I have NGINX set up on a Windows Server (let’s call it server 1) balancing two API’s hosted in other two Windows Servers (let’s call them server 2 and 3) through IIS.

When Active connections gets past 60, it starts throwing random Bad Gateway responses with this description in the logs: “upstream timed out (10060: A connection attempt failed because the connected party did not properly respond after a period of time, or established connection failed because connected host has failed to respond) while reading response header from upstream”.

Both servers 2 and 3 looks just fine when the error starts striking. If I’m not using NGINX and trying to access them directly, even with high loads, I have no problems, which leads me to believe I’m missing something in the config file.

Here is my conf file:

worker_processes  auto;

events {
    worker_connections  3000;
}

http {
    include       mime.types;
    default_type  application/octet-stream;

    sendfile        on;

    access_log off;
   

    keepalive_timeout  65;

    #gzip  on;

    upstream myapp {
        server (server 2 here);
        server (server 3 here);
    }

    server {
        listen       8080;

        location / {
            proxy_pass http://myapp;
            proxy_set_header Host $host;
            proxy_set_header X-Real-IP $remote_addr;
            proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
            proxy_set_header X-Forwarded-Proto $scheme;

            proxy_next_upstream error timeout invalid_header http_500 http_502 http_503 http_504;

            proxy_connect_timeout 5s;
            proxy_read_timeout 5s;
            proxy_send_timeout 5s;

        proxy_buffering on;
            proxy_buffers 8 16k;
            proxy_buffer_size 32k;
        }

        error_page   500 502 503 504  /50x.html;
        location = /50x.html {
            root   html;
        }

        location /nginx_status {
            stub_status;
            allow 127.0.0.1;  # Only allow access from localhost
            deny all;         # Deny access from all other IPs
        }
    }

}

I tried increasing worker_connections and increasing all the timeouts to 30, but nothing appears to work.

I was able to reproduce it on my local machine running two HTML files with python server on 8081 and 8082 to mimic the windows servers, while running nginx on 8080. I used apache jmeter to make the load testing and was able to get the error when I set the number of users to 400 and looping it for 30 seconds. Here’s the changes I made locally but also didn’t made any difference:

worker_processes auto;
worker_rlimit_nofile 10000;

events {
    worker_connections 10000;
    use poll;
    multi_accept on;
}

http {
    include       mime.types;
    default_type  application/octet-stream;

    sendfile        on;
    tcp_nopush     on;
    tcp_nodelay    on;

    # error_log logs/error.log crit;

    keepalive_timeout 65;
    keepalive_requests 100000;

    gzip  on;

    upstream myapp {
        server localhost:8081;
        server localhost:8082;
    }

    server {
        listen 8080;

        location / {
            proxy_pass http://myapp;
            proxy_set_header Host $host;
            proxy_set_header X-Real-IP $remote_addr;
            proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
            proxy_set_header X-Forwarded-Proto $scheme;

            proxy_next_upstream error timeout invalid_header http_500 http_502 http_503 http_504;
            # proxy_connect_timeout 5s;
            # proxy_read_timeout 5s;
            # proxy_send_timeout 5s;

            proxy_buffer_size 16k;
            proxy_buffers 4 32k;
            proxy_busy_buffers_size 64k;
            proxy_temp_file_write_size 64k;

            add_header Cache-Control "no-store, no-cache, must-revalidate, proxy-revalidate, max-age=0";
            add_header Pragma "no-cache";
            expires -1;
        }

        error_page 500 502 503 504 /50x.html;
        location = /50x.html {
            root html;
        }

        location /nginx_status {
            stub_status;
            allow 127.0.0.1;
            deny all;
        }
    }
}