企业中MySQL高可用集群架构三部曲之MM+keepalived
各位老铁们，老张与大家又见面了。看到各位在博客里面给我的留言和访问量的情况，我很是欣慰，也谢谢大家对我的认可。我写这些博客，就是想把自己对于MySQL数据库的一些看法和自己平时的实战经验分享出来，我们可以一起探讨，共同进步。也保证今后只要一有空就更新博文，推出更多的干货。我的学生经常对我说：“张老师，每次我遇到报错，有时还是会百度，但是最烦的是不知道百度哪篇帖子说的是正确的".其实这些呢，都是因为自己还没有对MySQL数据库核心知识的不熟悉，和对技术掌握的不牢固。平时下得功夫还是不到位。我们做技术这个行业，还是需要自己给自己加发条，促使自己每天都要学习一些新的知识。理论配合实验一起，先要学会多问自己几个问题，一个实验多做几遍，可能会得到不同的实验效果。学习知识要踏实下来，学会多做实验总结。我想今后再遇到报错，可能自己就会有一个清晰的解题思路，这个需要一定时间的磨练。
也有人经常问Oracle和MySQL到底有啥区别，其实MySQL数据库上手很简单，难的是后期架构的设计与维护。老张三部曲中第一部曲MHA希望对大家在线上部署方面有帮助。
今儿给大家介绍第二部曲，MM+keepalived的环境部署，我们会多种数据库的架构就可以灵活应用到我们的公司。根据公司业务的不同，选择合适的集群架构。
独家新课程上线>>MySQL体系结构深入剖析及实战DBA视频课程
MM+keepalived
简介：
双主配合keepalived这种架构设计，也是基于主从复制的原理而搭建的。
使用MySQL主主复制技术+Keepalived是一种简单、便捷的解决方案，在高可用集群环境中，keepalived使用VIP，使用Keepalived自带的服务监控功能和自定义脚本来实现MySQL故障时自动切换，非常灵活。
应用范围：
一般中小型公司都使用这种架构，搭建比较方便简单；
可以采用主从或者主主模式，在 master 节点发生故障后，利用 keepalived 高可用机制实现快速切换到 slave 节点。原来的从库变成新的主库。
个人建议：
[*] 一定要完善好切换脚本，keepalived 的切换机制要合理，避免切换不成功的现象发生。
[*] 从库的配置尽快要与主库一致，不能太次；避免主库宕机发生切换，新的主库（原来的从库）影响线上业务进行。
[*] 对于延迟的问题，在这套架构中，也不能避免。可以使用 mysql 5.7 中增强半同步完成。也可以改变架构使用 PXC，完成时时同步功能，基本上没有延迟；
[*] keepalived 无法解决脑裂的问题，因此在进行服务异常判断时，可以修改我们的判断脚本，通过对第三方节点补充检测来决定是否进行切换，可降低脑裂问题产生的风险。
[*] 采用 keepalived 这个架构，在设置两节点状态时，都要设置成不抢占模式，都是 backup 状态，通过优先级，来决定谁是主库。避免脑裂，冲突现象发生。
[*] 安装好 mysql 需要的一些依赖包；建议配置好 yum 源，用 yum 安装 keepalived 即可。
搭建架构之前理清思路：
[*] 首先需要装有两台mysql的数据库服务器，两者之间互为主从模式，都可读写。其实就只有一台服务器A负责数据的写入工作，而另一台服务器B作为我们的备用数据库；
[*] 安装keepalived的软件包，个人建议使用yum安装就可以，很方便。当然我们要知道yum安装之后的软件路径在什么位置。
[*] 整理好keepalived的配置文件，理清keepalived的三种状态信息。还要准备一个监控MySQL的脚本，便于检测到宕机顺利发生切换的过程。
[*] 所有提供服务的从服务器与备用服务器B进行主从同步。----双主从库模式
[*] 在两台服务器A和B，当配置keepalived的参数文件中，要注意两台机器都要采用backup这种状态，就是nopreempt这种非抢占模式，避免出现冲突，发生脑裂现象。
架构图展示：
实验部署环境介绍：
192.168.56.100 主 ---master1
192.168.56.101 备库---master2
都是干净环境没有任何数据
VIP：192.168.56.111
MySQL数据库版本5.7.14，采用GTID模式搭建主从环境
注意事项：
[*] 两台机器的防火墙必须是关闭状态。
[*] 两台MySQL数据库配置文件中server-id绝对不能一样，要不会报1593这个主从同步的错误，导致搭建不成功。
实战开始：
首先先要搭建两台MySQL数据库为互为主从的架构模式。
添加主从同步账号：
在192.168.56.100：
grant replication slave on *.* to 'bak'@'192.168.56.%' identified by '123456';
flush privileges;
在192.168.56.101：
grant replication slave on *.* to 'bak'@'192.168.56.%' identified by '123456';
flush privileges;
配置同步信息：
先在192.168.56.101上：
CHANGE MASTER TO MASTER_HOST='192.168.56.100',MASTER_USER='bak',MASTER_PASSWORD='123456',master_auto_position=1;
打开主从同步开关：
start slave；
查看主从同步状态：
root@db 10:49:[(none)]> show slave status\G;
*************************** 1. row ***************************
Slave_IO_State: Waiting for master to send event
Master_Host: 192.168.56.100
Master_User: bak
Master_Port: 3306
Connect_Retry: 60
Master_Log_File: mybinlog.000005
Read_Master_Log_Pos: 234
Relay_Log_File: node2-relay-bin.000010
Relay_Log_Pos: 365
Relay_Master_Log_File: mybinlog.000005
Slave_IO_Running: Yes
Slave_SQL_Running: Yes
Replicate_Do_DB:
Replicate_Ignore_DB:
Replicate_Do_Table:
Replicate_Ignore_Table:
