Keepalived 多主模型与 LVS 高可用

一.Keepalived多主模型

Keepalived多主模型概念

如上图，keepalived主从架构性能损耗较严重，如果业务分类明确，则可以配置keepalived多主模型降低损耗，两台keepalived互为主备，如：订单业务走keepalived1，keepalived2做备，商品业务走keepalived2，keepalived1做备。也就是keepalived1和keepalived2各拿一份vip

Keepalived多主模型配置

keepalived1订单业务：

vrrp_instance VI_2 {state BACKUP   #设置为BACKUPinterface ens33virtual_router_id 55priority 80advert_int 1authentication {auth_type PASSauth_pass 111156}virtual_ipaddress {192.168.80.40/24192.168.80.41/24192.168.80.42/24}notify_master "/root/sendemail.sh master"notify_backup "/root/sendemail.sh backup"notify_fault  "/root/sendemail.sh fault"
}

keepalived1商品业务：

vrrp_instance VI_1 {state MASTER   #设置为MASTER   interface ens33virtual_router_id 51priority 100advert_int 1authentication {auth_type PASSauth_pass 1111}virtual_ipaddress {192.168.80.50/24192.168.80.51/24192.168.80.52/24}notify_master "/root/sendemail.sh master"notify_backup "/root/sendemail.sh backup"notify_fault  "/root/sendemail.sh fault"
}

keepalived2订单业务：

vrrp_instance VI_2 {state MASTERinterface ens33virtual_router_id 55priority 80advert_int 1authentication {auth_type PASSauth_pass 111156}virtual_ipaddress {192.168.80.40/24192.168.80.41/24192.168.80.42/24}notify_master "/root/sendemail.sh master"notify_backup "/root/sendemail.sh backup"notify_fault  "/root/sendemail.sh fault"}

keepalived2商品业务：

vrrp_instance VI_1 {state MASTERinterface ens33virtual_router_id 51priority 80advert_int 1authentication {auth_type PASSauth_pass 1111}virtual_ipaddress {192.168.80.50/24192.168.80.51/24192.168.80.52/24}notify_master "/root/sendemail.sh master"notify_backup "/root/sendemail.sh backup"notify_fault  "/root/sendemail.sh fault"}

验证

如上图，keepalived是商品业务的主节点，拿到商品业务的vip，keepalived2是订单业务的主节点，拿到订单业务的vip，并且互为主备。

如果此时一个keepalived坏了，另一个keepalived将拥有两个业务的vip

二.实现IPVS 高可用

在分布式系统架构中，高可用性设计始终是核心命题。当我们使用 Keepalived 实现 VIP 飘移时，本质上只是解决了 IP 层的可用性问题，这就像为建筑搭建了稳固的地基，但要让大楼真正运转起来，还需要在服务层构建完整的容错机制。LVS（Linux Virtual Server）正是实现这一目标的关键组件，它与 Keepalived 的 VRRP 协议形成了完美的能力互补。

虚拟服务器配置

virtual_server IP port {                # 定义虚拟服务器，指定监听的IP和端口delay_loop<INT>                    # 健康检查的时间间隔（单位：秒）lib_algo rr|wr1|cln1c|lib|cln1dh    # 负载均衡算法（rr=轮询，wr1=加权轮询，cln1c=最少连接，cln1dh=目标地址哈希）lib_kind NAT|ON|TUN                 # 数据包转发模式（NAT=网络地址转换，TUN=隧道模式，ON=直接路由）persistence_timeout <INT>           # 会话保持时间（单位：秒，0表示不启用）protocol TCP|USIP|SCTP              # 协议类型（TCP/USIP[UDP]/SCTP）sorry_server <IPADDR> <PORT>        # 备用服务器地址，当所有真实服务器宕机时启用real_server <IPADDR> <PORT> {       # 定义真实服务器（后端节点）weight <INT>                    # 服务器权重（权重越高分配的请求越多）notify_up <STRING>              # 服务器上线时触发的脚本/命令notify_down <STRING>            # 服务器下线时触发的脚本/命令# 健康检查方法（只能选其一）：HTTP_GET|SSL_GET {              # HTTP/HTTPS 检查（需配置具体URL和状态码）url { path <PATH> }         # 示例：url { path "/health" status_code 200 }}TCP_CHECK { ... }               # TCP端口检查（默认检测端口连通性）SNTP_CHECK { ... }              # SNTP协议检查（用于时间服务器）MISC_CHECK { ... }             # 自定义脚本检查（需指定外部脚本路径）}
}

例如：

vrrp_instance VI_1 {state MASTERinterface eth0virtual_router_id 51priority 100advert_int 1authentication {auth_type PASSauth_pass 1111}virtual_ipaddress {192.168.1.100}
}virtual_server 192.168.1.100 80 {delay_loop 6lb_algo wrrlb_kind DRprotocol TCPreal_server 192.168.1.101 80 {weight 3TCP_CHECK {connect_timeout 3nb_get_retry 3delay_before_retry 3}}
}

该配置展示了