mod_weblogic

Well it had to happen eventually -- welcome to Weblogic Server :-)

Thought I'd walk through how to configure Apache as the front end server for a Weblogic Server.

Turns out, it's as easy as the proverbial:

1. Install WLS, create a domain, start it up
   
2. Download and install Apache (I used 2.2)
   
3. Copy the corresponding Apache module (for 2.2 I used mod_wl_22.so) from the WLS installation into $APACHE_HOME/modules directory
   
4. Edit $APACHE_HOME/conf/httpd.conf and enable the mod_weblogic as follows:
   
   # Load BEA Module
   LoadModule weblogic_module modules/mod_wl_22.so
   
   # Config file for WebLogic Server that defines the parameters
   Include conf/mod_weblogic.conf
   
   
5. Create the referenced mod_weblogic.conf file in which you place the configuration elements for mod_weblogic to use:
   
   # Basic config for mod_weblogic to point at my WLS instance
   
   WebLogicHost localhost
   WebLogicPort 7001
   Debug ON
   WLLogFile c:\temp\wlproxy.log
   WLTempDir c:\temp
   DebugConfigInfo On
   
   
   # Map the URLs to expose via Apache
   # Expose the console via Apache
   
   SetHandler weblogic-handler
   
   
   
6. Test the config with Apache
   
   $APACHE_HOME\bin\httpd -t
   
   
7. Start Apache
8. Access the console via Apache (assuming all default ports)
   
   http://localhost:8080/console
   

And that's the basic set of steps -- you should be looking at the WLS console being serviced via Apache.

With the config above, I've enabled the DebugConfigInfo option -- which allows you to see a bunch of useful debug/usage information by entering a URL serviced by mod_weblogic and tacking on a request parameter of the form ?__WebLogicBridgeConfig.

Classloader Loader Event Tracing

Classloader Event Tracing

If you are seeing classloader closed errors from time to time, then one really useful utility OC4J has is the ability to log the various events related to different aspects of the classloading mechanism.

So for example, lets say you are seeing errors occasionally containing this type of message: "This loader has been closed and should not be in use."

Then what you can do to help trace why this is occurring is to enable log tracing for the various loader events.

In the error message above, it should also explicitly provide the name of the loader that was being used, of the form appname.modulename[.module]:major.minor. For example myapp.webmodule.root:0.0

To observe classloading events relating to this specific loader, you configure OC4J to start as follows:

>java -Dclass.load.trace=loader:myapp.webmodule.* -Dclass.load.trace.level=all -jar oc4j.jar


That setup configures OC4J to output all the loader related classloading events, at the log level of ALL, where the loader name matches the pattern "myapp.webmodule.*".

In the output, you'll see entries such as:

ClassLoader created: myapp.webmodule.root:0.0.0 (instance 1cba87)
ClassLoader committed: myapp.webmodule.root:0.0.0 (instance 1cba87)
...
ClassLoader closing: myapp.webmodule..jsp13471351:0.0.0 (instance 14441b6)
...

That shows you where the loaders are being operated on, and confirms (one way or another) that the loader is actually being closed -- this may correlate with other actions you have performs -- say a redeployment operation.

Now to help further understand why/where the loader event occured, you can add an extra event filter. The "stack" event filter adds the associated stacktrace that caused the loader event to occur to the output -- thus for the loader closed events, you can trace back and identify the source of the loader close.

To add the stack event filter to the classloader tracing, simply concatentate it to the class.load.trace property:
>java -Dclass.load.trace=stack+loader:myapp.webmodule.* -Dclass.load.trace.level=all -jar oc4j.jar

That then causes a stacktrace to be appended to each event that is logged:

ClassLoader closing: myapp.webmodule:0.0.0 (instance 18aea9e)
at oracle.classloader.util.ClassLoadTracing$LoaderListener.loaderClosing(ClassLoadTracing.java:738)
at oracle.classloader.EventDispatcher.loaderClosing(EventDispatcher.java:232)
at oracle.classloader.PolicyClassLoader.close(PolicyClassLoader.java:1056)
at oracle.classloader.PolicyClassLoader.close(PolicyClassLoader.java:1041)
at com.evermind.server.http.HttpApplication.destroy(HttpApplication.java:5860)
at com.evermind.server.http.HttpApplication.stopCleanUp(HttpApplication.java:8437)
at com.evermind.server.http.HttpApplication.componentStop(HttpApplication.java:8383)
at com.evermind.server.Application.doStop(Application.java:2092)
at com.evermind.server.Application.stop(Application.java:2019)
at oracle.oc4j.admin.management.mbeans.J2EEStateManageableObjectBase.stop(J2EEStateManageableObjectBase.java:275)
at oracle.oc4j.admin.management.mbeans.J2EEApplication.stop(J2EEApplication.java:1542)
at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:39)
at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:25)
at java.lang.reflect.Method.invoke(Method.java:585)
at sun.reflect.misc.Trampoline.invoke(MethodUtil.java:36)
at sun.reflect.GeneratedMethodAccessor17.invoke(Unknown Source)
at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:25)
at java.lang.reflect.Method.invoke(Method.java:585)
at sun.reflect.misc.MethodUtil.invoke(MethodUtil.java:239)
at javax.management.modelmbean.RequiredModelMBean.invokeMethod(RequiredModelMBean.java:1071)
at javax.management.modelmbean.RequiredModelMBean.invoke(RequiredModelMBean.java:953)
at oracle.oc4j.admin.jmx.server.mbeans.model.DefaultModelMBeanImpl.invoke(DefaultModelMBeanImpl.java:700)
at com.sun.jmx.mbeanserver.DynamicMetaDataImpl.invoke(DynamicMetaDataImpl.java:213)
at com.sun.jmx.mbeanserver.MetaDataImpl.invoke(MetaDataImpl.java:220)
at com.sun.jmx.interceptor.DefaultMBeanServerInterceptor.invoke(DefaultMBeanServerInterceptor.java:815)
at com.sun.jmx.mbeanserver.JmxMBeanServer.invoke(JmxMBeanServer.java:784)
at oracle.oc4j.admin.jmx.server.state.ApplicationStateFilterMBeanServer.invoke(ApplicationStateFilterMBeanServer.java:572)
at oracle.oc4j.admin.jmx.shared.UserMBeanServer.invoke(UserMBeanServer.java:634)
at oracle.oc4j.admin.jmx.server.state.LocalizationFilterMBeanServer.invoke(LocalizationFilterMBeanServer.java:579)
at oracle.sysman.ias.studio.jmx.spi.JMXConnectorImpl.invoke(JMXConnectorImpl.java:389)
at oracle.sysman.ias.studio.oc4j.JMXAppAdminBean.stopApp(JMXAppAdminBean.java:276)
at oracle.sysman.ias.studio.oc4j.AppHelper.doStopAppOp(AppHelper.java:776)
at oracle.sysman.ias.studio.oc4j.AppHelper.access$600(AppHelper.java:116)
at oracle.sysman.ias.studio.oc4j.AppHelper$AppControlProcessingBean.appControlOperation(AppHelper.java:1628)
at oracle.sysman.ias.studio.oc4j.AppHelper$AppControlProcessingBean.doWork(AppHelper.java:1642)
at oracle.sysman.ias.studio.util.ProcessingBean$ProcessingThread.run(ProcessingBean.java:273)
at com.evermind.util.ReleasableResourcePooledExecutor$MyWorker.run(ReleasableResourcePooledExecutor.java:298)
at java.lang.Thread.run(Thread.java:595)


Looking at the stacktrace for the loader closing event, you can reasonably surmise that the event was caused by the application being "stopped" from ascontrol (sysman).

Now that's a contrived example, I agree! But I hope it serves in some small way to demonstrate how you can use the classloader tracing mechanism to get a handle on the different events that happen to loaders and why they have occurred.

FAQ for OC4J 10.1.3. Management

A few questions I see from time to time in the OTN OC4J Forum are covered in the OC4J 10.1.3 Management/Admin FAQ

http://www.oracle.com/technology/tech/java/oc4j/1013/whitepapers/OC4J-FAQ-MGT-10131.pdf

Creating Users in OC4J Groups

Another relatively common question:

The ascontrol group page only allows creation/configuration of JMS and JDBC resources at the group level. Is it possible to create users across all the OC4J instances in Group from ascontrol?