Replicate_Wild_Do_Table:
Replicate_Wild_Ignore_Table:
Last_Errno: 0
Last_Error:
Skip_Counter: 0
Exec_Master_Log_Pos: 234
Relay_Log_Space: 572
Until_Condition: None
Until_Log_File:
Until_Log_Pos: 0
Master_SSL_Allowed: No
Master_SSL_CA_File:
Master_SSL_CA_Path:
Master_SSL_Cert:
Master_SSL_Cipher:
Master_SSL_Key:
Seconds_Behind_Master: 0
Master_SSL_Verify_Server_Cert: No
Last_IO_Errno: 0
Last_IO_Error:
Last_SQL_Errno: 0
Last_SQL_Error:
Replicate_Ignore_Server_Ids:
Master_Server_Id: 3306100
Master_UUID: 913cc1d6-771a-11e7-8fe1-080027f1fd08
Master_Info_File: mysql.slave_master_info
SQL_Delay: 0
SQL_Remaining_Delay: NULL
Slave_SQL_Running_State: Slave has read all relay log; waiting for more updates
Master_Retry_Count: 86400
Master_Bind:
Last_IO_Error_Timestamp:
Last_SQL_Error_Timestamp:
Master_SSL_Crl:
Master_SSL_Crlpath:
Retrieved_Gtid_Set:
Executed_Gtid_Set: 5a668126-771d-11e7-b14c-08002783b39d:1-2,
913cc1d6-771a-11e7-8fe1-080027f1fd08:1-13
Auto_Position: 1
Replicate_Rewrite_DB:
Channel_Name:
Master_TLS_Version:
1 row in set (0.00 sec)从库两个工作的线程：IO，SQL都为yes，代表同步搭建成功
Slave_IO_Running: Yes
Slave_SQL_Running: Yes
再在192.168.56.100上：
配置主从同步信息：
CHANGE MASTER TO MASTER_HOST='192.168.56.101',MASTER_USER='bak',MASTER_PASSWORD='123456',master_auto_position=1;
开启主从开关：
start slave；
查看主从同步状态：
root@db 10:49:[(none)]> show slave status\G;
*************************** 1. row ***************************
Slave_IO_State: Waiting for master to send event
Master_Host: 192.168.56.101
Master_User: bak
Master_Port: 3306
Connect_Retry: 60
Master_Log_File: mybinlog.000005
Read_Master_Log_Pos: 234
Relay_Log_File: node1-relay-bin.000008
Relay_Log_Pos: 365
Relay_Master_Log_File: mybinlog.000005
Slave_IO_Running: Yes
Slave_SQL_Running: Yes
Replicate_Do_DB:
Replicate_Ignore_DB:
Replicate_Do_Table:
Replicate_Ignore_Table:
Replicate_Wild_Do_Table:
Replicate_Wild_Ignore_Table:
Last_Errno: 0
Last_Error:
Skip_Counter: 0
Exec_Master_Log_Pos: 234
Relay_Log_Space: 572
Until_Condition: None
Until_Log_File:
Until_Log_Pos: 0
Master_SSL_Allowed: No
Master_SSL_CA_File:
Master_SSL_CA_Path:
Master_SSL_Cert:
Master_SSL_Cipher:
Master_SSL_Key:
Seconds_Behind_Master: 0
Master_SSL_Verify_Server_Cert: No
Last_IO_Errno: 0
Last_IO_Error:
Last_SQL_Errno: 0
Last_SQL_Error:
Replicate_Ignore_Server_Ids:
Master_Server_Id: 3306101
Master_UUID: 5a668126-771d-11e7-b14c-08002783b39d
Master_Info_File: mysql.slave_master_info
SQL_Delay: 0
SQL_Remaining_Delay: NULL
Slave_SQL_Running_State: Slave has read all relay log; waiting for more updates
Master_Retry_Count: 86400
Master_Bind:
Last_IO_Error_Timestamp:
Last_SQL_Error_Timestamp:
Master_SSL_Crl:
Master_SSL_Crlpath:
Retrieved_Gtid_Set:
Executed_Gtid_Set: 5a668126-771d-11e7-b14c-08002783b39d:1-2,
913cc1d6-771a-11e7-8fe1-080027f1fd08:1-13
Auto_Position: 1
Replicate_Rewrite_DB:
Channel_Name:
Master_TLS_Version:
1 row in set (0.00 sec)同理
