 
              version="1.1.8" 
              home="/usr/local/IBMJava-1.1.8"/>

          <vm name="ibm141" 
              vendor="IBM" 
              version="1.4.1" 
              home="/usr/local/IBMJava-1.4.1"/>

          <vm name="blackdown118" 
              vendor="Blackdown" 
              version="1.1.8" 
              home="/usr/local/blackdown-1_1_8"/>

          <vm name="sun131" 
              vendor="Sun Microsystems" 
              version="1.3.1" 
              home="/usr/local/sunjdk_131"
              default="true"/>
        </sj>

The above system configuration contains a lot more information than our initial simple example.

Variables

There are explicitly declared variables that SJ will use when figuring out where things should be read from/written to.

Variables may be defined in terms of other variables, even if they have not been declared yet. The syntax for referring to variables thorughout is the same as Ant, ${variable_name}.

PID file

It states that by default a PID file should be written when running an application. This is most useful for multithreaded server applications where you want to be able to kill or HUP the server without figuring out what the lead process PID is.

Classpath

The classpath definition instructs SJ to look for the latest available versions of xalan, xerces and xml-apis and version 1.0.3 of the commons-cli libraries and add these to the classpath when running any program.

SJ will look for the libraries in the path defined by ${dir.lib}.

SJ is currently very simplistic about library versioning. If it needs to look for a specific version of a library then it simply looks for the library name and library version number joined by a single hyphen. So in the case of the commons-cli library SJ would look for commons-cli-1.0.3.jar in the ${dir.lib} directory.

You may be wondering how SJ figures out which is the 'latest' version of a JAR file if no version is specified. Quite simply it chooses the one whose filename begins with the required name and is lexicographically last in an ordered list of those filenames that match. It's not great but with any half-sensible version numbering scheme it will work.

Virtual Machines

There are four virtual machines defined here. In addition to the home directory for each there is now information regarding the vendor and version. Currently this is for informational purposes only but future versions of SJ should be able to choose a VM based on the version or vendor.

        <?xml version="1.0"?>
        <sj>
          <!-- define the class file to run -->
          <class>myclass</class>
          <vm ref="blackdown118"/>
        </sj>

        <?xml version="1.0"?>
        <sj>
          <!-- define the class file to run -->
          <class>myclass</class>
          <vm ref="ibm118"/>
        </sj>

        <?xml version="1.0"?>
        <sj>
        <!-- define the class file to run -->
          <class>myclass</class>
          <vm ref="sun131"/>
        </sj>

The above application configurations are exactly the same as the simple version with the addition of a vm reference tag to determine which VM they should be executed under.

High Granularity

In addition to being able to specify which VMs and libraries to use you have complete control over the environment that the VM is run under, the properties that are passed the the VM and even default command line options on a per system, VM and application basis.

        <?xml version="1.0"?>
        <sj>
          <!-- important locations -->
          <var name="dir.base" value="/usr/local"/>
          <var name="dir.pid" value="${dir.base}/var/run"/>
          <var name="dir.log" value="${dir.base}/var/log/sj"/>
          <var name="dir.tmp" value="/tmp"/>
          <var name="dir.lib" value="${dir.base}/lib/sj"/>
          <var name="dir.script" value="${dir.base}/var/sj/script"/>

          <!-- add these as -Dname=val VM properties-->
          <property name="ORBSingletonClass" value="jacorb.ORBSingleton"/>

          <!-- add these to the environment for every app -->
          <environment name="TZ" value="CET"/>
          <environment name="PAGER" value="less"/>

          <!-- add these to the command line parameters of every app -->
          <param name="--debuglevel" value="3"/>
          <param name="--colour" value="blue" sep="="/>
          <param name="-g" value="3" sep=":"/>

          <!-- write out a PID file for every program -->
          <pid/>        

          <!-- use this as our default CLASSPATH -->
          <classpath>
            <jar name="xalan"/>
            <jar name="xerces"/>
            <jar name="xml-apis"/>
            <jar name="commons-cli" version="1.0.3"/>
          </classpath>

          <vm name="ibm118" 
              vendor="IBM" 
              version="1.1.8" 
              home="/usr/local/IBMJava-1.1.8">
            <!-- set USE_JIT whenever this VM is chosen -->
            <environment name="USE_JIT" value="true"/>
          </vm>

          <vm name="ibm141" 
              vendor="IBM" 
              version="1.4.1" 
              home="/usr/local/IBMJava-1.4.1"/>

          <vm name="blackdown118" 
              vendor="Blackdown" 
              version="1.1.8" 
              home="/usr/local/blackdown-1_1_8">
            <!-- set this parameter for this VM only -->
            <param name="-Xmx" value="81920k" sep=""/>
          </vm>

          <vm name="sun131" 
              vendor="Sun Microsystems" 
              version="1.3.1" 
              home="/usr/local/sunjdk_131"
              default="true"/>
        </sj>

The above system configuration is identical to our useful configuration except we have now added directives that SJ will use to alter the environment and command line parameters passed to the application and VMs.

Using VM specific parameters you can make sure that the correct threading models are used or that memory limuts are enforced unless someone needs to tweak the settings.

In an application configuration file it is possible to override previously declared parameters such as the -Xmx directive above for the blackdown VM.

For example:

        <?xml version="1.0"?>
        <sj>
          <!-- define the class file to run -->
          <class>myclass</class>
          <vm ref="blackdown118">
                <param name="-Xmx" value="80m" sep=""/>
          </vm>
        </sj>

The Java::SJ::Config documentation describes every configuration directive in detail, also have a look in the sample directory for ideas.

TODO

Test, test, test.

BUGS

None known so far. Please report any and all to Nigel Rantor <wiggly@wiggly.org>

SUPPORT / WARRANTY

This module is free software. IT COMES WITHOUT WARRANTY OF ANY KIND.

LICENSE

The Java::SJ module is Copyright (c) 2003 Nigel Rantor. England. All rights reserved.

You may distribute under the terms of either the GNU General Public License or the Artistic License, as specified in the Perl README file.

AUTHORS

Nigel Rantor <wiggly@wiggly.org>

SEE ALSO

Java::SJ::Config.