利用LVS+Keepalived实现MySQL高可用(一)

利用LVS+Keepalived实现MySQL高可用

MySQL架构

MySQL的高可用，由一对互为主从的MySQL服务器组成，平时只有主库提供服务，备库不提供服务。当主库停止服务时，服务自动切换到备库。主备库通过双向复制，保证数据一致性。主备库以VIP对外统一服务接口。

Keepalived

MySQL对外服务接口的VIP，由Keepalived来实现心跳检查和动态漂移。因此MySQL和Keepalived部署在同一台服务器上，每台Keepalived配置的realserver即自身机器的MySQL服务。

环境安装

MySQL安装

两台服务器都安装MySQL，搭建主主复制架构。具体搭建略。

LVS+Keepalived安装

LVS源码安装

ln -s /usr/src/kernels/2.6.32-279.14.1.el6.x86_64/ /usr/src/linux
tar zxvf ipvsadm-1.24.tar.gz
cd ipvsadm-1.24
make && make install

Keepalived源码安装

安装依赖包

yum -y install popt-devel

安装Keepalived

tar zxvf keepalived-1.1.19.tar.gz
cd keepalived-1.1.19
./configure --prefix=/usr/local/keepalived
make && make install
cp /usr/local/keepalived/sbin/keepalived /usr/sbin/
cp /usr/local/keepalived/etc/sysconfig/keepalived /etc/sysconfig/
cp /usr/local/keepalived/etc/rc.d/init.d/keepalived /etc/init.d/
mkdir /etc/keepalived

添加自启动

mysqld和keepalived自启动

chkconfig --add mysqld
chkconfig --add keepalived
chkconfig  mysqld on
chkconfig  keepalived on

Keepalived配置

IP配置

这里的主备库IP配置如下

库	IP	服务VIP
主	10.0.0.77	10.0.0.76
备	10.0.0.78

Keepalived配置文件keepalived配置文件/etc/keepalived/keepalived.conf，主库为例

global_defs {
          router_id LVS_1
}
vrrp_sync_group test {
       group {
             mysql_ha
       }
}

vrrp_instance mysql_ha {
       state BACKUP         ## 主备库都是BACKUP，防止主库重启后，抢夺VIP资源
       interface eth0     ##  HA地址网卡
       virtual_router_id 61
       priority 150           ##  优先级，主库比备库高
       nopreempt           ## 非抢夺，在主库上配置，表示不抢夺VIP
       advert_int 1          ##  主备心跳检查间隔
       authentication {
              auth_type PASS
              auth_pass 1111
       }
       virtual_ipaddress {     ##  VIP，主备库一样
                     10.0.0.76
       }
   ##  以下脚本是Keepalived改变状态时触发的动作，分别是变为backup、master、fault状态
   ## 这是个简单的报警脚本而已
   notify_backup "/usr/local/keepalived_script/keepalived_notify.sh BACKUP"
   notify_master "/usr/local/keepalived_script/keepalived_notify.sh MASTER"
   notify_fault "/usr/local/keepalived_script/keepalived_notify.sh FAULT"
   notify "/usr/local/keepalived_script/keepalived_notify.sh "
}
virtual_server 10.0.0.76 3306 {
       delay_loop 2
       lb_algo rr
       lb_kind DR
       persistence_timeout 20
       protocol TCP
       real_server 10.0.0.77 3306 {   ## keepalived的realserver，主备库都取自身的mysql服务
              weight 3
              notify_down "/usr/local/keepalived_script/keepalived_notify.sh MYSQL_DOWN"    ##  检测到realserver服务不可用时的动作
              TCP_CHECK {
                     connect_timeout 3
                     nb_get_retry 3
                     delay_before_retry 3
                     connect_port 3306
              }
       }
}

测试

测试效果

从以下4个方面测试HA效果
    keepalived down
    mysqld down
    网络故障
    机器宕机

测试步骤和结果

keepalived down

步骤：kill master上的keepalived进程

结果：

>>slave的keepalived通知切换为master

mysqld down

步骤：kill master上的mysqld进程，或者提高master服务器的负载

结果：

>>master的keepalived通知因为mysqld down而停止keepalived

>>slave的keepalived通知切换为master

网络故障

步骤：master上ifconfig eth0 down，或者service network restart

结果：

>>master的keepalived通知变为fault状态

>>slave的keepalived通知切换为master

 

机器宕机

步骤：master机器重启

结果：

>>slave的keepalived通知切换为master

问题分析

1，keepalived切换为master时，arping失败

有的环境测试，keepalived切换为master时，没有成功发送arping，导致VIP不同。所以给出了arping脚本

2，重启网卡，master变为fault状态，但没继续变为backup状态。

检查日志发现有错误信息

Keepalived_vrrp: Netlink: filter function error

Keepalived_healthcheckers: Netlink: filter function error

查找网络资料得知，keepalive不支持网络设备热插拔。ip link可以看出设备id变化了。

此时需要reload keepalived或者直接restart。

办法：网卡发生重启后，需检查keepalived是否正常

3，keepalived异常终止，VIP没有解除绑定，导致及时重启keepalived后，这个VIP也不能正常绑定和解除绑定

猜测原因是异常终止keepalived后，VIP没解除绑定，重启以后keepalived不识别未解除绑定的VIP。

办法：机器发生重启后需检查keepalived是否正常

4，ipvsadm使用yum安装，keepalived使用源码安装，keepalived不能启动healthchecker进程

keepalived安装时指定kernel目录：./configure --prefix=/usr/local/keepalived --with-kernel-dir=/usr/src/kernels/2.6.18-308.1.1.el5-x86_64

5，关于virtual_router_id

virtual_router_id用来标识一对keepalived主从机的身份，不同keepalived对的这个id不能一样。取值范围1~255。如果不同组keepalived配置了同一个id，则会有如下报错：

Apr  7 14:31:11 localhost Keepalived_vrrp[22980]: ip address associated with VRID not present in received packet : 10.0.0.94

Apr  7 14:31:11 localhostKeepalived_vrrp[22980]: one or more VIP associated with VRID mismatch actual MASTER advert

Apr  7 14:31:11 localhost Keepalived_vrrp[22980]: bogus VRRP packet received on eth0.181 !!!

Apr  7 14:31:11 localhost Keepalived_vrrp[22980]: VRRP_Instance(mysql_ha) ignoring received advertisment...

6，端口转发

virtual_server配置的端口，与real_server配置的端口必须一样？