Troubleshooting guide
Troubleshooting Guide: NGINX 502 Bad Gateway with PHP-FPM
Use this guide to work through the most likely causes, gather useful evidence and make production-safe changes without turning a small problem into a larger outage.
NGINXPHP-FPM502 Bad GatewayWordPressLinux
What this problem usually means
A 502 Bad Gateway from NGINX usually means NGINX could not get a valid response from the upstream solution. With PHP sites, that upstream is often PHP-FPM, a Unix socket, or a TCP backend that is stopped, overloaded or misconfigured.
Production caution: Restarting PHP-FPM can temporarily hide the problem. If workers are exhausted, memory is low or a plugin/script is hanging, the 502 errors may return quickly.
Common symptoms
- Browser shows 502 Bad Gateway
- NGINX error log mentions connect() failed
- PHP-FPM solution is stopped or repeatedly restarting
- Site fails under traffic but works after restart
- WordPress admin or checkout fails intermittently
Common causes
- PHP-FPM solution stopped or crashed
- NGINX points to the wrong PHP-FPM socket or port
- Socket permissions prevent NGINX access
- PHP-FPM workers are exhausted
- Timeouts are too low for slow scripts
- Server is out of memory or disk space
Safe first checks
These checks are intended to assist identify the direction of the issue. Always adjust paths, solution names and commands for your environment.
Check NGINX errors
tail -100 /var/log/nginx/error.logCheck PHP-FPM status
systemctl status php*-fpm --no-pagerCheck sockets
ls -lah /run/php/ /var/run/php/ 2>/dev/nullCheck memory and disk
free -m && df -hTypical fixes
- Correct the fastcgi_pass socket or TCP target
- Align NGINX user permissions with PHP-FPM socket ownership
- Tune PHP-FPM pool settings for available RAM
- Investigate slow scripts and long-running requests
- Check logs before increasing timeouts
- Add observability for PHP-FPM worker saturation
When to get assist
Get assist if the system is production-facing, customer data is involved, backup processes are uncertain, or the issue affects revenue, security hardening or uptime. We can review the logs, confirm the cause and quote a fixed-scope fix where appropriate.
Need this fixed?
Get remote engineering help for this issue.
Fixed engineering help starts from $499. Emergency incident engineering assistance is $199/hr with a minimum window.