Slave_IO_Running: Yes
Slave_SQL_Running: Yes
从库两个工作的线程：IO，SQL都为yes，代表同步搭建成功。
分别两台机器上安装keepalived的软件包，可以使用yum安装方式
yum -y install keepalived;
Loaded plugins: fastestmirror, refresh-packagekit, security
Loading mirror speeds from cached hostfile
* c6-media:
file:///media/CentOS/repodata/repomd.xml: Could not open/read file:///media/CentOS/repodata/repomd.xml
Trying other mirror.
file:///media/cdrecorder/repodata/repomd.xml: Could not open/read file:///media/cdrecorder/repodata/repomd.xml
Trying other mirror.
c6-media | 4.0 kB 00:00 ...
Setting up Install Process
Resolving Dependencies
--> Running transaction check
---> Package keepalived.x86_64 0:1.2.7-3.el6 will be installed
--> Finished Dependency Resolution
Dependencies Resolved
======================================================================================================================================
Package Arch Version Repository Size
======================================================================================================================================
Installing:
keepalived x86_64 1.2.7-3.el6 c6-media 174 k
Transaction Summary
======================================================================================================================================
Install 1 Package(s)
Total download size: 174 k
Installed size: 526 k
Downloading Packages:
Running rpm_check_debug
Running Transaction Test
Transaction Test Succeeded
Running Transaction
Installing : keepalived-1.2.7-3.el6.x86_64 1/1
Verifying: keepalived-1.2.7-3.el6.x86_64 1/1
Installed:
keepalived.x86_64 0:1.2.7-3.el6
Complete! 证明keepalived软件已经存在：
rpm -qa|grep keepalived
keepalived-1.2.7-3.el6.x86_64
在两台机器上分别配置检测MySQL数据库的脚本：
首先进入到yum安装后的软件目录下：
cd /etc/keepalived
vim check_mysql.py 脚本通过查看MySQL进程的存在，和是否可以连接，来判断MySQL的状态
（这里只展现了脚本中判断的一部分）
def existMySQL(self):
cmd="ps -ef | egrep -i \"mysqld\" | grep %s | egrep -iv \"mysqld_safe\" | grep -v grep | wc -l" % self.dbport
mysqldNum = os.popen(cmd).read()
cmd ="netstat -tunlp | grep \":%s\" | wc -l" % self.dbport
mysqlPortNum= os.popen(cmd).read()
#print mysqldNum, mysqlPortNum
if ( int(mysqldNum) <= 0):
print "error"
return 1
if ( int(mysqldNum) > 0 andmysqlPortNum <= 0):
return 1
return 0
def connect(self):
# print "in db conn"
# print "host : %s, port: %d, user: %s, password: %s" % (self.dbhost, self.dbport, self.dbuser, self.dbpassword)
try:
self.conn=MySQLdb.connect(host="%s"%self.dbhost, port=self.dbport,user="%s"%dbuser, passwd="%s"%self.dbpassword)
except Exception, e:
# print " Error"
print e
return 1
return 0
def disconnect(self):
if (self.conn):
self.conn.close()
self.conn = None
在两台机器上修改keepalived的配置文件：
192.168.56.100上面
vim /etc/keepalived/keepalived.confvrrp_instance VI_100 {
state BACKUP
nopreempt
interface eth0
virtual_router_id 100
priority 100
advert_int 5
authentication {
auth_type PASS
auth_pass 1111
}
track_script {
vs_mysql_100
}
