Fixing Nginx 502 Errors

My setup includes a proxy server (nginx) and two other servers running nginx, php-fpm 7.4, and mariadb on a separate server. I've been experiencing random 502 errors throughout the day. Restarting servers, checking logs, and re-configuring nginx have not resolved the issue.

I'm looking for an experienced and expert system administrator with expertise in nginx, php-fpm, and mariadb to diagnose and fix the issue. Please provide examples of similar issues you've resolved in the past.

Note that we cannot access the server proxy, just able to access two other servers.
The fix should be made using Anydesk's connection to conduct the fix.

2024/08/14 22:55:16 [error] 1472680#1472680: *1387 recv() failed (104: Connection reset by peer) while reading response header from upstream, client: 10.100.15.249, server: abc.com, request: "GET /JOINTS/article/view/20301 HTTP/1.1", upstream: "fastcgi://unix:/run/php/1720061007114156.sock:", host: "x"

What we have done :
- adding more upstream (TCP and UNIX) for php-fpm
- Improve mysql max_connection
- improve synctl.conf on both servers


Requirement for work :
- Warranty if the server doesn't have other 502 error
- Non Disclosure data
- Commitment not adding any data or illegal backdoor to the server