Yes this can be done.

The solution is to use the Cluster MBean Browser provided in ascontrol and execute an operation on the J2EEApplication MBean via the desired J2EEServerGroup MBean.

Luckily its documented in the OC4J Security Guide, so I don't need to butcher the description any more!

http://download.oracle.com/docs/cd/B32110_01/web.1013/b28957/configxml.htm#sthref316

Starting OC4J from Ant

Good question today from a colleague in Dublin, who basically wanted to know how to start an OC4J instance from Ant using the oc4j:restart.

On the surface that looks like a reasonable thing to do. The problem though is that the OC4J Ant tasks actually require a connection to be made to the server in order to execute the operation -- so if the OC4J instance is down, you can't issue a restart since you can't connect to it.

For situations where the OC4J instance is being managed by OPMN, then one possible solution to this is to call out to the opmnctl executable and ask it to start the OC4J instance.

To do this, you can make use of the ant:exec task as shown below:

<project name="opmn" basedir="." default="status">
  <property name="oracle.home" value="d:/oracleas/10131/ohs_j2ee"/>
  <property name="opmn.cmd" value="${oracle.home}/opmn/bin/opmnctl.exe"/>
  
  <target name="base" description="Verify callout to opmnctl">
    <exec executable="${opmn.cmd}"/>
  </target>

  <target name="status" description="Show the current status">
    <exec executable="${opmn.cmd}">
      <arg line="status"/>
    </exec>
  </target>

  <!-- Start the oc4j.name instance-->
  <target name="start-oc4j" description="Start the oc4j.name instance" depends="get-oc4j-name">
    <echo message="${opmn.cmd} startproc process-type=${oc4j.name}"/>
    <exec executable="${opmn.cmd}">
      <arg line="startproc process-type=${oc4j.name}"/>
    </exec>
    <antcall target="status"/>
  </target>

  <target name="get-oc4j-name" unless="oc4j.name">
    <input message="Specify OC4J instance: " addproperty="oc4j.name"/>  
  </target>
</project>

When this executes, it prompts for the OC4J instance name, calls opmnctl to execute a startproc on it and then displays the status.

>ant -f opmn.xml start-oc4j
Buildfile: opmn.xml

get-oc4j-name:
    [input] Specify OC4J instance:
fred

start-oc4j:
     [echo] d:/oracleas/10131/ohs_j2ee/opmn/bin/opmnctl.exe startproc process-type=fred
     [exec] opmnctl: starting opmn managed processes...

status:

     [exec] Processes in Instance: ohs_j2ee_changed
     [exec] ---------------------------------+--------------------+---------+---------
     [exec] ias-component                    | process-type       |     pid | status
     [exec] ---------------------------------+--------------------+---------+---------
     [exec] OC4JGroup:default_group          | OC4J:foo           |    2128 | Alive
     [exec] OC4JGroup:default_group          | OC4J:fred          |    6012 | Alive
     [exec] OC4JGroup:default_group          | OC4J:home          |     N/A | Down

Application Supplied System Properties

We added a feature in the 10.1.3.1 release which enables an application to supply its own set of System properties, which are added and removed to the global System property set as the application is started and stopped. The documentation for it is not extensive, but there's a how-to on OTN which walks through the use of it.

Check it out on OTN.

The System properties (name, value) are defined in the orion-application.xml descriptor provided with the application.

<orion-application>
  <property name="application.tax.rate" value="10" />
  <property name="application.tax.code" value="AU" />
</orion-application>

During deployment with ASC, the deployment plan editor can be used to view the set of specified System properties in the application. It can also be used to add, remove, edit the defined properties so they can be tailored to the specific deployment being performed.

Run a Coherence Cache Server from JDeveloper

I was working on a Coherence based handson lab for the upcoming Open World conference last week. The lab configures Coherence to use a JPA based CacheStore to store/load cache entries into/from a database.

The labs are all based around an Ant script that runs the lab code, plus the Cache Servers. But one handy thing that could have been documented in the labs is the ability to run a Cache Server(s) directly from JDeveloper using its "Run Manager".

