Mysql双主互备+keeplived高可用架构介绍

一、Mysql双主互备+keeplived高可用架构介绍

Mysql主从复制架构可以在很大程度保证Mysql的高可用，在一主多从的架构中还可以利用读写分离将读操作分配到从库中，减轻主库压力。但是在这种架构中，主库出现故障时需要手动将一台从库提升为主库。在对写操作要求较高的环境中，主库故障在主从架构中会成为单点故障。因此需要主主互备架构，避免主节点故障造成写操作失效。

在双主互备的架构中，每台Mysql都充当主服务器，同时充当对方的从服务器。在任意一台服务器上的写操作都会被复制到另一台服务器上，从而保证了数据的可靠性。

在双主互备的基础上加上keeplived，在其中一台机器上绑定虚拟ip（VIP）。利用vip统一对外服务，可以避免在两个节点同时写数据造成冲突。同时当keeplived主节点发生故障时，keeplived会自动将VIP切换到备节点上，从而实现主服务器的高可用。

wKiom1fG93iwjIn9AAC06rhxUqo862.jpg-wh_50
Mysql双主互备+keeplived高可用架构图

二、Mysql双主互备架构部署

在上一篇Mysql主从复制操作中已经完成了Mysql主从架构的部署。在此基础上，需要在原Master上开启relay-log，在原Slave服务器上开启log-bin，同时在Mysql248上指定Mysql249为自己的主服务器并开启slave即可。

1.开启原Master上的relay-log，指定不复制的库
1
2
3
4
5
6
7
8
9
10
[root@Mysql-248 mysql-5.6.30]# grep -A8 'mysqld' my.cnf
[mysqld]
port = 3306
server_id = 1
log-bin=mysql-bin

以下部分为在原基础上新增的内容

relay-log = mysql-relay-bin
replicate-wild-ignore-table=mysql.%
replicate-wild-ignore-table=test.%
replicate-wild-ignore-table=information_schema.%

2.开启原Slave上的log-bin，指定不复制的库
[root@Mysql-249 mysql-5.6.30]# grep -A8 'mysqld' my.cnf
1
2
3
4
5
6
7
8
9
[mysqld]
port = 3306
server_id = 2
relay-log = mysql-relay-bin

以下部分为在原基础上新增的内容

log-bin=mysql-bin
replicate-wild-ignore-table=mysql.%
replicate-wild-ignore-table=test.%
replicate-wild-ignore-table=information_schema.%

3.在Mysql248服务器上指定Mysql249为自己的主服务器并开启slave
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23

① 在Mysql249上查看当前master信息，并建立复制用户

mysql> show master status;
+------------------+----------+--------------+------------------+-------------------+
| File | Position | Binlog_Do_DB | Binlog_Ignore_DB | Executed_GtidSet |
+------------------+----------+--------------+------------------+-------------------+
| mysql-bin.000001 | 493 | | | |
+------------------+----------+--------------+------------------+-------------------+
1 row in set (0.00 sec)
mysql> grant replication slave on ._ to 'repl_user'@'192.168.175.%' identified by '123456';
Query OK, 0 rows affected (0.00 sec)
mysql> flush privileges;
Query OK, 0 rows affected (0.00 sec)

② 在Mysql248上指定Mysql249为自己的主服务器，开启slave

mysql> change master to \
-> master_host='192.168.175.249',
-> master_user='repl_user',
-> master_password='123456',
-> master_log_file='mysql-bin.000001',
-> master_log_pos=493;
Query OK, 0 rows affected, 2 warnings (0.02 sec)
mysql> start slave;
Query OK, 0 rows affected (0.03 sec)

4.检验双主互备
①通过分别在两台服务器上使用show slave status\G，查询主库信息以及IO进程、SQL进程工作状态。若两台服务器的查询结果都为Slave_IO_Running: Yes，Slave_SQL_Running: Yes；则表示当前双主互备状态正常。
②在Mysql248数据库上建库建表，检查Mysql249上是否同步正常；然后在Mysql249上建库建表，检查Mysql248上是否同步正常。

