Most outage reports start as a status code somebody did not recognize. 502, 503, a timeout that returns nothing at all. Here is what the codes actually mean and the fastest way to check what a URL is really returning.
The five groups, without the textbook
1xx — hold on. Almost never seen by users. The server got the request headers and is still working. If you see 100-continue in a log, something is mid-handshake.
2xx — it worked. 200 is the only one most people meet. 201 means something was created, 204 means success with an empty body. If your health endpoint returns 204, make sure your uptime check does not expect content.
3xx — it moved. 301 is permanent, 302 and 307 are temporary. A monitor that does not follow redirects can report a healthy page as broken, or worse, a broken page as healthy because the redirect target still answers. Know which behavior your checks use.
4xx — your request is wrong. 401 unauthorized, 403 forbidden, 404 not found, 429 rate limited. A 429 during a deploy is usually your load balancer protecting the origin, not a crash.
5xx — the server is wrong. This is outage territory. 500 is a generic crash, 502 means a proxy got garbage from upstream, 503 means overloaded or down for maintenance, 504 means the upstream never answered. When users report 502s, the application behind the proxy is the suspect, not the proxy.
How to check what a URL returns
Browsers hide status codes behind error pages, so check them properly:
curl -sI https://example.com -o /dev/null -w '%{http_code}'returns the bare codecurl -vshows redirects, TLS and headers in one shot- Or skip the terminal: our free HTTP status code checker shows the code, redirect chain and response headers for any URL.
The gap between "200" and "working"
A status code is the front door, not the building. Your app can return 200 while the checkout is broken, the database is read-only, or latency went from 80 ms to 8 seconds. That is why real monitoring checks content and response time on top of the code — and alerts when any of the three drifts.
If you want that running continuously instead of ad-hoc checks, an Uptriva monitor watches the status code, a keyword on the page and the response time from multiple regions, then pages you on the first anomaly. The free tier covers 15 monitors, which is enough for the endpoints that pay your rent.