企业中MySQL高可用集群架构三部曲之MM+keepalived
Posted 小鲜肉1
tags:
篇首语:本文由小常识网(cha138.com)小编为大家整理,主要介绍了企业中MySQL高可用集群架构三部曲之MM+keepalived相关的知识,希望对你有一定的参考价值。
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。
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:
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
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
|
[email protected] 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) |
1
2
3
|
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:
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
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
|
[email protected] 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) |
1
2
3
4
|
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;
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
|
Loaded plugins: fastestmirror, refresh-packagekit, security Loading mirror speeds from cached hostfile * c6-media: file : ///media/CentOS/repodata/repomd .xml: [Errno 14] Could not open /read file : ///media/CentOS/repodata/repomd .xml Trying other mirror. file : ///media/cdrecorder/repodata/repomd .xml: [Errno 14] 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:
1
2
|
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:
1
2
|
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 板面的做法和配料
(这里只展现了脚本中判断的一部分)
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
|
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 and mysqlPortNum <= 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
1
|
vim /etc/keepalived/keepalived .conf |
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
|
vrrp_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:
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
|
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:
1
2
|
[[email protected] keepalived] # /etc/init.d/keepalived start Starting keepalived: [ OK ] |
观察日志中信息的变化:
cat /var/log/messages
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
|
Aug 3 10:52:17 node1 Keepalived_vrrp[6760]: Opening file ‘/etc/keepalived/keepalived.conf‘ . Aug 3 10:52:17 node1 Keepalived_vrrp[6760]: Configuration is using : 62711 Bytes Aug 3 10:52:17 node1 Keepalived_vrrp[6760]: Using LinkWatch kernel netlink reflector... Aug 3 10:52:17 node1 Keepalived_vrrp[6760]: VRRP_Instance(VI_100) Entering BACKUP STATE Aug 3 10:52:17 node1 Keepalived_vrrp[6760]: VRRP sockpool: [ifindex(2), proto(112), fd(11,12)] Aug 3 10:52:17 node1 Keepalived_healthcheckers[6759]: Opening file ‘/etc/keepalived/keepalived.conf‘ . Aug 3 10:52:17 node1 Keepalived_healthcheckers[6759]: Configuration is using : 4431 Bytes Aug 3 10:52:17 node1 Keepalived_healthcheckers[6759]: Using LinkWatch kernel netlink reflector... Aug 3 10:52:18 node1 Keepalived_vrrp[6760]: VRRP_Script(vs_mysql_100) succeeded Aug 3 10:52:33 node1 Keepalived_vrrp[6760]: VRRP_Instance(VI_100) Transition to MASTER STATE Aug 3 10:52:38 node1 Keepalived_vrrp[6760]: VRRP_Instance(VI_100) Entering MASTER STATE Aug 3 10:52:38 node1 Keepalived_vrrp[6760]: VRRP_Instance(VI_100) setting protocol VIPs. Aug 3 10:52:38 node1 Keepalived_vrrp[6760]: VRRP_Instance(VI_100) Sending gratuitous ARPs on eth0 for 192.168.56.111 Aug 3 10:52:38 node1 Keepalived_healthcheckers[6759]: Netlink reflector reports IP 192.168.56.111 added Aug 3 10:52:43 node1 Keepalived_vrrp[6760]: 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。
1
2
3
4
5
6
7
8
9
10
11
12
|
[[email protected] keepalived] # 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:
1
2
|
[[email protected] ~] # /etc/init.d/keepalived start Starting keepalived: [ OK ] |
cat /var/log/messages
1
2
3
4
5
6
7
8
9
10
11
12
13
14
|
Aug 3 10:53:28 node2 Keepalived_healthcheckers[8111]: Interface queue is empty Aug 3 10:53:28 node2 Keepalived_healthcheckers[8111]: Netlink reflector reports IP 192.168.56.101 added Aug 3 10:53:28 node2 Keepalived_healthcheckers[8111]: Netlink reflector reports IP fe80::a00:27ff:fe83:b39d added Aug 3 10:53:28 node2 Keepalived_healthcheckers[8111]: Registering Kernel netlink reflector Aug 3 10:53:28 node2 Keepalived_healthcheckers[8111]: Registering Kernel netlink command channel Aug 3 10:53:28 node2 Keepalived_healthcheckers[8111]: Opening file ‘/etc/keepalived/keepalived.conf‘ . Aug 3 10:53:28 node2 Keepalived_healthcheckers[8111]: Configuration is using : 4561 Bytes Aug 3 10:53:28 node2 Keepalived_healthcheckers[8111]: Using LinkWatch kernel netlink reflector... Aug 3 10:53:28 node2 Keepalived_vrrp[8112]: Opening file ‘/etc/keepalived/keepalived.conf‘ . Aug 3 10:53:28 node2 Keepalived_vrrp[8112]: Configuration is using : 62577 Bytes Aug 3 10:53:28 node2 Keepalived_vrrp[8112]: Using LinkWatch kernel netlink reflector... Aug 3 10:53:28 node2 Keepalived_vrrp[8112]: VRRP_Instance(VI_101) Entering BACKUP STATE Aug 3 10:53:28 node2 Keepalived_vrrp[8112]: VRRP sockpool: [ifindex(2), proto(112), fd(12,13)] Aug 3 10:53:28 node2 Keepalived_vrrp[8112]: 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:
1
|
mysqladmin -uroot -proot123 shutdown |
这时再查看一下keepalived日志的情况:
1
2
3
4
5
6
|
cat /var/log/messages Aug 3 11:04:17 node1 Keepalived_vrrp[6760]: VRRP_Script(vs_mysql_100) failed Aug 3 11:04:18 node1 Keepalived_vrrp[6760]: VRRP_Instance(VI_100) Entering FAULT STATE Aug 3 11:04:18 node1 Keepalived_vrrp[6760]: VRRP_Instance(VI_100) removing protocol VIPs. Aug 3 11:04:18 node1 Keepalived_healthcheckers[6759]: Netlink reflector reports IP 192.168.56.111 removed Aug 3 11:04:18 node1 Keepalived_vrrp[6760]: 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:
1
2
3
4
5
6
7
|
cat /var/log/messages Aug 3 11:04:29 node2 Keepalived_vrrp[8112]: VRRP_Instance(VI_101) Transition to MASTER STATE Aug 3 11:04:34 node2 Keepalived_vrrp[8112]: VRRP_Instance(VI_101) Entering MASTER STATE Aug 3 11:04:34 node2 Keepalived_vrrp[8112]: VRRP_Instance(VI_101) setting protocol VIPs. Aug 3 11:04:34 node2 Keepalived_vrrp[8112]: VRRP_Instance(VI_101) Sending gratuitous ARPs on eth0 for 192.168.56.111 Aug 3 11:04:34 node2 Keepalived_healthcheckers[8111]: Netlink reflector reports IP 192.168.56.111 added Aug 3 11:04:39 node2 Keepalived_vrrp[8112]: 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:
1
2
3
4
5
6
7
8
9
10
11
12
|
[[email protected] ~] # 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。
以上是关于企业中MySQL高可用集群架构三部曲之MM+keepalived的主要内容,如果未能解决你的问题,请参考以下文章