现场Keepalived配置测试

2021-11-29

字数统计: 1.3k | 阅读时长≈ 6 分钟

keepalived配置

配置主备，抢占和非抢占模式下的HA（高可用）

yum 安装 keepalived
为了通过web来测试，安装nginx，（nginx使用tengine的包）
1
2
yum -y install keepalived
yum -y install tengine

分别配置主副节点的keepalived.conf 配置，目前仅做简单配置

! Configuration File for keepalived

global_defs {
   router_id LVS_DEVEL  # 删除了email通知
}

# 配置一个vrrp_instacne 名字随便
vrrp_instance VI_1 {
    # 分别是主备
    state MASTER/BACKUP
    # 选择绑定那张网卡，需要是在用的
    interface eth1
    # 主备id要一样
    virtual_router_id 51
    # 主备优先级值一般设置不一样，但是应该遵循master_priority - backup_priority < weight (如果有设置vrrp_script，且设置了weight的情况下)
    priority 100
    # 通信时间
    advert_int 1
    # 认证方式，密码一样，方便主备通信
    authentication {
        auth_type PASS
        auth_pass 1111
    }
    # 绑定的VIP(虚拟IP，即对外用的IP地址)，可以是多个，每行一个。
    virtual_ipaddress {
        192.168.10.193
    }
}

分别启动服务，然后通过浏览器访问192.168.10.193:15003（我设置nginx监控15003端口，没有使用80），来确定keepalive生效

当master节点挂掉重启后，会自动抢占成为master

主节点关掉后，主节点日志：
Apr 14 00:02:12 ora12c Keepalived[1944]: Stopping Keepalived v1.2.13 (03/19,2015)
Apr 14 00:02:12 ora12c Keepalived_vrrp[1946]: VRRP_Instance(VI_1) sending 0 priority


备节点的状态：
Apr 13 16:14:00 interface16 Keepalived_vrrp[14033]: VRRP_Instance(VI_1) Transition to MASTER STATE
Apr 13 16:14:01 interface16 kernel: ACPI: No handler for Region [POWR] (ffff881821ff72b8) [IPMI]
Apr 13 16:14:01 interface16 Keepalived_vrrp[14033]: VRRP_Instance(VI_1) Entering MASTER STATE
Apr 13 16:14:01 interface16 Keepalived_vrrp[14033]: VRRP_Instance(VI_1) setting protocol VIPs.
Apr 13 16:14:01 interface16 Keepalived_vrrp[14033]: VRRP_Instance(VI_1) Sending gratuitous ARPs on em1 for 192.168.10.193
Apr 13 16:14:01 interface16 Keepalived_healthcheckers[14032]: Netlink reflector reports IP 192.168.10.193 added
Apr 13 16:14:06 interface16 Keepalived_vrrp[14033]: VRRP_Instance(VI_1) Sending gratuitous ARPs on em1 for 192.168.10.193

备用节点ip情况：
2: em1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq state UP qlen 1000
    link/ether b8:2a:72:de:40:c3 brd ff:ff:ff:ff:ff:ff
    inet 192.168.10.170/24 brd 192.168.10.255 scope global em1
    inet 192.168.10.193/32 scope global em1
    inet6 fe80::ba2a:72ff:fede:40c3/64 scope link
       valid_lft forever preferred_lft forever
此时VIP已经漂移到170这台机器上了

---

主节点：重启后，messages的日志
Apr 13 23:47:43 ora12c Keepalived_healthcheckers[1945]: Using LinkWatch kernel netlink reflector...
Apr 13 23:47:43 ora12c Keepalived_vrrp[1946]: VRRP_Instance(VI_1) Transition to MASTER STATE
Apr 13 23:47:43 ora12c Keepalived_vrrp[1946]: VRRP_Instance(VI_1) Received lower prio advert, forcing new election
Apr 13 23:47:44 ora12c Keepalived_vrrp[1946]: VRRP_Instance(VI_1) Entering MASTER STATE

