When one app process isn’t enough, run several and let Nginx spread traffic across them. Declare an upstream pool at the http { } level, then proxy_pass to it by name.

upstream app_backend {
    least_conn;                       # send to the server with fewest active connections

    server 127.0.0.1:3001;
    server 127.0.0.1:3002;
    server 127.0.0.1:3003 weight=2;   # gets ~2x the traffic
    server 10.0.0.9:3000 backup;      # only used if the others are down

    keepalive 32;                     # pooled connections to the backends
}

server {
    listen 80;
    server_name example.com;

    location / {
        proxy_pass http://app_backend;

        proxy_http_version 1.1;
        proxy_set_header Connection "";        # required for upstream keepalive
        proxy_set_header Host            $host;
        proxy_set_header X-Real-IP       $remote_addr;
        proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
    }
}

Balancing methods (pick one, default is round-robin):

round-robin (default) — even rotation.
least_conn; — fewest active connections; good for uneven request durations.
ip_hash; — same client IP always hits the same backend (sticky sessions without shared state).

Other knobs:

weight=N sends more traffic to bigger servers; backup marks a standby.
Passive health: Nginx marks a server failed after max_fails errors within fail_timeout (e.g. server … max_fails=3 fail_timeout=15s;) and stops sending to it temporarily.
keepalive 32; plus proxy_http_version 1.1 and an empty Connection header reuses backend connections — a big latency win. (Active health checks need Nginx Plus.)

sudo nginx -t && sudo nginx -s reload

Load-balance across multiple backends