三、Keeplived高可用部署
以下操作在两台机器上基本相同，不同之处请留意备注标识

1.keeplived编译安装
1
2
3
4
5
6
7
8
9
10

编译操作两台机器一样

[root@Mysql-248 src]# tar -zxf keepalived-1.2.12.tar.gz
[root@Mysql-248 src]# cd keepalived-1.2.12
[root@Mysql-248 keepalived-1.2.12]# ./configure --sysconf=/etc --with-kernel-dir=/usr/src/kernels/2.6.32-573.el6.x86_64/
[root@Mysql-248 keepalived-1.2.12]# make
[root@Mysql-248 keepalived-1.2.12]# make install
[root@Mysql-248 keepalived-1.2.12]# ln -s /usr/local/sbin/keepalived /sbin/
[root@Mysql-248 keepalived-1.2.12]# chkconfig --add keepalived
[root@Mysql-248 keepalived-1.2.12]# chkconfig --level 35 keepalived on

2.keeplived配置
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40

配置文件中需要注意主备节点的priority与nopreempt配置不同

[root@Mysql-248 ~]# cd /etc/keepalived/
[root@Mysql-248 keepalived]# cat keepalived.conf
! Configuration File for keepalived
global_defs {
notification_email {
acassen@firewall.loc
failover@firewall.loc
sysadmin@firewall.loc
}
notification_email_from Alexandre.Cassen@firewall.loc
smtp_server 192.168.200.1
smtp_connect_timeout 30
router_id MySQLHA_DEVEL
}
vrrp_script check_mysqld {
script "/etc/keepalived/mysqlcheck/check_slave.sh" #定义检测mysql复制状态的脚本
interval 2
}
vrrp_instance HA_1 {
state BACKUP #两台mysql服务器均配置为BACKUP
interface eth0
virtual_router_id 80
priority 100 #备用节点设置为90
advert_int 2
nopreempt #不抢占模式，只在优先级高的机器上设置即可，备用节点不加此参数
authentication {
auth_type PASS
auth_pass qweasdzxc
}

track_script {
    check_mysqld #调用mysql检测脚本
}
virtual_ipaddress {
    192.168.175.254/24 dev eth0     #mysql的对外服务IP，即VIP
}

}
[root@Mysql-248 keepalived]#

3.配置Mysql检测脚本
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38

检测脚本主备节点都一样

[root@Mysql-248 keepalived]# cat /etc/keepalived/mysqlcheck/check_slave.sh

!/bin/bash

This scripts is check for Mysql Slave status

Mysqlbin=/usr/local/mysql-5.6.30/bin/mysql
user=root
pw='qwe123``'
port=3306
host=127.0.0.1
sbm=120

Check for $Mysqlbin

if [ ! -f $Mysqlbin ];then
echo 'Mysqlbin not found,check the variable Mysqlbin'
exit 99
fi

Get Mysql Slave Status

IOThread=$Mysqlbin -h $host -P $port -u$user -p$pw -e 'show slave status\G' 2>/dev/null|grep 'Slave_IO_Running:'|awk '{print $NF}'
SQLThread=$Mysqlbin -h $host -P $port -u$user -p$pw -e 'show slave status\G' 2>/dev/null|grep 'Slave_SQL_Running:'|awk '{print $NF}'
SBM=$Mysqlbin -h $host -P $port -u$user -p$pw -e 'show slave status\G' 2>/dev/null|grep 'Seconds_Behind_Master:'|awk '{print $NF}'

Check if the mysql run

if [[ -z "$IOThread" ]];then
exit 1
fi

Check if the thread run

if [[ "$IOThread" = "No" || "$SQLThread" = "No" ]];then
exit 1
elif [[ $SBM -ge $sbm ]];then
exit 1
else
exit 0
fi
[root@Mysql-248 keepalived]#

4.启动keeplived
1
2
3
4
5
6
7

分别在Mysql248、Mysql249上启动keeplived

需要注意由于设置了nopreempt非强占模式，所以先启动的节点会成为主节点，后启动的成为备节点。

[root@Mysql-248 keepalived]# /etc/init.d/keepalived start
Starting keepalived: [ OK ]
[root@Mysql-249 keepalived]# /etc/init.d/keepalived start
Starting keepalived: [ OK ]

