Unless you think you can predict every single bug that you'll ever write, and also how it will affect your customers' experience, how do you know you can track it at all?
And if you're not logging it, once you do figure out that 50% of your customers are leaving because your payment endpoint throws an exception on Amex cards, how do you know what's causing that? Now it's just a silly detective game for something you could have been just logging in the first place.
And if you're only able to find out via your customer metrics, it's already too late. If you were just logging the exception, you'd have said "huh that's funny" when you deployed the code, rather than when you have enough data to realise that you've already lost the customers.
> If you were just logging the exception, you'd have said "huh that's funny" when you deployed the code, rather than when you have enough data to realise that you've already lost the customers.
I'm assuming that you still have the local logs on the local server and you're looking at one instance as you deploy to find those exceptions.
I'm also assuming that we're talking about a system too complex to actually watch logs go by in real time.
If the log level is light enough that you can actually read the logs as you deploy, then by all means do so. But if you're aggregating them just to be able to search them because you can't read them as they go by, then I'm saying there are better things to be doing than storing all those logs.
And if you're not logging it, once you do figure out that 50% of your customers are leaving because your payment endpoint throws an exception on Amex cards, how do you know what's causing that? Now it's just a silly detective game for something you could have been just logging in the first place.
And if you're only able to find out via your customer metrics, it's already too late. If you were just logging the exception, you'd have said "huh that's funny" when you deployed the code, rather than when you have enough data to realise that you've already lost the customers.
Personally I'd rather just have the logs.