nginx + PHP-FPM Production Config

Also Known As

TL;DR

Explanation

nginx as a reverse proxy in front of PHP-FPM provides: serving static files directly (no PHP overhead), buffering slow client uploads before passing to FPM (protecting worker count), SSL termination, gzip compression, and security headers. PHP-FPM pool configuration controls worker count, memory limits, and request timeouts. Key settings: pm.max_children (max concurrent PHP processes), pm.max_requests (restart workers after N requests to prevent memory leaks), fastcgi_read_timeout (PHP execution time limit in nginx).

Common Misconception

Why It Matters

Common Mistakes

• pm.max_children set too high — workers exceed available RAM, causing swap and severe slowdowns.
• No fastcgi_read_timeout — nginx kills long-running PHP processes before they complete.
• Not setting PHP-FPM pm.max_requests — workers accumulate memory leaks without restarts.
• Missing fastcgi_param PHP_VALUE to override php.ini per virtual host.

Code Examples

# Minimal nginx config — missing critical settings:
server {
    listen 80;
    root /var/www/html;
    index index.php;
    location ~ \.php$ {
        fastcgi_pass 127.0.0.1:9000;
        include fastcgi_params;
        fastcgi_param SCRIPT_FILENAME $document_root$fastcgi_script_name;
        # No timeout, no buffering, no security headers
    }
}
; php-fpm.conf — no limits:
pm = dynamic
pm.max_children = 50  ; May exceed available RAM

# Production nginx config:
server {
    listen 443 ssl http2;
    root /var/www/html/public;
    
    # Serve static files directly:
    location ~* \.(css|js|png|jpg|woff2)$ {
        expires 1y; add_header Cache-Control 'public, immutable';
    }
    location ~ \.php$ {
        fastcgi_pass unix:/run/php-fpm.sock;
        fastcgi_read_timeout 30;
        fastcgi_buffers 16 16k;
        fastcgi_buffer_size 32k;
        include fastcgi_params;
        fastcgi_param SCRIPT_FILENAME $realpath_root$fastcgi_script_name;
    }
}
; php-fpm pool.conf:
pm = dynamic
pm.max_children = 20     ; RAM / ~50MB per worker
pm.max_requests = 500     ; Restart after 500 requests

References

• ↗ https://www.nginx.com/resources/wiki/start/topics/examples/phpfcgi/

Tags

Related Terms