NGINX: Common FastCGI Errors involving PHP-FPM

nginx_fastcgi_errors.md

Common FastCGI Errors involving PHP-FPM

1. 504 Gateway Timeout

   upstream timed out (110: Connection timed out) while reading response header from upstream
   

   Means NGINX timed out reading the HTTP response header from upstream. This is controlled by:

   fastcgi_read_timeout 60s;
   

2. 502 Bad Gateway

   recv() failed (104: Connection reset by peer) while reading response header from upstream
   readv() failed (104: Connection reset by peer) while reading upstream
   

   Means PHP-FPM reset the connection while NGINX was trying to read the HTTP response header. It is not NGINX's fault.

   This could happen due to either being killed, meeting a fatal error (possibly from extensions), hitting the memory limit, or the execution time limit inside PHP.

   You can tune the parameters inside PHP-FPM and the php.ini and change these limits and see if anything works. For PHP-FPM see:

   process_control_timeout = 0
   request_terminate_timeout = 0
   php_admin_value[memory_limit] = 128M
   listen.backlog = 65535
   rlimit_files = 4096
   

   For php.ini see:

   max_execution_time = 80
   max_input_time = 60
   memory_limit = 128M
   default_socket_timeout = 60
   

   It may not be your script per se, the worker might have just crashed due to PHP memory leaks, and you may need PHP-FPM to restart the worker, or kill zombie workers.

   Also see: http://stackoverflow.com/questions/1434451/what-does-connection-reset-by-peer-mean