MMM配置官方文檔: http://mysql-mmm.org/mmm2:guide
本文環境:
主庫：CentOS6.7 x64 192.168.0.201 mysql-5.6.29
主庫：CentOS6.7 x64 192.168.0.202 mysql-5.6.29
備庫：CentOS6.7 x64 192.168.0.203 mysql-5.6.29
備庫：CentOS6.7 x64 192.168.0.204 mysql-5.6.29
管理：CentOS6.7 x64 192.168.0.205
MMM服務器角色對應關係:
node1(主1) 192.168.0.201 db1 1
node2(主2) 192.168.0.202 db2 2
node3(從1) 192.168.0.203 db3 11
node4(從1) 192.168.0.204 db4 12
node5(監控) 192.168.0.205 mon -
配置完成後，使用下面的VIP訪問MySQL Cluster。
192.168.0.211 writer
192.168.0.212 reader
192.168.0.213 reader
192.168.0.214 reader
192.168.0.215 reader
一、雙主環境，以及從服務器環境準備
1，node1(主1) /etc/my.cnf
[mysqld]
server-id = 1
binlog-format = ROW
log-bin = master-bin
log-bin-index = master-bin.index
log-slave-updates = true
auto_increment_offset = 1
auto_increment_increment = 2
mysql> show master status;
+-------------------+----------+--------------+------------------+-------------------+
| File | Position | Binlog_Do_DB | Binlog_Ignore_DB | Executed_Gtid_Set |
+-------------------+----------+--------------+------------------+-------------------+
| master-bin.000001 | 120 | | | |
+-------------------+----------+--------------+------------------+-------------------+
1 row in set (0.00 sec)
mysql>
2. node1(主2) /etc/my.cnf
[mysqld]
server-id = 2
binlog-format = ROW
log-bin = master-bin
log-bin-index = master-bin.index
log-slave-updates = true
auto_increment_offset = 2
auto_increment_increment = 2
mysql> show master status;
+-------------------+----------+--------------+------------------+-------------------+
| File | Position | Binlog_Do_DB | Binlog_Ignore_DB | Executed_Gtid_Set |
+-------------------+----------+--------------+------------------+-------------------+
| master-bin.000001 | 120 | | | |
+-------------------+----------+--------------+------------------+-------------------+
1 row in set (0.00 sec)
mysql>
3. node1,node2 雙主服務器配置同步複製帳號
grant replication slave on *.* to 'repl'@'%' identified by '123456';
flush privileges;
4, node3（從）/etc/my.cnf
[mysqld]
binlog-format = ROW
log-bin = mysql-bin
relay-log = slave-relay-bin
relay-log-index = slave-relay-bin.index
log-slave-updates = true
server-id = 11
skip-name-resolve
5， node4（從）/etc/my.cnf
[mysqld]
binlog-format = ROW
log-bin = mysql-bin
relay-log = slave-relay-bin
relay-log-index = slave-relay-bin.index
log-slave-updates = true
server-id = 12
skip-name-resolve
6. node1從庫配置同步過程
(1) 配置同步，手動執行同步參數，該配置會寫入master.info文件中。
mysql >
CHANGE MASTER TO
MASTER_HOST='192.168.0.202',
MASTER_PORT=3306,
MASTER_USER='repl',
MASTER_PASSWORD='123456',
MASTER_LOG_FILE='master-bin.000001',
MASTER_LOG_POS=120;
mysql> start slave;
mysql> show slave status\G;
*************************** 1. row ***************************
Slave_IO_State: Waiting for master to send event
Master_Host: 192.168.0.202
Master_User: repl
Master_Port: 3306
Connect_Retry: 60
Master_Log_File: master-bin.000001
Read_Master_Log_Pos: 407
Relay_Log_File: node1-relay-bin.000002
Relay_Log_Pos: 571
Relay_Master_Log_File: master-bin.000001
Slave_IO_Running: Yes
Slave_SQL_Running: Yes
Replicate_Do_DB:
7. node1從庫配置同步過程
(1) 配置同步，手動執行同步參數，該配置會寫入master.info文件中。
mysql >
CHANGE MASTER TO
MASTER_HOST='192.168.0.201',
MASTER_PORT=3306,
MASTER_USER='repl',
MASTER_PASSWORD='123456',
MASTER_LOG_FILE='master-bin.000001',
MASTER_LOG_POS=120;
mysql> start slave;
mysql> show slave status\G;
*************************** 1. row ***************************
Slave_IO_State: Waiting for master to send event
Master_Host: 192.168.0.201
Master_User: repl
Master_Port: 3306
Connect_Retry: 60
Master_Log_File: master-bin.000001
Read_Master_Log_Pos: 694
Relay_Log_File: node2-relay-bin.000002
Relay_Log_Pos: 571
Relay_Master_Log_File: master-bin.000001
Slave_IO_Running: Yes
Slave_SQL_Running: Yes
Replicate_Do_DB:
8. 測試雙主相互複製
(1) node1上創建數據庫
mysql> create database mydb;
Query OK, 1 row affected (0.00 sec)
