Load Balancing with upstream

Multiple Backends

An upstream group balances traffic across identical app instances:

upstream app_servers {
    server 10.0.0.11:8080;
    server 10.0.0.12:8080;
    server 10.0.0.13:8080;
}

server {
    listen 80;
    location / {
        proxy_pass http://app_servers;
    }
}

Balancing Methods

• Round-robin (default) - every backend gets a fair turn.
• least_conn - send to the backend with fewest connections.
• ip_hash - route a client to the same backend every time (nice for sticky-but-not-sticky sessions).
• weight - give bigger servers more traffic: server 10.0.0.11 weight=3;.

Health Checks and Failover

Nginx Open Source marks backends down when requests fail hard:

upstream app_servers {
    server 10.0.0.11:8080 max_fails=3 fail_timeout=30s;
    server 10.0.0.12:8080 backup;   # used only if the others fail
}

Nginx Plus adds active HTTP health checks with health_check interval=5s;.

Caveat: Sticky Sessions

If the app stores sessions in memory, ip_hash keeps users pinned. Better: move sessions to a shared store (Redis/DB) so any backend can serve anyone - that is how you scale.

Key Points

• upstream + proxy_pass balances and tolerates backend failures.
• Choose least_conn/weight by your workload; ip_hash for sticky needs.
• Stateless apps let load balancing do its best work.