四、检验Mysql双主互备+keeplived高可用架构

1.启动后检查message日志信息
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19

在248上查看message日志，发现启动时节点进入MASTER状态，并且绑定了192.168.175.254的VIP

[root@Mysql-248 ~]# tail -f /var/log/messages
Sep 1 08:16:41 Mysql-248 Keepalived_vrrp[71143]: VRRP_Instance(HA_1) Transition to MASTER STATE
Sep 1 08:16:43 Mysql-248 Keepalived_vrrp[71143]: VRRP_Instance(HA_1) Entering MASTER STATE
Sep 1 08:16:43 Mysql-248 Keepalived_vrrp[71143]: VRRP_Instance(HA_1) setting protocol VIPs.
Sep 1 08:16:43 Mysql-248 Keepalived_vrrp[71143]: VRRP_Instance(HA_1) Sending gratuitous ARPs on eth0 for 192.168.175.254
Sep 1 08:16:43 Mysql-248 Keepalived_healthcheckers[71142]: Netlink reflector reports IP 192.168.175.254 added
Sep 1 08:16:48 Mysql-248 Keepalived_vrrp[71143]: VRRP_Instance(HA_1) Sending gratuitous ARPs on eth0 for 192.168.175.254

在249上查看message日志，发现启动时节点进入BACKUP状态

[root@Mysql-249 ~]# tail -f /var/log/messages
Sep 1 08:19:15 Mysql-249 Keepalived_vrrp[54929]: Truncating auth_pass to 8 characters
Sep 1 08:19:15 Mysql-249 Keepalived_vrrp[54929]: Configuration is using : 65520 Bytes
Sep 1 08:19:15 Mysql-249 Keepalived_healthcheckers[54928]: Configuration is using : 7520 Bytes
Sep 1 08:19:15 Mysql-249 Keepalived_vrrp[54929]: Using LinkWatch kernel netlink reflector...
Sep 1 08:19:15 Mysql-249 Keepalived_vrrp[54929]: VRRP_Instance(HA_1) Entering BACKUP STATE
Sep 1 08:19:15 Mysql-249 Keepalived_healthcheckers[54928]: Using LinkWatch kernel netlink reflector...
Sep 1 08:19:15 Mysql-249 Keepalived_vrrp[54929]: VRRP sockpool: [ifindex(2), proto(112), unicast(0), fd(10,11)]

若两台服务器启动后都进入MASTER状态，都绑定了VIP，则需要检查防火墙是否放行了VRRP。
防火墙放行VRRP后，会有一台机器进入BACKUP状态。
1
2
3
4

分别在两台服务器上放行vrrp协议(Centos5/6配置方法)：

[root@Mysql-248 keepalived]#iptables -I INPUT -p vrrp -d 224.0.0.18/32 -j ACCEPT

保存放行规则。若是通过开机脚本控制防火墙规则，请注意修改对应脚本。

[root@Mysql-248 keepalived]#/etc/init.d/iptables

2.检查两台服务器的IP，并在客户机通过VIP连接数据库
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27

在248上查看，发现此时vip在此机器

[root@Mysql-248 keepalived]# ip addr
1: lo: <loopback,up,lower_up>mtu 65536 qdisc noqueue state UNKNOWN
link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
inet 127.0.0.1/8 scope host lo
inet6 ::1/128 scope host
valid_lft forever preferred_lft forever
2: eth0: <broadcast,multicast,up,lower_up>mtu 1500 qdisc pfifo_fast state UP qlen 1000
link/ether 00:0c:29:47:a4:ea brd ff:ff:ff:ff:ff:ff
inet 192.168.175.248/24 brd 192.168.175.255 scope global eth0
inet 192.168.175.254/24 scope global secondary eth0
inet6 fe80::20c:29ff:fe47:a4ea/64 scope link
valid_lft forever preferred_lft forever

在249上查看，并无VIP

[root@Mysql-249 keepalived]# ip addr
1: lo: <loopback,up,lower_up>mtu 65536 qdisc noqueue state UNKNOWN
link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
inet 127.0.0.1/8 scope host lo
inet6 ::1/128 scope host
valid_lft forever preferred_lft forever
2: eth0: <broadcast,multicast,up,lower_up>mtu 1500 qdisc pfifo_fast state UP qlen 1000
link/ether 00:0c:29:e9:01:e0 brd ff:ff:ff:ff:ff:ff
inet 192.168.175.249/24 brd 192.168.175.255 scope global eth0
inet6 fe80::20c:29ff:fee9:1e0/64 scope link
valid_lft forever preferred_lft forever
wKiom1fHBN2gWP_IAAA89YqIno0039.png-wh_50
客户端通过VIP连接数据库