mysql> use mydb;
Database changed
mysql> show tables;
+----------------+
| Tables_in_mydb |
+----------------+
| user |
+----------------+
1 row in set (0.01 sec)
mysql> select * from user;
+----+----------+----------+
| id | username | password |
+----+----------+----------+
| 1 | koumm | 123456 |
+----+----------+----------+
1 row in set (0.01 sec)
mysql>
(2) node2上的操作如下：
mysql> show databases;
+--------------------+
| Database |
+--------------------+
| information_schema |
| mydb |
| mysql |
| performance_schema |
+--------------------+
4 rows in set (0.00 sec)
mysql> use mydb;
Database changed
mysql> CREATE TABLE `user` (
-> `id` varchar(20) NOT NULL,
-> `username` varchar(20) NOT NULL,
-> `password` char(32) NOT NULL,
-> PRIMARY KEY (`id`)
-> ) ;
Query OK, 0 rows affected (0.14 sec)
mysql> INSERT INTO user VALUES ('1', 'koumm', '123456');
Query OK, 1 row affected (0.00 sec)
mysql>
二、配置node3, node4同步node1(主)
1，node1主庫鎖表，導數據庫
(1) 主庫鎖表
mysql> flush tables with read lock;
Query OK, 0 rows affected (0.00 sec)
mysql> show master status;
+-------------------+----------+--------------+------------------+-------------------+
| File | Position | Binlog_Do_DB | Binlog_Ignore_DB | Executed_Gtid_Set |
+-------------------+----------+--------------+------------------+-------------------+
| master-bin.000001 | 1226 | | | |
+-------------------+----------+--------------+------------------+-------------------+
1 row in set (0.00 sec)
(2) 主庫備份
[root@master ~]# mysqldump -uroot -p -B mydb > mydb.sql
說明：-B參數有建庫語句。
(3) 主庫解開鎖表功能
mysql> unlock tables;
Query OK, 0 rows affected (0.00 sec)
mysql>
2. node3,node4從庫導入數據庫
# mysql -uroot -padmin < mydb.sql
3. node3,node4 從庫配置同步過程
(1) 配置同步，手動執行同步參數，該配置會寫入master.info文件中。
mysql >
CHANGE MASTER TO
MASTER_HOST='192.168.0.201',
MASTER_PORT=3306,
MASTER_USER='repl',
MASTER_PASSWORD='123456',
MASTER_LOG_FILE='master-bin.000001',
MASTER_LOG_POS=1226;
mysql> start slave;
Query OK, 0 rows affected (0.04 sec)
mysql> show slave status\G;
*************************** 1. row ***************************
Slave_IO_State: Waiting for master to send event
Master_Host: 192.168.1.121
Master_User: repl
Master_Port: 3306
Connect_Retry: 60
Master_Log_File: master-bin.000005
Read_Master_Log_Pos: 730
Relay_Log_File: slave-relay-bin.000002
Relay_Log_Pos: 284
Relay_Master_Log_File: master-bin.000005
Slave_IO_Running: Yes
Slave_SQL_Running: Yes
Replicate_Do_DB:
Replicate_Ignore_DB:
可以簡單測試一下同步數據情況。
三、配置MMM過程
1, 安裝mmm軟件包
(1) node1,node2,node3,node4,node5節點
# wget http://download.fedoraproject.org/pub/epel/6/x86_64/epel-release-6-8.noarch.rpm
# rpm -ivh epel-release-6-8.noarch.rpm
(2) node5(監控節點)
# yum install mysql-mmm mysql-mmm-agent mysql-mmm-tools mysql-mmm-monitor
(3) node1,node2,node3,node4 (mysql節點)
# yum install mysql-mmm-agent
2，所有的MySQL節點添加以下兩個用戶，這裏可以在主庫做該操作，會自動同步到其它節點。
mysql> grant replication client on *.* to 'mmm_monitor'@'192.168.0.%' identified by '123456';
mysql> grant super, replication client, process on *.* to 'mmm_agent'@'192.168.0.%' identified by '123456';
mysql> flush privileges;
3，所有主機node1,node2,node3,node4,node5上創建如下文件
[root@node1 ~]# vi /etc/mysql-mmm/mmm_common.conf
active_master_role writer
<host default>
cluster_interface eth0
pid_path /var/run/mysql-mmm/mmm_agentd.pid
bin_path /usr/libexec/mysql-mmm/
replication_user repl
replication_password 123456
agent_user mmm_agent
agent_password 123456
</host>
<host db1>
ip 192.168.0.201
mode master
peer db2
</host>
<host db2>
ip 192.168.0.202
mode master
peer db1
</host>
<host db3>
ip 192.168.0.203
mode slave
</host>
<host db4>
ip 192.168.0.204
mode slave
</host>
<role writer>
hosts db1, db2
ips 192.168.0.211
mode exclusive
</role>
<role reader>
hosts db1, db2, db3, db4
ips 192.168.0.212,192.168.0.213,192.168.0.214,192.168.0.215
mode balanced
</role>
複製該配置文件到所有節點上:
