linux - keepalived 和两个事件服务器

Question

我正在尝试使用 keepalived 配置两个事件负载均衡器服务器。

它与标准配置略有不同，我们有两台服务器和一个虚拟 IP。

我想要的只是当 loadbalancer_1 出现故障时，loadbalancer_2 会获取他的 IP 地址，反之亦然。

loadbalancer_1 IP 地址:xx.237.193.132/27loadbalancer_2 IP 地址:xx.237.193.134/27

当我第一次启动 loadbalancer_1 时，它也会使用 loadbalancer_2 IP 地址，因为在该特定时刻 keepalived 尚未在 loadbalancer_2 上启动。但当我启动 loadbalancer_2 时，loadbalancer_1 上仍然有他的 IP 地址。因此，loadbalancer_1。是两个 IP 地址的主站。

当我首先启动 loadbalancer_2 时，也会发生同样的情况，因此较早启动的那些服务器将成为两个 IP 地址的主服务器。

登录loadbalancer_1:

Jul 19 19:18:00 001 Keepalived_vrrp[9751]: Registering Kernel netlink reflector
Jul 19 19:18:00 001 Keepalived_vrrp[9751]: Registering Kernel netlink command channel
Jul 19 19:18:00 001 Keepalived_vrrp[9751]: Registering gratuitous ARP shared channel
Jul 19 19:18:00 001 Keepalived_vrrp[9751]: Opening file '/etc/keepalived/keepalived.conf'.
Jul 19 19:18:00 001 Keepalived_healthcheckers[9750]: Registering Kernel netlink reflector
Jul 19 19:18:00 001 Keepalived_healthcheckers[9750]: Registering Kernel netlink command channel
Jul 19 19:18:00 001 Keepalived_healthcheckers[9750]: Opening file '/etc/keepalived/keepalived.conf'.
Jul 19 19:18:00 001 Keepalived_vrrp[9751]: Configuration is using : 71869 Bytes
Jul 19 19:18:00 001 Keepalived_vrrp[9751]: Using LinkWatch kernel netlink reflector...
Jul 19 19:18:00 001 Keepalived_vrrp[9751]: VRRP_Instance(web_loadbalancer_2) Entering BACKUP STATE
Jul 19 19:18:00 001 Keepalived_healthcheckers[9750]: Configuration is using : 12587 Bytes
Jul 19 19:18:00 001 Keepalived_healthcheckers[9750]: Using LinkWatch kernel netlink reflector...
Jul 19 19:18:01 001 Keepalived_vrrp[9751]: VRRP_Instance(web_loadbalancer_1) Transition to MASTER STATE
Jul 19 19:18:02 001 Keepalived_vrrp[9751]: VRRP_Instance(web_loadbalancer_1) Entering MASTER STATE
Jul 19 19:18:04 001 Keepalived_vrrp[9751]: VRRP_Instance(web_loadbalancer_2) Transition to MASTER STATE
Jul 19 19:18:05 001 Keepalived_vrrp[9751]: VRRP_Instance(web_loadbalancer_2) Entering MASTER STATE

登录loadbalancer_2:

Jul 19 19:18:08 002 Keepalived_vrrp[468]: Registering Kernel netlink reflector
Jul 19 19:18:08 002 Keepalived_vrrp[468]: Registering Kernel netlink command channel
Jul 19 19:18:08 002 Keepalived_vrrp[468]: Registering gratuitous ARP shared channel
Jul 19 19:18:08 002 Keepalived_vrrp[468]: Opening file '/etc/keepalived/keepalived.conf'.
Jul 19 19:18:08 022 Keepalived_vrrp[468]: Configuration is using : 71869 Bytes
Jul 19 19:18:08 002 Keepalived_vrrp[468]: Using LinkWatch kernel netlink reflector...
Jul 19 19:18:08 002 Keepalived_vrrp[468]: VRRP_Instance(web_loadbalancer_1) Entering BACKUP STATE
Jul 19 19:18:08 002 Keepalived_healthcheckers[467]: Registering Kernel netlink reflector
Jul 19 19:18:08 002 Keepalived_healthcheckers[467]: Registering Kernel netlink command channel
Jul 19 19:18:08 002 Keepalived_healthcheckers[467]: Opening file '/etc/keepalived/keepalived.conf'.
Jul 19 19:18:08 002 Keepalived_healthcheckers[467]: Configuration is using : 12587 Bytes
Jul 19 19:18:08 002 Keepalived_healthcheckers[467]: Using LinkWatch kernel netlink reflector...
Jul 19 19:18:09 002 Keepalived_vrrp[468]: VRRP_Instance(web_loadbalancer_2) Transition to MASTER STATE
Jul 19 19:18:10 002 Keepalived_vrrp[468]: VRRP_Instance(web_loadbalancer_2) Entering MASTER STATE
Jul 19 19:18:11 002 Keepalived_vrrp[468]: VRRP_Instance(web_loadbalancer_2) Received lower prio advert, forcing new election

loadbalancer_1的配置:

global_defs {
    router_id loadbalancer_1
    lvs_id loadbalancer_1
}
vrrp_instance web_loadbalancer_1 {
    state MASTER
    interface eth0
    virtual_router_id 10
    priority 110
    advert_int 1
    virtual_ipaddress {
        xx.237.193.132/27 dev eth0
    }
    authentication {
        auth_type PASS
        auth_pass 45f3fgv1
    }
}
vrrp_instance web_loadbalancer_2 {
    state BACKUP
    interface eth0
    virtual_router_id 20
    priority 200
    smtp_alert
    advert_int 1
    virtual_ipaddress {
        xx.237.193.134/27 dev eth0
    }
    authentication {
        auth_type PASS
        auth_pass 45f3fgv2
    }
}

loadbalancer_2的配置:

global_defs {
    router_id loadbalancer_2
    lvs_id loadbalancer_2
}
vrrp_instance web_loadbalancer_1 {
    state BACKUP
    interface eth0
    virtual_router_id 10
    priority 100
    advert_int 1
    virtual_ipaddress {
        xx.237.193.132/27 dev eth0
    }
    authentication {
        auth_type PASS
        auth_pass 45f3fgv1
    }
}
vrrp_instance web_loadbalancer_2 {
    state MASTER
    interface eth0
    virtual_router_id 20
    priority 210
    advert_int 1
    virtual_ipaddress {
        xx.237.193.134/27 dev eth0
    }
    authentication {
        auth_type PASS
        auth_pass 45f3fgv2
    }
}

Answer 1

我建议向两个 MASTER 部分添加 preempt_delay 10。

查看 server2 的最后一个日志条目

Jul 19 19:18:11 002 Keepalived_vrrp[468]: VRRP_Instance(web_loadbalancer_2) Received lower prio advert, forcing new election

它似乎刚刚意识到自己可以接任 VRRP 主设备。如果没有发生这种情况，也许稍后可以查看日志。

第一台服务器启动是正常的，因为没有其他服务器而假设为主服务器，然后当第二台服务器变为事件状态时，vrrp 重新选举应该选择配置最高优先级的服务器。