在高流量业务场景中，单台服务器宕机会导致整个服务不可用。Keepalived配合HAProxy构建的双机热备集群是实现服务器高可用的经典方案，通过VRRP协议实现VIP故障切换，配合HAProxy的七层负载均衡，可将服务可用性提升至99.99%。本文记录一套从零搭建到验证完整的双机热备部署流程。

双机热备架构设计与环境准备

架构拓扑：两台物理服务器（Master/Backup）各运行HAProxy实例，Keepalived通过VRRP协议维护一个浮动VIP。正常情况下Master持有VIP并处理所有流量，Master故障时Backup在秒级接管VIP。

# 环境信息
# Master节点: 192.168.10.11  (hostname: lb-master)
# Backup节点: 192.168.10.12  (hostname: lb-backup)
# VIP:        192.168.10.100
# 后端Web服务器: 192.168.10.21, 192.168.10.22

# 系统环境
cat /etc/os-release  # Rocky Linux 9.3
uname -r            # 5.14.0-362.el9.x86_64

# 安装依赖
dnf install -y keepalived haproxy
systemctl enable keepalived haproxy

Keepalived VRRP配置主备切换

Keepalived的核心是VRRP（虚拟路由冗余协议）。Master节点优先级高于Backup，定期发送VRRP通告包。Backup在3个通告周期内未收到Master通告，则认为Master故障并提升自身为Master。

Master节点配置文件/etc/keepalived/keepalived.conf：

global_defs {
    router_id LB_MASTER
    enable_script_security
    script_user root
}

vrrp_script chk_haproxy {
    script "/etc/keepalived/check_haproxy.sh"
    interval 2
    weight -20
    fall 3
    rise 2
}

vrrp_instance VI_1 {
    state MASTER
    interface eth0
    virtual_router_id 51
    priority 100
    advert_int 1

    authentication {
        auth_type PASS
        auth_pass MySecret2026
    }

    virtual_ipaddress {
        192.168.10.100/24 dev eth0
    }

    track_script {
        chk_haproxy
    }

    notify_master "/etc/keepalived/notify.sh master"
    notify_backup "/etc/keepalived/notify.sh backup"
    notify_fault  "/etc/keepalived/notify.sh fault"
}

Backup节点配置几乎相同，仅state和priority不同：

vrrp_instance VI_1 {
    state BACKUP
    interface eth0
    virtual_router_id 51
    priority 90          # 低于Master的100
    advert_int 1
    # ... 其余配置与Master一致
}

健康检查脚本/etc/keepalived/check_haproxy.sh：

#!/bin/bash
# 检查HAProxy进程是否存活
if ! killall -0 haproxy 2>/dev/null; then
    systemctl restart haproxy
    sleep 2
    if ! killall -0 haproxy 2>/dev/null; then
        echo "HAProxy is down" >&2
        exit 1
    fi
fi

# 检查HAProxy管理端口是否响应
if ! curl -s -o /dev/null -w '%{http_code}' http://127.0.0.1:8404/stats | grep -q '200\|503'; then
    exit 1
fi

exit 0

chmod +x /etc/keepalived/check_haproxy.sh
chmod +x /etc/keepalived/notify.sh

HAProxy七层负载均衡配置

HAProxy作为前置代理，将VIP流量分发到后端Web服务器。关键配置包括后端健康检查、会话保持、连接复用等。

配置文件/etc/haproxy/haproxy.cfg：

global
    log         127.0.0.1 local2
    chroot      /var/lib/haproxy
    pidfile     /var/run/haproxy.pid
    maxconn     20000
    user        haproxy
    group       haproxy
    daemon
    stats socket /var/lib/haproxy/stats mode 660 level admin

defaults
    mode                    http
    log                     global
    option                  httplog
    option                  dontlognull
    option http-server-close
    option forwardfor       except 127.0.0.0/8
    timeout connect         5s
    timeout client          30s
    timeout server          30s
    timeout http-keep-alive 10s

# 管理统计页面
frontend stats
    bind *:8404
    mode http
    stats enable
    stats uri /stats
    stats admin if TRUE
    stats auth admin:HaProxy2026

# HTTP前端
frontend web_http
    bind *:80
    mode http
    option httplog
    
    # ACL规则
    acl is_api path_beg /api
    acl is_static path_end .css .js .png .jpg .gif .ico .woff2
    
    # 路由
    use_backend api_servers if is_api
    use_backend static_servers if is_static
    default_backend web_servers

# 后端Web服务器组
backend web_servers
    mode http
    balance roundrobin
    option httpchk GET /healthcheck
    
    # 会话保持：基于Cookie
    cookie SERVERID insert indirect nocache
    
    server web1 192.168.10.21:80 check cookie web1 inter 2s rise 2 fall 3
    server web2 192.168.10.22:80 check cookie web2 inter 2s rise 2 fall 3

# API服务器组
backend api_servers
    mode http
    balance leastconn
    option httpchk GET /api/health
    
    # 连接复用
    option http-keep-alive
    http-reuse safe
    
    server api1 192.168.10.21:8080 check inter 2s rise 2 fall 3
    server api2 192.168.10.22:8080 check inter 2s rise 2 fall 3