[root@node1 ~]# scp /etc/mysql-mmm/mmm_common.conf node2:/etc/mysql-mmm/
[root@node1 ~]# scp /etc/mysql-mmm/mmm_common.conf node3:/etc/mysql-mmm/
[root@node1 ~]# scp /etc/mysql-mmm/mmm_common.conf node4:/etc/mysql-mmm/
[root@node1 ~]# scp /etc/mysql-mmm/mmm_common.conf node5:/etc/mysql-mmm/
4. 配置agent代理節點，所有mysql主從服務器均爲代理節點
(1) node1,node2,node3,node4 的MySQL節點配置mmm_agent.conf
# vi /etc/mysql-mmm/mmm_agent.conf
include mmm_common.conf
this db1 #在四臺mysql節點上設置對應的db，分別爲db1、db2、db3、db4
(2) 默認爲啓用，可以不用修改
# cat /etc/default/mysql-mmm-agent
# mysql-mmm-agent defaults
ENABLED=1
(3) mysql節點啓動服務
# chkconfig mysql-mmm-agent on
# /etc/init.d/mysql-mmm-agent start
5，配置MMM管理監控節點node5
(1) 配置監控
[root@node5 ~]# vi /etc/mysql-mmm/mmm_mon.conf
include mmm_common.conf
<monitor>
ip 127.0.0.1
pid_path /var/run/mysql-mmm/mmm_mond.pid
bin_path /usr/libexec/mysql-mmm
status_path /var/lib/mysql-mmm/mmm_mond.status
ping_ips 192.168.0.201,192.168.0.202,192.168.0.203,192.168.0.204
auto_set_online 30
# The kill_host_bin does not exist by default, though the monitor will
# throw a warning about it missing. See the section 5.10 "Kill Host
# Functionality" in the PDF documentation.
#
# kill_host_bin /usr/libexec/mysql-mmm/monitor/kill_host
#
</monitor>
<host default>
monitor_user mmm_monitor
monitor_password 123456
</host>
debug 0
(2) mysql節點啓動服務
# chkconfig mysql-mmm-monitor on
# /etc/init.d/mysql-mmm-monitor start
6， 查看狀態
(1) 管理服務器上查看狀態
[root@node1 ~]# mmm_control show
db1(192.168.0.201) master/ONLINE. Roles: reader(192.168.0.215), writer(192.168.0.211)
db2(192.168.0.202) master/ONLINE. Roles: reader(192.168.0.213)
db3(192.168.0.203) slave/ONLINE. Roles: reader(192.168.0.212)
db4(192.168.0.204) slave/ONLINE. Roles: reader(192.168.0.214)
[root@node1 ~]# mmm_control checks all
db4 ping [last change: 2016/04/27 16:45:49] OK
db4 mysql [last change: 2016/04/27 16:45:49] OK
db4 rep_threads [last change: 2016/04/27 16:45:49] OK
db4 rep_backlog [last change: 2016/04/27 16:45:49] OK: Backlog is null
db2 ping [last change: 2016/04/27 16:45:49] OK
db2 mysql [last change: 2016/04/27 16:45:49] OK
db2 rep_threads [last change: 2016/04/27 16:45:49] OK
db2 rep_backlog [last change: 2016/04/27 16:45:49] OK: Backlog is null
db3 ping [last change: 2016/04/27 16:45:49] OK
db3 mysql [last change: 2016/04/27 16:45:49] OK
db3 rep_threads [last change: 2016/04/27 16:45:49] OK
db3 rep_backlog [last change: 2016/04/27 16:45:49] OK: Backlog is null
db1 ping [last change: 2016/04/27 16:45:49] OK
db1 mysql [last change: 2016/04/27 16:45:49] OK
db1 rep_threads [last change: 2016/04/27 16:45:49] OK
db1 rep_backlog [last change: 2016/04/27 16:45:49] OK: Backlog is null
