When you install FusionReactor on an IBM J9 JVM platform (e.g. IBM Java2 SDK 5.0-5.1 for Linux), you may find the CPU graph always reads zero. You may not be able to retrieve stack information on J9 for Java 4 (1.4).
Additionally, you may see the following exception in your servlet container's output:
In this case, please mail support for an updated FusionReactor native library.
You will need to install the new native library over the old one:
- Stop your servlet container
- Copy the special build native library libFusionReactor.so into the folder /opt/FusionReactor/etc/lib, overwriting the copy which is already there (your FusionReactor installation folder may vary)
- Configure your Servlet Container to use this path as the java.library.path
- E.g. for Tomcat, make sure to export CATALINA_OPTS=-Djava.library.path=/opt/FusionReactor/etc/lib
- Restart the container.
Caveat for Java 4 (1.4) J9 JVMs
The FusionReactor native library for IBM J9 does not attempt to coerce the format of stack traces into a parsable format for FusionReactor.
This means that when you select Stack Trace in FusionReactor's Requests or Threads view, you will see an unformatted single dump of all running threads. This dump, while somewhat large, is still useful – simply scroll or search to find the thread name required, then read the frame data as before.
IBM Java 5.0 (1.5) JVMs do not suffer this restriction, since they use Java 5's internal (non-native) tracing functionality to produce colorized output.
|Last Updated:||10/Sep/07 3:02 PM|