Two is one, one is none

One encryption key went missing and every scheduled job stopped for 55 hours. The only spare copy turned up in a backup file that should never have been in the code repository. The leak saved them.

That is what one of everything looks like. One person who can deploy. One copy of the key. One alert channel nobody owns. One server a customer depends on. Each of them fine on a normal day.

The old rule is two is one and one is none. I would go further for anything your revenue runs through. Two or three systems that can do the job. Two or three people who can run them. None of them will fail today. The day one does, you want the question to be "which backup," never "who do we call."

-Christopher