We have two drupal apache web servers behind ALB. Nginx proxy_read_timeout as 300 no other
proxy_connect_timeout
proxy_send_timeout
keepalive_timeout has been defined/configured. Occasionally getting upstream timed out (110: Connection timed out) while connecting to upstream.
However, when we look at all of the upstream components they are working. The ALB is working, and all targets are healthy. The individual servers are up and healthy. It appears that "something happened" causing it to break, but whatever that event was it was resolved. However, after the resolution the proxy connections were not reset, still resulting in timeouts.
We can resolve the issue by restarting nginx, However, we want nginx to reset the proxy connection on it's own without restarting the server.