此时主节点ip情况
2: eth1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq state UP qlen 1000
    link/ether 00:0c:29:10:4f:bc brd ff:ff:ff:ff:ff:ff
    inet 192.168.10.191/24 brd 192.168.10.255 scope global eth1
    inet 192.168.10.193/32 scope global eth1
设置的对外IP（即VIP）绑定到 191这台机器上了


备节点此时状态：
Apr 13 16:15:22 interface16 Keepalived_vrrp[14033]: VRRP_Instance(VI_1) Received higher prio advert
Apr 13 16:15:22 interface16 Keepalived_vrrp[14033]: VRRP_Instance(VI_1) Entering BACKUP STATE
Apr 13 16:15:22 interface16 Keepalived_vrrp[14033]: VRRP_Instance(VI_1) removing protocol VIPs.
Apr 13 16:15:22 interface16 Keepalived_healthcheckers[14032]: Netlink reflector reports IP 192.168.10.193 removed
Apr 13 16:15:24 interface16 ntpd[2105]: Deleting interface #7 em1, 192.168.10.193#123, interface stats: received=0, sent=0, dropped=0, active_time=81 secs

此时备节点ip情况：
2: em1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq state UP qlen 1000
    link/ether b8:2a:72:de:40:c3 brd ff:ff:ff:ff:ff:ff
    inet 192.168.10.170/24 brd 192.168.10.255 scope global em1
    inet6 fe80::ba2a:72ff:fede:40c3/64 scope link
       valid_lft forever preferred_lft forever

非抢占模式

非抢占模式配置：

主备文件中都在vrrp_instance中添加参数nopreempt，表示不争抢

在vrrp_instance中将state都改为BACKUP



修改完成后先停用主节点，重启备用节点的keepalived进程，
目的是，让备用节点先拿到VIP，验证启动主节点进程后，VIP不会漂移到主节点  --ok，因为配置了nopreempt参数，所以主节点不会抢占，

此时备用节点ip情况：
2: em1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq state UP qlen 1000
    link/ether b8:2a:72:de:40:c3 brd ff:ff:ff:ff:ff:ff
    inet 192.168.10.170/24 brd 192.168.10.255 scope global em1
    inet 192.168.10.193/32 scope global em1

杀掉备用节点进程
[root@interface16 ~]# killall keepalived

此时主节点会变成MASTER主节点，日志：
Apr 14 00:20:16 ora12c Keepalived_vrrp[2213]: VRRP_Instance(VI_1) Transition to MASTER STATE
Apr 14 00:20:17 ora12c Keepalived_vrrp[2213]: VRRP_Instance(VI_1) Entering MASTER STATE
Apr 14 00:20:17 ora12c Keepalived_vrrp[2213]: VRRP_Instance(VI_1) setting protocol VIPs.
Apr 14 00:20:17 ora12c Keepalived_vrrp[2213]: VRRP_Instance(VI_1) Sending gratuitous ARPs on eth1 for 192.168.10.193
Apr 14 00:20:17 ora12c Keepalived_healthcheckers[2212]: Netlink reflector reports IP 192.168.10.193 added
Apr 14 00:20:22 ora12c Keepalived_vrrp[2213]: VRRP_Instance(VI_1) Sending gratuitous ARPs on eth1 for 192.168.10.193

主节点ip情况：
2: eth1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq state UP qlen 1000
    link/ether 00:0c:29:10:4f:bc brd ff:ff:ff:ff:ff:ff
    inet 192.168.10.191/24 brd 192.168.10.255 scope global eth1
    inet 192.168.10.193/32 scope global eth1
    inet6 fe80::20c:29ff:fe10:4fbc/64 scope link
       valid_lft forever preferred_lft foreve


综上，当设置nopreempt非抢占模式时，会根据priority大小选举一个MASTER节点，但是当MASTER节点的keepalived进程出现问题，即使恢复后，就不会再主动抢占MASTER，
但是如果其他备用节点的keepalived进程都出现问题时，该节点会自动接管成为MASTER。
--所以，这也是网上教程中都在提的双主模式下，通过杀掉keepalived进程，实现出现问题时完成主备切换。

版权声明： 本博客所有文章除特别声明外，著作权归作者所有。转载请注明出处！