The first ten minutes
A new server should be brought to a known state before anything is deployed to it: apply available updates, create an unprivileged user, install your SSH key, disable password authentication, and close every port that is not needed.
Doing this by hand once is fine. Doing it by hand repeatedly is how environments drift apart, which is why the same steps are usually captured in a script or image.
Make it reproducible
The useful question about any server is how quickly you could rebuild it. If the answer involves remembering what was configured, the configuration belongs in version control instead.
Reproducible servers also make scaling and recovery ordinary events rather than emergencies.
Monitor from the start
Basic monitoring — processor, memory, disk, and whether the service answers — costs very little to set up and is the difference between noticing a problem and being told about it by a user.


