Whats the overhead of creating a Log4j Logger

I have some webservices in my application and i want to log them to diferent files, based on the webservice name. For that i am creating loggers with

myLogger = Logger.getLogger(logKey);

I am wondering if i should cache these loggers to avoid creating them for every call, or can i ignore the overhead.

From stackoverflow

• Loggers are already cached by log4j using the default log repository (Hierarchy). In other words, it's just a hashtable lookup.

  However, in my experience you tend to make the logger static, so it only ends up being called once per class anyway.

  willcodejavaforfood : +1 but it is more fun voting up people with less reputation

  Nicolas : Wouldn't making the logger static prevent you from changing its level at runtime (through an MBean for example)? In theory not, but I thought that was the case with java logging.

  Jon Skeet : I don't *think* so. I think the level stuff is checked dynamically, rather than being frozen in the constructed logger.

• This method Logger.getLogger(logKey) looks in logger cache for a logger with the name passed in logKey. If it doesn't exist it creates one. First call for a logger name, a Logger will be created but later calls will get it from cache so you don't need to handle this in your code.