容器云负载均衡之五：使用keepalived管理基于容器的IPVS

一、前言

IPVS的director上会绑定一个VIP，这个VIP是下游客户的唯一接入点，所以这个VIP需要满足高可用性的要求。另外，不同的VIP策略需要根据后端real server的部署状态进行更新。

Keepalived是一个开源软件，主要提供loadbalancing（负载均衡）和 high-availability（高可用）功能，负载均衡实现需要依赖Linux的虚拟服务内核模块（IPVS），而高可用是通过VRRP协议实现多台机器之间的故障转移服务。 

转载自https://blog.csdn.net/cloudvtech

二、配置keepalived

2.1 测试环境配置 

Director node 1: 200.222.0.73
Director node 2: 200.222.0.74
Real Server 1: 200.222.0.87
Real Server 2: 200.222.0.89
VIP: 200.222.0.113 

2.2 设置iptables允许VRRP协议交互

iptables -t filter -A IN_public_allow -p udp -m udp --dport 112  -m conntrack --ctstate NEW -j ACCEPT

2.3 在CentOS容器内安装keepalived

docker pull centos
docker run -td --privileged --net=host --name=keepalived centos
docker exec -it keepalived bash
yum install -y net-tools iproute
yum install -y keepalived ipvsadm

将容器打包成docker image

2.4 在两个Director node启动容器

2.5 配置Director node 1的keepalived

/etc/keepalived/keepalived.conf

global_defs {
   notification_email {
     [email protected]
     [email protected]
   }
   notification_email_from [email protected]
   smtp_server localhost
   smtp_connect_timeout 30
}

vrrp_instance VI_1 {
    state MASTER
    interface ens192
    virtual_router_id 51
    priority 100
    advert_int 1
    unicast_src_ip 200.222.0.73
    unicast_peer {
        200.222.0.74
    }
    authentication {
        auth_type PASS
        auth_pass 1111
    }
    virtual_ipaddress {
        200.222.0.113
    }
}

virtual_server 200.222.0.113 80 {
    delay_loop 6
    lb_algo wlc
    lb_kind DR
    persistence_timeout 600
    protocol TCP

    real_server 200.222.0.87 80 {
        weight 100
        TCP_CHECK {
            connect_timeout 3
        }
    }
    real_server 200.222.0.89 80 {
        weight 100
        TCP_CHECK {
            connect_timeout 3
        }
    }
}

2.6 配置Director node 2的keepalived

/etc/keepalived/keepalived.conf

global_defs {
   notification_email {
     [email protected]
     [email protected]
   }
   notification_email_from [email protected]
   smtp_server localhost
   smtp_connect_timeout 30
}

vrrp_instance VI_1 {
    state BACKUP
    interface ens192
    virtual_router_id 51
    priority 99
    advert_int 1
    unicast_src_ip 200.222.0.74
    unicast_peer {
        200.222.0.73
    }
    authentication {
        auth_type PASS
        auth_pass 1111
    }
    virtual_ipaddress {
        200.222.0.113
    }
}

virtual_server 200.222.0.113 80 {
    delay_loop 6
    lb_algo wlc
    lb_kind DR
    persistence_timeout 600
    protocol TCP

    real_server 200.222.0.87 80 {
        weight 100
        TCP_CHECK {
            connect_timeout 3
        }
    }
    real_server 200.222.0.89 80 {
        weight 100
        TCP_CHECK {
            connect_timeout 3
        }
    }
}

2.7 启动Director node 1/2的keepalived

可以看到keepalived仅仅在node1上绑定了VIP

2.8 查看Director node 1/2的VRRP协议交互信息

tcpdump -vvv -an -i ens192 | grep "vrid 51" 

可以看到active node持续向backup node发送VRRP协议信息

2.9 使用keepalived进行IPVS director failover

在Director node1运行pkill keepalived 

在Director node1可以看到VIP被绑定

也可以看到现在是新的active node(Director node 2)持续向node1发送VRRP协议信息

2.10 重新启动node1的keepalived，node1继续成为active的IPVS director

转载自https://blog.csdn.net/cloudvtech