My monitor says the server is up but the app is broken — why?

My monitor says the server is up but the app is broken — why?

Q: My check shows green, port 443 responds, but the actual app returns errors to users. What's the disconnect?

A: Shallow checks are the trap. A reader monitored only TCP port 443 — was the port accepting connections? It always was, even when the app behind it was returning HTTP 500 to every request. During one incident the app threw errors for 52 minutes while the monitor sat green the whole time.

They deepened the check in stages: from TCP, to an HTTP 200 check, to finally asserting a specific string in the response body ('"status":"ok"' from a health endpoint that pings the database). The next time the DB connection pool exhausted, the body check failed in under a minute.

The follow-up: don't go overboard and check 20 things in one probe — when it fails you won't know which. Use one deep check that proves the critical path works end to end, and keep it readable.

Got a question? Drop it in the comments.