How to do error handling in Windows .NET Services

Question

This feels like a question that has probably been asked before but I couldn't find it so am asking (possibly again).

I'm writing a windows service to do a bit of file monitoring and processing. Its all doing what I want but I am aware that there are several places that my code might throw errors when doing read/writes to the file system or event log. I'm goign to put in some try/catches where I expect there to be errors but I was wanting to do some global error handling to pick up all those unforseen problems. I'm not sure how to do this though.

My first thought was to look for some kind of global error handler similar to what I find in ASP.NET projects but couldn't see anything along these lines.

The other thought was that I could just get all my top level function calls and put a try/catch around them. As I see it that would mean putting a try/catch around all my event handler code and my service Start and Stop methods. This doesn't feel like a good way of doing it so I'm thinking I'm probably approaching this in the wrong way. Can anybody suggest what I should be doing in terms of error handling?

I assume that in a service I don't want it to just throw errors because it will potentially stop the service or should I in fact be letting the service collapse and automatically restarting it through teh recovery mechanism I've seen mentioned?

I'm very uncertain here so please share your wisdom with me... :)

Answer 1

There is no problem in putting a try/catch in the main method of the service and log the errors so you can fix them. If the error caught is unexpected then you may close the service cleanly.

Also, besides the main try/catch, you will have to add a try/catch for every potential exceptions that you cannot avoid to happen (i.e. network issues) so you can act accordingly in order to restore the correct behavior of the application, avoiding the service to crash.

Finally, keep in mind that you will have to do this for every top-method of every distinct thread you have, because the exceptions are not propagated outside the scope of a thread.

Answer 2

We've used this successfully in many Windows services we've written over the years.

We always have the functionality in a dll. In the service when we instantiate and run the function in the dll we wrap the call in a try/catch. Then the service can log or respond to any unforseen exceptions that are thrown.

You would, of course, wrap suspect code in the dll in try/catch blocks and do what is appropriate there as well.

Answer 3

A service should never fail. Usually things in a service are triggered by an event (network call or timer)... The code that runs for each event should be in a try catch, such that the service never goes down.

I suggest logging to the EventLog for Errors and other crucial information. You could use log4net + the EventLogAppender for doing that.

It can still happen, that the service goes down because of hardware problems like full disks or memory issues. If you want it to restart automatically you can set the recovery in the NT Service Options Dialog or in the Installer.