[root@node1 ~]#
(2) 服務器的啓動VIP地址
[root@node1 ~]# ip a
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:ff:33:6a brd ff:ff:ff:ff:ff:ff
inet 192.168.0.201/24 brd 192.168.0.255 scope global eth0
inet 192.168.0.215/32 scope global eth0
inet 192.168.0.211/32 scope global eth0
inet6 fe80::20c:29ff:feff:336a/64 scope link
valid_lft forever preferred_lft forever
[root@node1 ~]#
[root@node2 ~]# ip a
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:29:8c:60:58 brd ff:ff:ff:ff:ff:ff
inet 192.168.0.202/24 brd 192.168.0.255 scope global eth0
inet 192.168.0.213/32 scope global eth0
inet6 fe80::20c:29ff:fe8c:6058/64 scope link
valid_lft forever preferred_lft forever
[root@node2 ~]#
服務器讀寫採有VIP地址進行讀寫，出現故障時VIP會漂移到其它節點，由其它節點提供服務。
四、MMM高可用測試
首先查看整個集羣的狀態，可以看到整個集羣狀態正常。
[root@node5 ~]# mmm_control show
db1(192.168.0.201) master/ONLINE. Roles: reader(192.168.0.212), writer(192.168.0.211)
db2(192.168.0.202) master/ONLINE. Roles: reader(192.168.0.213)
db3(192.168.0.203) slave/ONLINE. Roles: reader(192.168.0.215)
db4(192.168.0.204) slave/ONLINE. Roles: reader(192.168.0.214)
[[email protected] ~]#
1. 模擬node2宕機，手動停止mysql服務，觀察monitor日誌node2由關閉到啓動的日誌如下：
[[email protected] ~]# tail -f /var/log/mysql-mmm/mmm_mond.log
2016/04/29 09:58:33 FATAL State of host 'db2' changed from ONLINE to HARD_OFFLINE (ping: OK, mysql: not OK)
2016/04/29 10:00:54 FATAL State of host 'db2' changed from HARD_OFFLINE to AWAITING_RECOVERY
2016/04/29 10:01:24 FATAL State of host 'db2' changed from AWAITING_RECOVERY to ONLINE because of auto_set_online(30 seconds). It was in state AWAITING_RECOVERY for 30
seconds
重新查看集羣的最新狀態：
[root@node5 ~]# mmm_control show
db1(192.168.0.201) master/ONLINE. Roles: reader(192.168.0.212), writer(192.168.0.211)
db2(192.168.0.202) master/HARD_OFFLINE. Roles: #從日誌發現db2的狀態有ONLINE轉換爲HARD_OFFLINE
db3(192.168.0.203) slave/ONLINE. Roles: reader(192.168.0.213), reader(192.168.0.215)
db4(192.168.0.204) slave/ONLINE. Roles: reader(192.168.0.214)
[root@node5 ~]# mmm_control checks all
db4 ping [last change: 2016/04/29 09:01:08] OK
db4 mysql [last change: 2016/04/29 09:01:08] OK
db4 rep_threads [last change: 2016/04/29 09:01:08] OK
db4 rep_backlog [last change: 2016/04/29 09:01:08] OK: Backlog is null
db2 ping [last change: 2016/04/29 09:01:08] OK
db2 mysql [last change: 2016/04/29 09:58:33] ERROR: Connect error (host = 192.168.0.202:3306, user = mmm_monitor)! Lost connection to MySQL server at 'reading
initial communication packet', system error: 111 能ping通，說明只是服務死掉了。
db2 rep_threads [last change: 2016/04/29 09:01:08] OK
db2 rep_backlog [last change: 2016/04/29 09:01:08] OK: Backlog is null
