Question

At this link http://en.wikipedia.org/wiki/Tracing_(software) they point out differences between server logs and trace logs. As a developer, I have always been sufficed by server logs and never needed trace logs. What situations require looking into trace logs?

Answer 1

As @bkail mentions, WebSphere Application Server’s built-in server tracing is typically for IBM support. It is generally too fine-grained and tightly coupled with IBM’s closed source code to be of use to customers.

However, there are also uses of the trace logs for application support as well. If your application utilizes java.util.logging, these log events will be written to WAS’s log files (e.g. SystemOut.log, trace.log). The log messages written to SystemOut.log (Level.CONFIG and higher) are typically intended for system administrators. Log messages written to trace.log (Level.FINE and lower), on the other hand, are messages that are typically intended for developers or troubleshooting and debugging purposes; these messages may be tightly coupled with the code or contain extensive diagnostic information useful in troubleshooting situations. Generally, you only want to enable tracing during troubleshooting or development, as this type of extensive logging can be expensive and potentially impact the performance of your applications.

As a developer, you should make a first-class distinction between logging intended for your system administrators and logging (tracing) intended for developers or troubleshooting. Logging is a great method for communicating with system administrators and can be an invaluable for troubleshooting, but each of these use cases should be handled differently. This is one of the primary reasons that logging APIs (including java.util.logging) provide multiple logging levels. The article you referenced seems to do a great job distinguishing between logging and tracing (which translates to SystemOut.log and trace.log in WAS). IBM’s documentation also provides a good overview of the differences.