HTTP status codes in the 400s say the request was rejected or could not be fulfilled as sent. Codes in the 500s say the server failed to complete a request that looked valid. That distinction decides who has to act: usually the person or code making the request for a 4xx, and the site owner for a 5xx. Real systems blur the line, though, so this page shows where the rule helps and where it misleads.
The core difference
| 4xx (client error) | 5xx (server error) | |
|---|---|---|
| Meaning | The request has a problem or is not allowed | The server could not fulfill a valid request |
| Usually fixed by | Changing the request, URL, credentials or permissions | Fixing the server, app, upstream or capacity |
| Retry advice | Do not retry unchanged (with exceptions like 408 and 429) | Retrying later may work |
| Counts as "site down"? | Usually not, but sometimes | Almost always |
| Typical SEO effect | 404 and 410 remove pages from the index over time | Sustained 5xx can slow crawling and drop pages |
Common 4xx codes
- 400 Bad Request. Malformed syntax, an oversized header or a bad cookie. Clearing cookies often fixes it for a browser user.
- 401 Unauthorized. Credentials missing or invalid. Despite the name it is about authentication.
- 403 Forbidden. Authenticated or not, you are not allowed. Often a permissions rule, IP block or firewall rule.
- 404 Not Found. The URL does not exist, or the server will not say it does.
- 405 Method Not Allowed. The URL exists but not for this verb, for example a POST to a read-only route.
- 408 Request Timeout. The server gave up waiting for the client to finish sending.
- 410 Gone. Intentionally removed, stronger than 404.
- 413 Content Too Large. The request body exceeds a limit; see nginx 413.
- 429 Too Many Requests. Rate limiting. A
Retry-Afterheader may tell you how long to wait.
Common 5xx codes
- 500 Internal Server Error. A generic failure, usually an unhandled exception or bad configuration. See fix 500.
- 502 Bad Gateway. A proxy or gateway got an invalid response from the upstream. See fix 502.
- 503 Service Unavailable. Overloaded or in maintenance. It may include
Retry-After. See fix 503. - 504 Gateway Timeout. The upstream took too long. See fix 504.
The full list lives in the HTTP status code reference, with a page per code such as 404 और 503.
When the rule bends
The 4xx/5xx split is a convention, and real software breaks it.
- A 404 that is really a server problem. A misdeployed app, missing static files or a wrong document root can return 404 for pages that should exist. If the whole site suddenly returns 404, treat it as an outage.
- A 403 from a security layer. A WAF or CDN may return 403 to monitoring probes or certain countries. The server is fine; your check is being blocked. See Cloudflare error 1020.
- A 5xx caused by a bad request. Some apps crash on unexpected input and return 500, even though the client's input was the trigger.
- A 200 with an error page. Some apps return 200 for an error message. A status code check will not notice; use a keyword check instead.
What to check first
- Look at the exact code. Use the HTTP status checker so you see what an external client sees, not what your cached browser shows.
- Is it every URL or one URL? All URLs failing points to server, proxy or deploy; one URL points to routing, a missing file or permissions.
- Is it every client? If it works from your phone but not from the office, suspect a firewall, geo rule or rate limit.
- Read the headers. The HTTP header checker shows
Server,Via,Retry-Afterand cache headers that reveal which layer answered. - Follow redirects. A chain that ends in 4xx or 5xx is easier to read in the redirect checker.
- Find the layer. The outage root cause finder tests DNS, TCP, TLS and HTTP in order.
How monitoring should treat each
For an uptime monitor, 5xx should alert. 4xx depends on the URL. A homepage returning 404 or 403 is a real outage for visitors, so alert. An API route returning 401 without a token might be expected, in which case set the expected status to 401 or send the token. Pair status checks with a keyword check to catch 200 responses that carry an error page. The guides on reducing false alerts और monitoring API endpoints go deeper.
अगले कदम
- Check any URL with the HTTP status checker.
- Browse the HTTP status code reference.
- Start with why is my website down.
- Set up an uptime monitor with an expected status code.
- Read website downtime and SEO.
सामान्य प्रश्न
Is a 404 a server error?
No, it is a client error by definition. But if pages that should exist return 404 after a deploy, the cause is on your server.
Should I retry a 4xx?
Not unchanged, with exceptions such as 408 and 429, where waiting or slowing down can help. Retrying a 5xx after a delay is often reasonable.
Do 5xx errors hurt SEO?
Prolonged 5xx responses can reduce crawling and cause pages to drop from results. Brief 503s with Retry-After are the recommended way to signal temporary maintenance.
Why does my monitor report 403 when the site works for me?
A firewall or CDN is probably blocking the monitor's IP or user agent. Allow it or adjust the rule.
Try SiteDownCheck free
3 monitors, 5-minute checks, email and webhook alerts and a public status page. No credit card.
मुफ़्त मॉनिटरिंग शुरू करेंOpinions here are our own; verify vendor details yourself. See our संपादकीय नीति और अस्वीकरण.