Nginx负载均衡策略需求：分区域固定主备服务器转发流量

Hey there! That massive weight workaround you're using does technically get the job done, but you’re totally right—it’s not the clean or intended way to set up primary/backup failover in Nginx. Luckily, there’s a built-in, standard directive made exactly for this scenario: backup.

How It Works #

The backup flag tells Nginx to only send traffic to that server if all non-backup servers are unavailable. This aligns perfectly with your requirement: keep 100% of traffic on the primary server when it’s healthy, and switch to the backup only when the primary fails.

Revised Configuration #

Here’s how to rewrite your upstreams using the proper backup parameter, while keeping your existing session persistence with ip_hash:

upstream US_UPSTREAM { 
    ip_hash; 
    # Primary server: handles all traffic when healthy
    server 2.2.2.2 max_fails=10 fail_timeout=3600s; 
    # Backup server: only used if primary is marked down
    server 1.1.1.1 backup; 
} 

upstream EU_UPSTREAM { 
    ip_hash; 
    # Primary server for EU region
    server 1.1.1.1 max_fails=10 fail_timeout=3600s; 
    # Backup server for EU region
    server 2.2.2.2 backup; 
}

Key Details to Keep in Mind #

• ip_hash compatibility: Your ip_hash directive works seamlessly with backup—users will stick to the same server (primary or backup) as long as their IP stays consistent, preserving session persistence.
• Failover thresholds: Your current max_fails=10 and fail_timeout=3600s mean Nginx will mark the primary as down after 10 failed connection attempts, and keep it marked down for an hour. Adjust these values if you want faster/slower failover (e.g., max_fails=3 fail_timeout=30s for quicker fault detection).
• Explicit intent: Unlike the massive weight hack, this configuration makes your primary/backup relationship crystal clear—anyone reading the config will immediately understand your traffic routing strategy, making maintenance way easier.

Testing the Failover #

To confirm it works as expected:

1. Stop the primary server for a region.
2. Send traffic to that upstream—you’ll see it route exclusively to the backup server.
3. Restart the primary server; after the fail_timeout period elapses, Nginx will automatically shift traffic back to the primary.

内容的提问来源于stack exchange，提问作者ace