db3 ping [last change: 2016/04/29 09:01:08] OK
db3 mysql [last change: 2016/04/29 09:01:08] OK
db3 rep_threads [last change: 2016/04/29 09:01:08] OK
db3 rep_backlog [last change: 2016/04/29 09:01:08] OK: Backlog is null
db1 ping [last change: 2016/04/29 09:01:08] OK
db1 mysql [last change: 2016/04/29 09:56:33] OK
db1 rep_threads [last change: 2016/04/29 09:01:08] OK
db1 rep_backlog [last change: 2016/04/29 09:01:08] OK: Backlog is null
[root@node5 ~]#
啓動node2，node2由HARD_OFFLINE轉到AWAITING_RECOVERY狀態。這裏db2再次接管讀請求。
[root@node5 ~]# mmm_control show
db1(192.168.0.201) master/ONLINE. Roles: reader(192.168.0.212), writer(192.168.0.211)
db2(192.168.0.202) master/ONLINE. Roles: reader(192.168.0.213)
db3(192.168.0.203) slave/ONLINE. Roles: reader(192.168.0.215)
db4(192.168.0.204) slave/ONLINE. Roles: reader(192.168.0.214)
[root@node5 ~]#
2. 模擬node1主庫宕機，查看集羣狀態。
[root@node5 ~]# mmm_control show
db1(192.168.0.201) master/HARD_OFFLINE. Roles:
db2(192.168.0.202) master/ONLINE. Roles: reader(192.168.0.213), writer(192.168.0.211)
db3(192.168.0.203) slave/ONLINE. Roles: reader(192.168.0.212), reader(192.168.0.215)
db4(192.168.0.204) slave/ONLINE. Roles: reader(192.168.0.214)
node1主庫再啓動：
[root@node5 ~]# mmm_control show
db1(192.168.0.201) master/ONLINE. Roles: reader(192.168.0.212)
db2(192.168.0.202) master/ONLINE. Roles: reader(192.168.0.213), writer(192.168.0.211)
db3(192.168.0.203) slave/ONLINE. Roles: reader(192.168.0.215)
db4(192.168.0.204) slave/ONLINE. Roles: reader(192.168.0.214)
可以看到主庫啓動用會自動轉到讀的角色，不會接管主，只到現有的主再次宕機。
[[email protected] ~]# tail -f /var/log/mysql-mmm/mmm_mond.log
2016/04/29 10:03:25 FATAL State of host 'db1' changed from ONLINE to HARD_OFFLINE (ping: OK, mysql: not OK)
2016/04/29 10:06:20 FATAL State of host 'db1' changed from HARD_OFFLINE to AWAITING_RECOVERY
2016/04/29 10:06:51 FATAL State of host 'db1' changed from AWAITING_RECOVERY to ONLINE because of auto_set_online(30 seconds). It was in state AWAITING_RECOVERY for 31
seconds
3. 測試小結：
(1)node2備選主節點宕機不影響集羣的狀態，就是移除了node2備選節點的讀狀態。
(2)node1主節點宕機，由node2備選主節點接管寫角色，node3,node4指向新node2主庫進行復制，node3,node4會自動change master到node2.
(3)問題來了：
如果node1主庫宕機，node2複製應用又落後於node1時就變成了主可寫狀態，這進的數據主無法保證一致性。
如果node2,node3,node4延遲於node1主，這個時node1宕機，node3,node4將會等待數據追上db1後，再重新指向新的主node2進行復制操作，這時的數據也無法保證同步的一致性。
(4)如查採用MMM高可用架構，主，主備選節點機器配置一樣，而且開啓半同步進一步提高安全性或採用MariaDB進行多線程從複製，提高複製的性能。
說明：MMM（Master-Master Replication Manager for MySQL）主主複製管理器, MMM集羣套件具有良好的穩定性、高可用性和可擴展性。當活動的Master節點出現故障時，備用Master節點可以立即接管，而其他的Slave節點也能自動切換到備用Master節點繼續進行同步複製，而無需人爲干涉；MMM架構需要多個節點、多個IP，對服務器數量有要求，在讀寫非常繁忙的業務系統下表現不是很穩定，可能會出現複製延時、切換失效等問題。MMM方案並不太適應於對數據安全性要求很高，並且讀、寫繁忙的環境中。