健康检查机制

Nginx 通过主动/被动健康检查实现后端服务器状态监控，自动摘除故障节点，保障服务高可用。

被动健康检查

内置机制

upstream backend {
    server 10.0.0.1:8080 max_fails=3 fail_timeout=30s;
    server 10.0.0.2:8080 max_fails=3 fail_timeout=30s;
    server 10.0.0.3:8080 backup;
    server 10.0.0.4:8080 down;
}

核心参数

失败判定条件

• 连接超时
• 连接拒绝
• HTTP 5xx 响应码（需 proxy_next_upstream）

proxy_next_upstream error timeout http_500 http_502 http_503 http_504;
proxy_next_upstream_tries 3;
proxy_next_upstream_timeout 10s;

被动检查仅在实际请求时触发，无法主动发现未承载流量的服务器故障。

主动健康检查

Nginx Plus 方案

upstream backend {
    zone backend 64k;
    server 10.0.0.1:8080;
    server 10.0.0.2:8080;

    health_check interval=10s fails=3 passes=2
                 uri=/healthz match=status_200;
}

match status_200 {
    status 200;
    header Content-Type = text/html;
    body ~ "OK";
}

OpenResty 方案

-- healthcheck.lua
local healthcheck = require("resty.upstream.healthcheck")

local ok, err = healthcheck.spawn{
    name = "backend",
    type = "http",
    http_req = "GET /healthz HTTP/1.1\r\nHost: backend\r\n\r\n",
    interval = 5000,   -- 5s
    timeout = 2000,    -- 2s
    fall = 3,          -- 连续3次失败标记为 down
    rise = 2,          -- 连续2次成功标记为 up
    valid_statuses = {200, 204},
    concurrency = 10,
}

主动检查参数

状态监控

状态接口

# Nginx Plus API
location /api {
    api write=on;
    allow 127.0.0.1;
    deny all;
}

# 查询健康状态
# GET /api/7/http/upstreams/backend/peers/

响应结构

{
  "server": "10.0.0.1:8080",
  "state": "up",
  "active": 5,
  "health_checks": {
    "checks": 150,
    "fails": 2,
    "unhealthy": 0,
    "last_passed": true
  }
}

故障转移策略

优雅摘除

upstream backend {
    server 10.0.0.1:8080 weight=5;
    server 10.0.0.2:8080 weight=3;
    server 10.0.0.3:8080 weight=2;

    # 慢启动（新节点逐步承接流量）
    server 10.0.0.4:8080 slow_start=30s;
}

强制下线

# 动态 API
# POST /api/7/http/upstreams/backend/peers/
{ "down": true }

要点总结

• 被动检查依赖实际请求，通过 max_fails/fail_timeout 标记故障节点
• 主动检查定期探测健康端点，Nginx Plus 内置，OpenResty 可通过 resty.upstream 实现
• proxy_next_upstream 定义失败重试条件，支持 error/timeout/http_5xx
• 健康状态可通过 Nginx Plus API 查询，支持动态上下线
• 慢启动参数 slow_start 避免新节点突然承接全量流量