My own experience is that logging is essential, even in production. We
log everything - SQL commands, debug logs we've inserted... everything.
The problem is that when a customer writes in and says "I can't login
to your web site", without logs you are left with telling them "clear
your cache and try again". I can't even count the number of times a
screwed-up browser has sent us total nonsense and confused our scripts.
In a perfect world, everything would be bullet-proof before making it
available to customers, but I think the very fast cycle times for
web development in smaller companies make that impossible.
Our server logs (not access logs - server logs) are 90MB a day, and we
have even gone so far as to log the start and end of every hit before,
without a noticeable effect on performance.
Jim
>
> The thought never occurred to us that logging on a production server would
> be so extensive that it would require such filtering by severity. Usually
> production servers should only spew out the most interesting messages:
> Warning, Error and Fatal. If you're routinely logging on a production
> server with Notice you might want to reconsider the idea of development vs.
> production and how you report your server's activity. Just a thought.
>
> Kris
>
