A common question we get is centered around removing a backend from a Fastly rotation for maintenance, or some other temporary reason.
If you’re planning on putting the backend back into service in a short amount of time, the best option is to set a failing health check. When this happens, Varnish pulls the backend out of rotation until the health check returns to normal.
So, if you’re planning for future work, I’d recommend pointing your health check to an obscure URL: something like
my.example.backend.com/path-to-check.html, and setting up a 200 response on your backends for this path. That way, when you’re ready to pull a machine out of rotation, you only need to adjust the
path-to-check.html path to return a response other than 200. Varnish will to the rest, and you won’t have to adjust your configuration at all.