502 Bad Gateway: Causes and How to Fix It (Nginx, PHP-FPM and Cloudflare)

Applies to: Nginx, Apache or LiteSpeed in front of PHP-FPM, Node.js, Docker or another backend; error messages tested with Nginx 1.31.6

502 Bad Gateway means that a server acting as a gateway or proxy (Nginx, a load balancer or Cloudflare) asked the server behind it for the page and got no valid answer. The visitor’s browser is almost never the cause: the problem is the backend, such as PHP-FPM, a Node.js app or a container.

1. Find which layer shows the error

• A Cloudflare error page means Cloudflare could not get a valid response from your server. Check your server next.
• A plain page with 502 Bad Gateway and nginx underneath comes from Nginx on your server.

To test your server directly, bypassing Cloudflare (use your server’s IP address):

curl -sI --resolve example.com:443:203.0.113.10 https://example.com/ | head -1

2. Read the error log

sudo tail -n 50 /var/log/nginx/error.log

These are the messages we reproduced and what they mean:

• connect() failed (111: Connection refused) while connecting to upstream: nothing listens on that port. The backend is stopped, or the port in proxy_pass is wrong.
• connect() to unix:/run/php/php8.3-fpm.sock failed (2: No such file or directory): PHP-FPM is stopped, or the socket path in fastcgi_pass doesn’t match (common after changing the PHP version).
• upstream prematurely closed connection while reading response header from upstream: the backend accepted the request but died before answering, often a crash, a fatal error or the out-of-memory killer.

3. Fix the backend

1. Is it running?

   sudo systemctl status php8.3-fpm        # Ubuntu, Debian (php-fpm on AlmaLinux, Rocky Linux)
   sudo systemctl restart php8.3-fpm
   sudo docker ps                           # for containers; then: sudo docker logs <name>

2. Does the address match? Compare ls /run/php/ (or the app’s port, see which process uses a port) with the fastcgi_pass or proxy_pass line, then run sudo nginx -t && sudo systemctl reload nginx.
3. Is memory running out? sudo journalctl -k | grep -i "killed process" shows processes stopped by the out-of-memory killer. Add RAM, lower PHP-FPM’s pm.max_children or fix the memory-hungry script.
4. Is PHP-FPM full? The PHP-FPM log says server reached pm.max_children setting. Raise it only if the server has enough RAM.

cPanel servers

With Nginx in front of Apache: fix an Engintron 502 Bad Gateway and Nginx 502 and a missing error log.

A backend that answers too slowly gives a different error, 504. If your server keeps running out of resources, compare our KVM VPS plans or ask an engineer.