Just curious, why Push the guid onto the NDC and Pop it off before the call to “Proceed” and then Push/Pop afterwards? I have not used NLog much, but I assume that, having pushed something onto the NDC, that makes it available to be logged via a layout renderer for each message that is logged while the guid is on the stack. So, when your messages

log.Info(“{0}() called”, methodName) and log.Info(“{0}() completed”, methodName)

are logged, they will (or can be, depending on your logging format layout) each be tagged with the guid that is specific to this instance of this call.

If you leave the guid in the NDC, any logging inside of the method can also be tagged with the guid.

Also, I think that if you want to leave the guid in the NDC, you can do something like this (Push returns an IDisposable that does a Pop in its Dispose method):

var log = LogManager.GetCurrentClassLogger();

using (NLog.NDC.Push(Guid.NewGuid())
{
log.Info(“Entering blah blah”);

// snip some code here

try
{
eventArgs.Proceed();
}
catch (Exception ex)
{
//do stuff in case of exception
}

NLog.MDC.Set(“Elapsed time …”);

log.Info(“Completed blah blah …”);
}

Thus, the using (NLog.NDC.Push(Guid.NewGuid())) allows the Pop to happen implicitly and also allows all statements inside of the “real” method (eventArgs.Proceed()) to be tagged with the same correlation id. Of course, if each subsequent layer of the call hierarchy is instrumented similarly, then each new Guid would be pushed onto the stack and, again, the logging statements from a specific level in the hierarchy

Maybe that is not what you were after and you really just want the enter/exit log messages to be tagged with the correlation id.

Anyway, thought I would mention this to see if you have any thoughts on this approach (leaving guid “Push”-ed for whole method call rather than Push/Pop before and Push/Pop after).

Thanks again for the nice AOP/NLog example.