Here's how it works:

1. Have a project set up which has the Coherence JAR files added as a library.

2. Create a new Run Configuration for your Cache Server.

3. In the "Default Run Target" field enter:

C:\coherence-ejb3\coherence\lib\tangosol.jar!\com\tangosol\net\DefaultCacheServer.class

4. Specify any additional JVM or Coherence properties as desired.

Click save and you're done.



Now to run a Cache Server, simply select your Cache Server run target from the list and the Cache Server will start. A log window will open up showing the output from the Cache Server. The process running the Cache Server will be shown in the Process Manager window.



The other Run Configuration I found helpful was to launch the Console that comes with Coherence. This little application is a cache client, and allows you to interactively perform operations against the cache. Handy to quickly see the size of a cache or list out some elements from the cache to see that its containing the expected data objects.

To run this Coherence application, you create a new Run Configuration as above, except specifying the Default Run Target as:

C:\coherence-ejb3\coherence\lib\coherence.jar!\com\tangosol\coherence\component\application\Console.class



Since the Coherence console is an interactive application, one extra little step you need to do here is to enable "Allow Program Input" for this running process.

In the Run Configuration dialog, select the "Tool Settings" node and then check the "Allow Program Input" option.



Save the entry and execute it. You should then the Coherence Console application fire up and prompt for your command in the "Input: [ ]" field.

Locating the process that has a specific network port open

Ever tried to start a process on Windows, only to be told that the network port it needs is unavailable?

TNSLSNR for 32-bit Windows: Version 10.2.0.1.0 - Production
Error listening on: (DESCRIPTION=(ADDRESS=(PROTOCOL=TCP)(HOST=mypc.oracle.com)(PORT=1521)))
TNS-12542: TNS:address already in use

Bugger. Now, which process is using that port? The task manager doesn't show you any of these sorts of details.

There's two ways I know of to work out which process which already has the port open.

1. If you have cygwin or MKS or some Unix styled set of command utilities -and who using Windows commands shells wouldn't? -- then you can use the grep and ps commands it provides with the standard netstat command.

First off, use the netstat command with the -ao switch and pipe it into grep to locate the listening details. The -o switch tells netstat to output the process-id of the owner of the port.

D:\>netstat -ao | grep 1521
TCP mypc:1521 mypc:0 LISTENING 2944

Then using the ps command, search for the process-id:

D:\>ps -eaf | grep 2944
SYSTEM 2944 1488 0 11:37:44 CONIN$ 0:00 vmware-authd

This could be combined into a nice awk script I suspect.

2. Use the very handy SysInternals utilities to locate the owning process. These used to be available on the sysinternals.com website, but the site now redirects to Microsoft, so presumably they have purchased them. Good idea Microsoft.

The specific utility for this operation is TCPView:

http://www.microsoft.com/technet/sysinternals/utilities/TcpView.mspx

This comes in two flavours, GUI and cmd line and identifies each process that is using a network port. The GUI is easy to use, just fire it up, look for the port in question which then identifies the process in the process table.

I prefer using the command line utility, tcpvcon.exe.

This results in a nicely formatted text output that shows the process and network port details for all processes that have a network port in use in some manner.

D:\>tcpvcon.exe -a
...
[TCP] vmware-authd.exe
PID: 2944
State: ESTABLISHED
Local: mypc:1521
Remote: mypc:0
...

To make it easier to identify a specific process which owns a specific port, then you can ask the output to be provided in CSV format, which lists each process on a single line, and which you can then pipe into find to locate a specific port

D:\>tcpvcon.exe -c -a | find "1521"
TCP,VMNET.EXE,2944,LISTENING,mypc:1521,mypc:0

----------------
Listening to: John Butler Trio - Good Excuse

OC4J 10.1.3.3 Ant Taskname Error

OC4J 10.1.3.x supplies a set of Ant tasks that enable operations such as application deployment, resource configuration and server lifecycle tasks to be performed.

The abridged history of the tasks is:

• OC4J 10.1.3.0 introduced the basic tasks
  
• OC4J 10.1.3.1 and later versions extends the set of tasks to support JDBC/JMS resource configuration options and server shutdown/restart tasks

The OC4J Configuration and Administration guide documents the Ant tasks.

