Question

Well…I do know it’s good to log errors when the website is in production and error reporting is off. But what should I actually log? I can’t log all errors, so I should to choose some most errors-probable places in the code and call the function with some message to log this. Should I use try-catch to do this? And is it actually the best way to do this?

Answer 1

You should have several levels of error severity, all of which are handled differently:

• debug level messages are just for debugging, e.g. very verbosely outputting information about everything that’s happening
• info level messages inform about important events, like “trying to contact web service”
• notices are not unusual, but noteworthy events
• warnings are problems which need to be corrected, but do not cause immediate problems
• errors should be logged in detail and monitored
• critical errors are unexpected emergencies and should be logged in detail and possibly emailed automatically
• alerts require immediate attention and should trigger automated SMS notifications
• emergencies should automatically have the national guard deploy

You intersperse your app with all these different types of errors/events. Notices and up should be logged, errors and up should be logged in detail (dump the system status into a separate file) and for critical errors and up somebody should be automatically notified.

See https://github.com/Seldaek/monolog for a good component which can help you do this.