virtual_ipaddress {
192.168.56.111
}
} 192.168.56.101上面：
vrrp_instance VI_101 {
state BACKUP
nopreempt
interface eth0
virtual_router_id 100
priority 90
advert_int 5
authentication {
auth_type PASS
auth_pass 1111
}
track_script {
vs_mysql_101
}
virtual_ipaddress {
192.168.56.111
}
} 总结：可以看到两台机器的state都是backup并且都是非抢占模式nopreempt，通过优先级的高低来决定谁是主库。（这里192.168.56.100是主）还有注意virtual_router_id(虚拟路由id）要保持一致。
接下来可以启动两台机器的keepalived进程：
192.168.56.100：
# /etc/init.d/keepalived start
Starting keepalived: 观察日志中信息的变化：
cat /var/log/messages
Aug3 10:52:17 node1 Keepalived_vrrp: Opening file '/etc/keepalived/keepalived.conf'.
Aug3 10:52:17 node1 Keepalived_vrrp: Configuration is using : 62711 Bytes
Aug3 10:52:17 node1 Keepalived_vrrp: Using LinkWatch kernel netlink reflector...
Aug3 10:52:17 node1 Keepalived_vrrp: VRRP_Instance(VI_100) Entering BACKUP STATE
Aug3 10:52:17 node1 Keepalived_vrrp: VRRP sockpool:
Aug3 10:52:17 node1 Keepalived_healthcheckers: Opening file '/etc/keepalived/keepalived.conf'.
Aug3 10:52:17 node1 Keepalived_healthcheckers: Configuration is using : 4431 Bytes
Aug3 10:52:17 node1 Keepalived_healthcheckers: Using LinkWatch kernel netlink reflector...
Aug3 10:52:18 node1 Keepalived_vrrp: VRRP_Script(vs_mysql_100) succeeded
Aug3 10:52:33 node1 Keepalived_vrrp: VRRP_Instance(VI_100) Transition to MASTER STATE
Aug3 10:52:38 node1 Keepalived_vrrp: VRRP_Instance(VI_100) Entering MASTER STATE
Aug3 10:52:38 node1 Keepalived_vrrp: VRRP_Instance(VI_100) setting protocol VIPs.
Aug3 10:52:38 node1 Keepalived_vrrp: VRRP_Instance(VI_100) Sending gratuitous ARPs on eth0 for 192.168.56.111
Aug3 10:52:38 node1 Keepalived_healthcheckers: Netlink reflector reports IP 192.168.56.111 added
Aug3 10:52:43 node1 Keepalived_vrrp: VRRP_Instance(VI_100) Sending gratuitous ARPs on eth0 for 192.168.56.111 可以看到它从backup状态切换到master的状态了，并且发送了一个广播协议，证明192.168.56.111已经在本台机器上面，其他机器不要再使用了。
# ip addr show
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 16436 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 08:00:27:f1:fd:08 brd ff:ff:ff:ff:ff:ff
inet 192.168.56.100/24 brd 192.168.56.255 scope global eth0
inet 192.168.56.111/32 scope global eth0
inet6 fe80::a00:27ff:fef1:fd08/64 scope link
valid_lft forever preferred_lft forever 192.168.56.101:
# /etc/init.d/keepalived start
Starting keepalived: cat /var/log/messages
Aug3 10:53:28 node2 Keepalived_healthcheckers: Interface queue is empty
Aug3 10:53:28 node2 Keepalived_healthcheckers: Netlink reflector reports IP 192.168.56.101 added
Aug3 10:53:28 node2 Keepalived_healthcheckers: Netlink reflector reports IP fe80::a00:27ff:fe83:b39d added
Aug3 10:53:28 node2 Keepalived_healthcheckers: Registering Kernel netlink reflector
Aug3 10:53:28 node2 Keepalived_healthcheckers: Registering Kernel netlink command channel
Aug3 10:53:28 node2 Keepalived_healthcheckers: Opening file '/etc/keepalived/keepalived.conf'.