However in the OC4J 10.1.3.3 release several of the tasks are incorrectly defined in the antlib.xml and therefore not available using the names as described in the documentation.

The documentation is correct and shows how the tasks should be named. However if you use these tasks as documented you will get errors. The screen capture below highlights the specific differences.



The workarounds are either:

1. Revert to using the names as defined in antlib.xml for these tasks:

• addDataSourceConnectionPool == createJDBCConnectionPool
• addManagedDataSource == createManagedDataSource
• addNativeDataSource == createNativeDataSource
• testDataSourceConnectionPool == testConnectionPool

2. Update the /j2ee/utilities/ant-oracle-classes.jar!/oracle/antlib.xml to change the names of the tasks so they reflect the correct names as specified in the documentation:

Change: taskdef name="createJDBCConnectionPool"
To: taskdef name="addDataSourceConnectionPool"

Change: taskdef name="createManagedDataSource"
To: taskdef name="addManagedDataSource"

Change: taskdef name="addManagedDataSource"
To: taskdef name="createNativeDataSource"

Change: taskdef name="testConnectionPool"
To: taskdef name="testDataSourceConnectionPool"

A bug has been filed for this and a patch will be released which fixes the issue along the lines of option 2 shown above.

Bring on UST + Stans

My Turner Flux is shod with a Mavic Crossmax SL wheelset which is UST compatible. I've been riding it with tubes + Maxxis Crossmarks (which I think are awesome) so far since I've been a bit slow to pick up some UST tires -- I'm commerced out I think. Plus I also figured with all the other knobs and dials to tweak, I had enough new stuff to worry about.

But last Friday some new Maxxis Crossmax LUSTs I'd purchased arrived. I was hoping to put them on over the but I didn't quite get a chance to do.

And as fate or that little Irish Murphy bloke would have it, my early Monday morning blast around the Eagle MTB park resulted in a pinch flat just at the end of the SouthSide trail.

I didn't have a spare tube, but thought I was lucky that I had some patches. But of course that little Murphy bloke, smiling ever more cunningly at this stage, somehow conspired to make the CO2 cartridge I carry to refill a tube be empty.

Bugger, a 2km walk, mostly uphill back to the car.

Yes, it's my fault, I know. Always carry a spare tube and a full canister. And if you use the spare tube, make sure you buy another one to replace it. And check the canister occasionally.

The good thing is this means I now have a hard requirement to put on the USTs. I picked up some Stans sealant from Pete at BMCR, so hopefully I won't be suffering too many more punctures.

But I'll will carry a spare tube just in case.

Update: Job done!

I just spent an short while out in the shed fitting the UST tires to the wheels. It's pretty straightforward to do. I think the best piece of advice I received was to do the initial inflation using an air compressor, to get enough volume into the tire to push the beads into place in the rim. The beauty of the Mavic wheelset I discovered is that they provide you with a schraeder screw-on adapter, which lets you use the standard car tire type fitting to pump out the tire. Once the tire is seated, deflate it and make sure there are no points around the rim where the bead has come away. Once you are satisfied pump it back up to the desired using a floor pump. Easy!

Adding Stans was easy too -- just read the direction and make sure you follow the inversion tips to ensure the micro particles are in the sealant portion being added to the tire. Unset the bead from a small portion of the tire, tip in the sealant, reseat and you are done.

Total time was about 30 mins for the first tire, with a bit of mucking around and working out how to do it. The next tire took about 5 minutes to do.

Should be able to get out for a few trails this weekend, I'll update this note when I see how they ride.

Update: Stans 1, Blue Gums 0

Out on my early monday morning ride, cruising across the top of the Blue Gums trail -- which is quite rocky -- I heard a noise that I thought sounded like I had something stuck in my spokes. It went away before I could stop to check it out. When I did stop, I saw a white patch on my tire. Ahh, Stans to the rescue. I must have punctured on a sharp rock and Stans sealed it up almost immediately. I lost a little pressure, but it was still very rideable and got me back to the car easily. Sure beats walking. Pumped it back up at the car and there was no further leakage.

Rock on Stan!