Trapping/Catching/Logging/Handling Errors in production

[scrupul0us]

I know that the scope of this question is rather large and encompassing, but, I was wondering what you folks tend to use in your bigger production projects to handle errors, logging and so forth...

I'm not so much worried about the mechanics of the storage (db, flatfile, etc) but rather your implementation

Is it wise convention to add reporting to every instance in which you are doing something

e.g.

if you can/cant connect to the db, if you can/cant write a record, if a page isnt found, if an if/else tree fails to the last else...

things of that nature

and at that, do you have something setup such that these items are considered logging at a different level so you can be as detailed or not in your reporting

thanks

[bpat1434]

If you logged everything that happened during the course of a request, your log files would be enormous!

I tend to like to only log those items where I know failure presents a bad user experience (i.e. not a mis-entered form datum). Only then will I typically log it, send an email, and let the user know I know about it now.

I've seen some really interesting error handlers. Like taking a user notice and making that completely stop your script just because it used trigger_error. Then I've seen some really bad ones where the errors just got shown on the screen in a production environment.

It really depends on how pro-active you want to be with your support. Most of the clients I work with expect a few hour turn-around, even at 3 am.

[Weedpacket]

Trap it
Catch it
Log it
Handle it

[Shrike]

When you are asked to debug why "it doesn't work" or "the page didn't load" you can never, ever have enough information. Log everything.