HTTP · Operations
Status codes as evidence
An HTTP response means DNS, a network connection, and usually TLS have already succeeded. The next task is to determine which component generated that response and why.
Read the class first
2xx · Success
The server accepted and completed the request, though the returned content can still be wrong.
3xx · Redirection
The client must make another request. Inspect every location in the chain.
4xx · Client-side condition
The responding server rejected or could not match the request. This can still indicate a routing mistake.
5xx · Server-side failure
A gateway, proxy, or application could not complete the request. Identify the actual responder before editing the origin.
Identify the responder
Compare response headers, body style, timing, and request IDs with known components. A CDN, reverse proxy, and application may each have a distinct error page. Treat the Server header as a clue, not proof: it can be removed or rewritten.
A custom error page should preserve the correct status. Returning a friendly page with 200 OK for a missing resource makes monitoring and search indexing unreliable.
Follow redirects deliberately
Record each status and Location value. Loops commonly arise when two layers disagree about the original scheme or hostname. A single redirect from HTTP to HTTPS is normal; an alternating host or scheme is not.
Account for caches
A cache can return a valid but old response, including an error. Note Age, Cache-Control, validators, and any cache-status header. Purging a cache can be appropriate after the source is corrected, but it should not replace finding the source.