3.在客户机通过VIP连接Mysql，同时测试不间断ping vip254，并停止Mysql248上的Mysql服务器。停止后新建一个数据库测试主节点故障后的写操作是否正常。
1
2
3
[root@Mysql-248 keepalived]# /etc/init.d/mysqld stop
Shutting down MySQL............ SUCCESS!
[root@Mysql-248 keepalived]#

wKiom1fHB4Kgdgi7AABHJzFU0fI837.png-wh_50
不间断ping测试故障切换时间

wKiom1fHB4PTgS-jAAAh8cVrLkQ200.png-wh_50
故障切换后测试写入数据

在模拟故障的过程中，VIP切换时网络会出现一下中断，几次测试发现VIP切换的时间在几百毫秒延迟到一个丢包之间，切换时间非常短。但是测试写入数据提示连接中断，重试后发现可以正常写入数据。所以此架构需要程序端有数据库断开时重连的机制。在此情况下，主节点故障对业务的影响时间就降低到了秒级，无需人为干预才能恢复业务。
一台节点故障后，仅剩下一台节点能够正常工作，只需要配合监控发现故障后，及时修复故障节点，就可以将恢复后的节点重新成为keeplived的备用节点，以保证架构的持续高可用。

4.故障切换过程
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16

在故障节点248上查看日志，发现VRRP_Script检测到Mysql状态faild，节点进入FAULT状态并移除vip

[root@Mysql-248 ~]# tail -f /var/log/messages
Sep 1 08:32:51 Mysql-248 Keepalived_vrrp[71143]: VRRP_Script(check_mysqld) failed
Sep 1 08:32:52 Mysql-248 Keepalived_vrrp[71143]: VRRP_Instance(HA_1) Entering FAULT STATE
Sep 1 08:32:52 Mysql-248 Keepalived_vrrp[71143]: VRRP_Instance(HA_1) removing protocol VIPs.
Sep 1 08:32:52 Mysql-248 Keepalived_vrrp[71143]: VRRP_Instance(HA_1) Now in FAULT state
Sep 1 08:32:52 Mysql-248 Keepalived_healthcheckers[71142]: Netlink reflector reports IP 192.168.175.254 removed

249服务器进入MASTER状态，并绑定VIP

[root@Mysql-249 ~]# tail -f /var/log/messages
Sep 1 08:32:54 Mysql-249 Keepalived_vrrp[54929]: VRRP_Instance(HA_1) Transition to MASTER STATE
Sep 1 08:32:56 Mysql-249 Keepalived_vrrp[54929]: VRRP_Instance(HA_1) Entering MASTER STATE
Sep 1 08:32:56 Mysql-249 Keepalived_vrrp[54929]: VRRP_Instance(HA_1) setting protocol VIPs.
Sep 1 08:32:56 Mysql-249 Keepalived_vrrp[54929]: VRRP_Instance(HA_1) Sending gratuitous ARPs on eth0 for 192.168.175.254
Sep 1 08:32:56 Mysql-249 Keepalived_healthcheckers[54928]: Netlink reflector reports IP 192.168.175.254 added
Sep 1 08:33:01 Mysql-249 Keepalived_vrrp[54929]: VRRP_Instance(HA_1) Sending gratuitous ARPs on eth0 for 192.168.175.254

5.故障节点恢复
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23

将测试时关闭的数据库打开，查看日志发现节点进入BACKUP状态

查看数据库发现故障时候创建的数据库已经同步过来了

此时又恢复Mysql双主复制的高可用状态了

[root@Mysql-248 keepalived]# /etc/init.d/mysqld start
Starting MySQL.. SUCCESS!
[root@Mysql-248 keepalived]# tail -5 /var/log/messages
Sep 1 08:32:52 Mysql-248 Keepalived_vrrp[71143]: VRRP_Instance(HA_1) removing protocol VIPs.
Sep 1 08:32:52 Mysql-248 Keepalived_vrrp[71143]: VRRP_Instance(HA_1) Now in FAULT state
Sep 1 08:32:52 Mysql-248 Keepalived_healthcheckers[71142]: Netlink reflector reports IP 192.168.175.254 removed
Sep 1 08:51:42 Mysql-248 Keepalived_vrrp[71143]: VRRP_Script(check_mysqld) succeeded
Sep 1 08:51:44 Mysql-248 Keepalived_vrrp[71143]: VRRP_Instance(HA_1) Entering BACKUP STATE
[root@Mysql-248 keepalived]#
[root@Mysql-248 mysql-5.6.30]# mysql -uroot -p'qwe123``' -e 'show databases;'
Warning: Using a password on the command line interface can be insecure.
+--------------------+
| Database |
+--------------------+
| information_schema |
| cubix |
| db1 |
| mysql |
| performance_schema |
+--------------------+