Monitoring from the outside

A service that reports its own health will happily say it is fine while nobody can reach it.

The trap

Checks that run on the server see the server: the process is up, the port is open, the disk has space. None of that says whether a person on another network can use the service. The path between them is the part that fails most often, and it is invisible from the inside.

A better check

Run the check from where the users are and make it do what a user does. For a file service that means downloading a real file, not asking for a status page. A few megabytes are enough to tell a working link from one that accepts the connection and then stalls.

Quiet by default

An alert should mean that something needs attention. If it fires every night while everyone sleeps, people learn to ignore it, and then it is worse than no alert at all.