Aug3 10:53:28 node2 Keepalived_healthcheckers: Configuration is using : 4561 Bytes
Aug3 10:53:28 node2 Keepalived_healthcheckers: Using LinkWatch kernel netlink reflector...
Aug3 10:53:28 node2 Keepalived_vrrp: Opening file '/etc/keepalived/keepalived.conf'.
Aug3 10:53:28 node2 Keepalived_vrrp: Configuration is using : 62577 Bytes
Aug3 10:53:28 node2 Keepalived_vrrp: Using LinkWatch kernel netlink reflector...
Aug3 10:53:28 node2 Keepalived_vrrp: VRRP_Instance(VI_101) Entering BACKUP STATE
Aug3 10:53:28 node2 Keepalived_vrrp: VRRP sockpool:
Aug3 10:53:28 node2 Keepalived_vrrp: VRRP_Script(vs_mysql_101) succeeded 这台服务器就是正常的backup状态，时刻准备着接管主库的服务。
模拟一下主库宕机的故障切换：
主库192.168.56.100上面执行关闭MySQL服务操作：
mysqladmin -uroot -proot123 shutdown 这时再查看一下keepalived日志的情况：
cat /var/log/messages
Aug3 11:04:17 node1 Keepalived_vrrp: VRRP_Script(vs_mysql_100) failed
Aug3 11:04:18 node1 Keepalived_vrrp: VRRP_Instance(VI_100) Entering FAULT STATE
Aug3 11:04:18 node1 Keepalived_vrrp: VRRP_Instance(VI_100) removing protocol VIPs.
Aug3 11:04:18 node1 Keepalived_healthcheckers: Netlink reflector reports IP 192.168.56.111 removed
Aug3 11:04:18 node1 Keepalived_vrrp: VRRP_Instance(VI_100) Now in FAULT state 已经把vip removed了，并且keepalived的状态变成了fault
在备库192.168.56.101上面查看日志：
cat /var/log/messages
Aug3 11:04:29 node2 Keepalived_vrrp: VRRP_Instance(VI_101) Transition to MASTER STATE
Aug3 11:04:34 node2 Keepalived_vrrp: VRRP_Instance(VI_101) Entering MASTER STATE
Aug3 11:04:34 node2 Keepalived_vrrp: VRRP_Instance(VI_101) setting protocol VIPs.
Aug3 11:04:34 node2 Keepalived_vrrp: VRRP_Instance(VI_101) Sending gratuitous ARPs on eth0 for 192.168.56.111
Aug3 11:04:34 node2 Keepalived_healthcheckers: Netlink reflector reports IP 192.168.56.111 added
Aug3 11:04:39 node2 Keepalived_vrrp: VRRP_Instance(VI_101) Sending gratuitous ARPs on eth0 for 192.168.56.111 备库已经从backup状态切换成master状态了。并且VIP（192.168.56.111）已经切换过来。
查看ip地址：
# ip addr show
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 16436 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 08:00:27:83:b3:9d brd ff:ff:ff:ff:ff:ff
inet 192.168.56.101/24 brd 192.168.56.255 scope global eth0
inet 192.168.56.111/32 scope global eth0
inet6 fe80::a00:27ff:fe83:b39d/64 scope link
valid_lft forever preferred_lft forever 主备库切换成功。
实战演练过程结束，希望对大家学习MySQL高可用集群有帮助。
今后我们可能还会遇到其他的MySQL高可用架构，学习它的时候，先不要忙于搭建，要先弄清原理，整理好实验过程的思路，遇到报错，一步步地去排查。自己的水平也会在这个历练的过程中，得到提升。
今后我们可以一起讨论，在写博的过程中难免可能会有一些笔误，或是想不周全的地方，希望大家谅解。有不对的地方欢迎各位老铁指定。（MySQL高可用集群第二部曲完结）
第一步曲直通车>> 企业中MySQL主流高可用架构实战三部曲之MHA
页:
[1]