# 静态资源服务器组
backend static_servers
    mode http
    balance roundrobin
    option httpchk GET /favicon.ico
    
    # 静态资源缓存
    http-response set-header Cache-Control "public, max-age=86400"
    
    server static1 192.168.10.21:80 check inter 5s rise 2 fall 3
    server static2 192.168.10.22:80 check inter 5s rise 2 fall 3

VRRP抢占模式与非抢占模式选择

默认配置下，原Master恢复后会重新抢占VIP。某些场景下这种行为不理想——例如Master频繁抖动会导致VIP反复切换。可通过nopreempt实现非抢占模式：

# 非抢占模式配置（Backup节点需设为MASTER初始状态）
vrrp_instance VI_1 {
    state BACKUP        # 两台都设为BACKUP
    interface eth0
    virtual_router_id 51
    priority 100        # 仍然是100，但因为nopreempt不会抢占
    advert_int 1
    
    nopreempt           # 关键：禁止抢占
    
    # ... 其余配置不变
}

非抢占模式下，原Master恢复后不会自动接管VIP，需手动执行systemctl restart keepalived触发切换。生产环境建议采用非抢占模式配合手动切换，避免抖动场景下的来回切换。

故障切换验证与监控告警

配置完成后需进行完整的故障切换验证。通知脚本/etc/keepalived/notify.sh：

#!/bin/bash
# 故障切换通知脚本
STATE=$1
HOSTNAME=$(hostname)
VIP="192.168.10.100"
DATE=$(date '+%Y-%m-%d %H:%M:%S')

case "$STATE" in
    master)
        MSG="${DATE} | ${HOSTNAME} 成为MASTER，接管VIP ${VIP}"
        ;;
    backup)
        MSG="${DATE} | ${HOSTNAME} 降级为BACKUP，释放VIP ${VIP}"
        ;;
    fault)
        MSG="${DATE} | ${HOSTNAME} 进入FAULT状态"
        ;;
esac

# 记录日志
logger -t keepalived "$MSG"

# 发送告警（示例：企业微信机器人）
WEBHOOK="https://qyapi.weixin.qq.com/cgi-bin/webhook/send?key=YOUR_KEY"
curl -s -X POST "$WEBHOOK" \
    -H 'Content-Type: application/json' \
    -d "{\"msgtype\":\"markdown\",\"markdown\":{\"content\":\"**高可用告警**\n${MSG}\"}}"

exit 0

验证步骤：

# 1. 启动服务（两台都执行）
systemctl start keepalived haproxy

# 2. 确认Master持有VIP
ip addr show eth0 | grep 192.168.10.100
# 应在Master节点看到VIP

# 3. 模拟故障：停止Master的Keepalived
ssh lb-master "systemctl stop keepalived"

# 4. 在Backup节点确认VIP切换
ip addr show eth0 | grep 192.168.10.100
# 应在Backup节点看到VIP，延迟通常1-3秒

# 5. 持续压测验证无中断
ab -n 100000 -c 100 http://192.168.10.100/
# 检查失败请求数是否为0或极低

# 6. 恢复Master
ssh lb-master "systemctl start keepalived"
# 抢占模式下VIP会切回Master

监控层面建议对以下指标设置告警：vrrp_script检查失败次数、VIP切换事件、后端服务器健康检查失败率、HAProxy当前连接数超过阈值。可使用Prometheus + haproxy_exporter采集指标，配合Alertmanager实现多渠道告警通知。

性能调优与内核参数优化

高流量场景下需调优系统内核参数和HAProxy配置以避免连接耗尽：

# /etc/sysctl.conf 网络参数优化
# 增加文件描述符上限
fs.file-max = 2097152

# TCP连接复用
net.ipv4.tcp_tw_reuse = 1
net.ipv4.tcp_fin_timeout = 15

# 增加SYN队列
net.ipv4.tcp_max_syn_backlog = 8192
net.core.somaxconn = 8192

# 增加连接跟踪表
net.netfilter.nf_conntrack_max = 1048576
net.netfilter.nf_conntrack_tcp_timeout_established = 7200

# 应用配置
sysctl -p

# /etc/security/limits.conf
haproxy  soft  nofile  200000
haproxy  hard  nofile  200000
root     soft  nofile  200000
root     hard  nofile  200000

HAProxy层面，maxconn在global和frontend中需匹配设置，避免单前端连接数超过服务端处理能力。对于SSL终结场景（HTTPS），建议使用nbthread参数启用多线程并将CPU亲和性绑定到特定核心：

global
    nbthread 4          # 与CPU核心数匹配
    cpu-map auto:1/1-4  # 自动绑定线程到CPU核心
    maxconn 20000

# SSL终结配置
frontend web_https
    bind *:443 ssl crt /etc/haproxy/certs/domain.pem alpn h2,http/1.1
    mode http
    # HTTP/2多路复用
    http-response set-header Strict-Transport-Security "max-age=31536000"
    default_backend web_servers

这套Keepalived + HAProxy双机热备方案已在实际生产环境中承载单日数亿请求量。核心运维要点：定期演练故障切换流程，确保团队熟悉切换流程和恢复操作；监控VRRP心跳包传输延迟异常，提前发现网络抖动风险；保持两节点配置严格一致，使用配置管理工具（Ansible/Puppet）统一分发配置变更。