Decide who responds
Agree ownership for each monitored system in advance. An unowned alert is a notification, not a response.
How it works
The loop is short on purpose. Every step that a monitoring service adds between a failure and a response is a step where the failure gets more expensive.
The loop
You should be able to get from nothing to genuinely watched in one sitting, and you should be able to explain the whole arrangement to a colleague in a sentence.
Start with the websites and servers whose failure you'd want to hear about at 3am. Not everything you run — the things people actually depend on.
BinaryCanary becomes the layer watching those systems from the outside, independently of the infrastructure it's watching.
When something changes, the point is that you find out before your customers do — while the problem is still small enough to be uninteresting.
A signal has done its job when the next action is obvious. Someone owns it, they know what broke, and they get on with it.
Why outside-in
“If your monitoring lives on the box that just went down, you don't have monitoring — you have a coincidence.”
Checking from outside your own infrastructure is what makes the answer trustworthy. It is also why a monitoring service is worth buying rather than building on the side of another job.
What comes after the alert
Monitoring tells you something is wrong. What turns that into a short incident rather than a long one is groundwork you do while nothing is on fire.
Agree ownership for each monitored system in advance. An unowned alert is a notification, not a response.
Check the signal, then say something — internally and, where it affects them, to customers. Silence costs more than an imperfect update.
Every incident tells you something about your coverage. The useful question afterwards is what would have told you sooner.
Next step
14 days free, then from $2 a month. Already have an account? Log in.