5 Techniques to Improve Your Server Logging | Takipi Blog
4. Uncaught handlers
Westeros has the Wall as its last line of defense. You have Thread.uncaughtExceptionHandler. So make sure you use them. If you don't install one of these handlers, you run the risk of throwing exceptions into the wild with very little context, and very little control of if and where they end up being logged.
Finding out that your code has been encountering massive errors without them being logged, or with them being logged with little to no state data is an epic fail.
Notice that even within an uncaught exception handler, which on the surface of things has no access to any variable within that thread (which has already terminated) you still get a reference to the actual Thread object. If you adhered to step #1, you'll still get a meaningful thread.getName() value which you can log.
5. Catch external calls
Whenever you make an external API call which leaves the JVM, chances of an exception increase dramatically. This includes Web service, HTTP, DB, File system, OS and any other JNI calls. Treat each call as it will explode (it most likely will at some point).
For the most part, the reasons for external API failures have to do with unexpected input provided to them. Having those inputs readily available for you in the log is a key part in fixing the code.
You may choose not to log an error at this point, but have the exception thrown, which is fine. In that case, simply collect as many of the relevant arguments passed to that call and parse them into the exception error message.
Read full article from 5 Techniques to Improve Your Server Logging | Takipi Blog
No comments:
Post a Comment