Apache error log codes

# About Apache error codes

Every line in the Apache error log carries a unique AHxxxxx code — the letters AH plus five digits, like AH01630. Each marks one spot in the Apache source, so a code always points at the same condition.

# Relation to HTTP status codes

An Apache log code is not an HTTP status. Apache log codes live in the error log, statuses in the access log, the mapping is many-to-one — several codes share one status, many map to none. That is the value: a 500 only says something failed internally, while the code says whether a PHP-FPM process died, a rewrite loop ran away, or a rule denied the request.

On fortrabbit, Apache talks to PHP-FPM over FastCGI (proxy_fcgi). The codes below come from that path, from access control, or from Apache's core. Find them in the apache log source — see logs.

# Access and authorization

Apache uses access rules and Require directives to control who can reach your app. These errors occur when a request is blocked by those rules.

# AH01630: client denied by server configuration

[authz_core:error] AH01630: client denied by server configuration: /srv/app/htdocs/.env

An access rule blocked the request before it reached the app. The path at the end of the message is what was requested — often a protected file like .env or a directory closed off by a Require directive. Maps to a 403.

Check the .htaccess rules and any Require directives against the requested path. See the .htaccess intro for background. Requests for .env, .git, or dotfiles are usually scanners probing for secrets and can be ignored.

# AH01631: user authorization failure

[authz_core:error] AH01631: user jane: authorization failure for "/admin/":

Credentials were accepted but the authenticated user was not allowed into the requested area, or the authorization directives for that area are incomplete. Maps to a 403 — or a 401 when AuthzSendForbiddenOnFailure is off.

Review the Require user / Require group rules for the protected path and confirm the user belongs there.

# Missing files and bad requests

Apache logs these errors when it cannot find a requested file or receives a malformed URI that it cannot parse.

# AH00128: file does not exist

[core:info] AH00128: File does not exist: /srv/app/htdocs/public/favicon.ico

Apache could not find the requested file under the root path. Maps to a 404. Most lines are harmless — browsers and bots probing for favicon.ico, apple-touch-icon.png, or stale source maps.

If a real page 404s, confirm the root path points at the framework's public directory and that the front-controller rewrite (index.php) is in place.

# AH00126: invalid URI in request

[core:error] AH00126: Invalid URI in request GET /%c0%ae/ HTTP/1.1

The request line held a malformed URI. Maps to a 400. These almost always come from bots, malformed proxies, or scanners sending broken paths — rarely from the app itself.

# Redirect loops

Redirect loops occur when rewrite rules in .htaccess accidentally redirect requests to themselves, causing Apache to give up after a maximum number of internal redirects.

# AH00124: request exceeded the limit of internal redirects

[core:error] AH00124: Request exceeded the limit of 10 internal redirects due to probable configuration error.

A rewrite rule kept redirecting the request to itself until Apache gave up. Maps to a 500. The cause is nearly always a loop in .htaccess — a RewriteRule whose target matches its own pattern.

Review recent .htaccess changes and the RewriteRule targets. See the .htaccess intro.

# PHP-FPM and FastCGI

Since fortrabbit apps communicate with PHP-FPM over FastCGI, several error codes reflect failures in that connection or crashes in the PHP process.

# AH01071: PHP-FPM error output

[proxy_fcgi:error] AH01071: Got error 'PHP message: PHP Fatal error: Uncaught Error: Call to undefined function foo()'

PHP-FPM wrote to its error output and Apache surfaced it. The quoted text is the PHP message itself — a notice, warning, or fatal. A fatal pairs with a 500; a warning may still return 200 while noise piles up in the log.

Read the quoted PHP message for the real cause, then open the php log source for the full stack trace. See logs.

# AH01075: error dispatching request

[proxy_fcgi:error] AH01075: Error dispatching request to : (reading input brigade): Connection reset by peer

Apache started talking to PHP-FPM but the exchange broke off, commonly with Connection reset by peer. The PHP worker died mid-request — a fatal, a memory limit, a hard timeout, or a pool restart. Maps to a 503.

Look for a matching AH01071 fatal at the same time, and check memory_limit and execution time against what the request needs.

# AH01067: failed to read FastCGI header

[proxy_fcgi:error] AH01067: Failed to read FastCGI header

PHP-FPM returned no valid response header at all — the worker crashed, timed out, or was killed before it could answer. Maps to a 503. Close cousin of AH01075; both point at a PHP process that never finished.

Check for out-of-memory kills and fatals in the php log, and confirm memory_limit and max_execution_time fit the workload.