云服务器Nginx proxy_next_upstream配置实现故障转移

upstream backend_pool {
    server 192.0.2.1:8080;
    server 192.0.2.2:8080;
    server 192.0.2.3:8080 backup;  # 备用节点（仅当主节点全挂时启用）
}

server {
    location / {
        proxy_pass http://backend_pool;
        proxy_next_upstream error timeout invalid_header http_502 http_503 http_504;
        proxy_next_upstream_tries 3;  # 最多尝试3个节点
        proxy_next_upstream_timeout 5s;  # 5秒内完成故障转移
    }
}

upstream backend_pool {
    server 192.0.2.1:8080 max_fails=3 fail_timeout=30s;  # 3次失败后标记为不可用，30秒后恢复检查
    server 192.0.2.2:8080;
}

-- 在Nginx的init_by_lua_block中加载共享字典
init_by_lua_block {
    local node_status = ngx.shared.node_status
    node_status:set("192.0.2.1", 100)  -- 初始权重100
}

-- 在content_by_lua_block中动态选择节点
location / {
    access_by_lua_block {
        local node_status = ngx.shared.node_status
        local backend = nil
        local min_load = 999

        -- 遍历所有节点，选择负载最低的可用节点
        for _, server in ipairs({"192.0.2.1", "192.0.2.2"}) do
            local load = node_status:get(server) or 100
            if load < min_load then
                min_load = load
                backend = server
            end
        end

        if backend then
            ngx.var.backend = backend
        else
            ngx.exit(503)
        end
    }

    proxy_pass http://$backend;
}

-- 通过外部脚本更新节点负载（如每5秒执行一次）
-- curl -X POST "http://nginx-server/update_load?node=192.0.2.1&load=75"

upstream backend_pool {
    # 通过Consul的DNS接口动态获取节点列表
    server backend.service.consul:8080 resolve;
    server backup.service.consul:8080 backup;
}

server {
    location / {
        proxy_pass http://backend_pool;
        proxy_next_upstream error timeout;
        resolver 8.8.8.8 valid=30s;  # 配置DNS解析器
    }
}

upstream geo_backend {
    # 主数据中心（北京）
    server 192.0.2.1:8080 max_fails=2 fail_timeout=10s;
    server 192.0.2.2:8080;

    # 备用数据中心（上海）
    server 203.0.113.1:8080 backup;
    server 203.0.113.2:8080 backup;
}

server {
    location / {
        proxy_pass http://geo_backend;
        proxy_next_upstream error timeout http_502 http_503;
        proxy_next_upstream_tries 4;  # 尝试2个主节点+2个备用节点

        # 基于GeoIP的智能路由（需安装ngx_http_geoip_module）
        geo $preferred_region {
            default main;
            10.0.0.0/8 main;  # 内部网络走主数据中心
            203.0.113.0/24 backup;  # 上海用户直接访问本地节点
        }

        proxy_set_header X-Region $preferred_region;
    }
}

# 记录故障转移次数与耗时
- name: nginx_proxy_next_upstream
  rules:
  - record: job:nginx_proxy_next_upstream_total:count
    expr: sum(rate(nginx_http_requests_total{status=~"502|503|504"}[1m])) by (job)
  - record: job:nginx_proxy_next_upstream_duration_seconds:avg
    expr: sum(rate(nginx_http_request_duration_seconds_bucket{le="0.5"}[1m])) by (job) / sum(rate(nginx_http_request_duration_seconds_count[1m])) by (job)

groups:
- name: nginx-proxy-alert
  rules:
  - alert: HighFailureRate
    expr: rate(nginx_http_requests_total{status=~"502|503|504"}[5m]) > 10  # 每秒10次故障
    for: 1m
    labels:
      severity: critical
    annotations:
      summary: "Nginx故障转移率过高"
      description: "后端服务 {{ $labels.instance }} 在5分钟内发生 {{ $value }} 次故障转移，请检查节点状态！"