MMM架構實現MySQL高可用讀寫分離（進階版,包含Amoeba）
環境介紹：
開始部署
第一部分MMM部署安裝
一.安裝mysql5.6-34(雙主雙從都進行安裝)
1.yum -y install gcc perl lua-devel pcre-devel openssl-devel gd-devel gcc-c++ ncurses-devel libaio autoconf
2.tar xvf mysql-5.6.34-linux-glibc2.5-x86_64.tar.gz
3.useradd mysql
4.mv mysql-5.6.34-linux-glibc2.5-x86_64 /data/mysql && chown -R mysql.mysql /data/mysql/
5.vim my.cnf(4臺服務器server_id不同)
[mysqld] socket = /data/mysql/mysql.sock pid-file = /data/mysql/mysql.pid basedir = /data/mysql datadir = /data/mysql/data tmpdir = /data/mysql/data log_error = error.log relay_log = relay.log binlog-ignore-db=mysql,information_schema character_set_server=utf8 log_bin=mysql_bin server_id=1 log_slave_updates=true sync_binlog=1 auto_increment_increment=2 auto_increment_offset=1
6.cd /data/mysql/ && scripts/mysql_install_db --user=mysql --basedir=/data/mysql --datadir=/data/mysql/data/
7.vim /root/.bash_profile 修改 PATH=$PATH:$HOME/bin 為：
PATH=$PATH:$HOME/bin:/data/mysql/bin:/data/mysql/lib
8.source /root/.bash_profile
9.cp support-files/mysql.server /etc/init.d/mysql
10.service mysql start
二.實現雙主復制
因為我的環境已經是雙主復制，這裏只把實現的命令寫出來
1.在192.168.6.30(master1)上執行
grant replication slave on *.* to 'replication'@'192.168.6.%' identified by '123456'; show master status; change master to master_host='192.168.6.31',master_user='replication',master_password='123456',master_log_file='mysql_bin.000001',master_log_pos=330; start slave; show slave status\G;
2.,在192.168.6.31(master2)上執行
grant replication slave on *.* to 'replication'@'192.168.6.%' identified by '123456'; show master status; change master to master_host='192.168.6.30',master_user='replication',master_password='123456',master_log_file='mysql_bin.000001',master_log_pos=334; start slave; show slave status\G;
觀察是否是兩個YES，如果是則雙主同步成功
三.如何加入正在運行的mysql雙主復制（不停服務加slave復制）
其實平時能讓運維從0搭建一套Mysql集群是很少見的，大都是現成的環境然後進行升級，裏面有數據，並且有重要業務不能停服務，筆者在做實驗的時候以為可以直接加入，結果每秒都在變化的POS根本不能直接加入進去，於是上網查了一些資料，找到了不停服務也能加入的方法。
1.導出全庫，將binlog點和pos寫到文件中
mysqldump --skip-lock-tables --single-transaction --master-data=2 -A > ~/dump.sql
--master-data：默認等於1，將dump起始（change master to）binlog點和pos值寫到結果中，等於2是將change master to寫到結果中並註釋。-A就是備份全庫
--single_transaction：導出開始時設置事務隔離狀態，並使用一致性快照開始事務，然後unlock tables;而lock-tables是鎖住一張表不能寫操作，直到dump完畢。
2.查詢裏面的POS值和binlog值
head dump.sql -n80 | grep "MASTER_LOG_POS"
CHANGE MASTER TO MASTER_LOG_FILE='mysql_bin.000149', MASTER_LOG_POS=120;
3.將文件進行壓縮
gzip ~/dump.sql
4.拷貝文件到兩個從服務器，解壓後導入
scp dump.gz 192.168.6.107:/root scp dump.gz 192.168.6.108:/root gunzip dump.gz mysql < dump.sql
5.加入兩臺Slave服務器
在192.168.6.107(slave1)上執行:
change master to master_host='192.168.6.30',master_user='replication',master_password='123456',master_log_file='mysql_bin.000149',master_log_pos=120; start slave; show slave status\G;
在192.168.6.108(slave2)上執行:
change master to master_host='192.168.6.30',master_user='replication',master_password='123456',master_log_file='mysql_bin.000149',master_log_pos=120; start slave; show slave status\G;
觀察是否是兩個YES，如果是則主從同步成功
四.安裝部署MMM服務器
1.安裝mmm軟件包（4臺都裝）
yum -y install mysql-mmm* #Centos7和6安裝的包數量不一致，這個可以暫時忽略不計
2.在Monitor端編輯mysql-mmm配置文件
vim /etc/mysql-mmm/mmm_common.conf
active_master_role writer
<host default>
cluster_interface ens160 #Centos6的網卡名稱為eth0
pid_path /run/mysql-mmm-agent.pid #Centos6的路徑為/var/run/mysql-mmm-agent.pid
bin_path /usr/libexec/mysql-mmm/
replication_user replication
replication_password 123456
agent_user mmm_agent
agent_password RepAgent
</host>
<host db1> #agent連接時使用的名稱
ip 192.168.6.30
mode master #角色
peer db1
</host>
<host db2>
ip 192.168.6.31
mode master
peer db2
</host>
<host db3>
ip 192.168.6.107
mode slave
</host>
<host db4>
ip 192.168.6.108
mode slave
</host>
<role writer>
hosts db1, db2
ips 192.168.6.35 #VIP
mode exclusive
</role>
<role reader>
hosts db3, db4
ips 192.168.6.36,192.168.6.37
mode balanced #負載均衡
</role>3.將此文件復制到4個服務器端的指定位置,並更改上面指定位置的文字
scp /etc/mysql-mmm/mmm_common.conf 192.168.6.30:/etc/mysql-mmm/mmm_common.conf scp /etc/mysql-mmm/mmm_common.conf 192.168.6.31:/etc/mysql-mmm/mmm_common.conf scp /etc/mysql-mmm/mmm_common.conf 192.168.6.107:/etc/mysql-mmm/mmm_common.conf scp /etc/mysql-mmm/mmm_common.conf 192.168.6.108:/etc/mysql-mmm/mmm_common.conf
4.在Monitor端編輯mmm-mon文件
vim /etc/mysql-mmm/mmm_mon.conf
include mmm_common.conf
<monitor>
ip 127.0.0.1
pid_path /run/mysql-mmm-monitor.pid #因為Monitor端是Centos7所以不需要更改
bin_path /usr/libexec/mysql-mmm
status_path /var/lib/mysql-mmm/mmm_mond.status
ping_ips 192.168.6.30,192.168.6.31,192.168.6.107,192.168.6.108 #把4臺服務器的ip寫好
auto_set_online 10 #設置10秒無響應就顯示離線
</monitor>
<host default>
monitor_user mmm_monitor
monitor_password RepMonitor
</host>
debug 05.啟動監控
systemctl start mysql-mmm-monitor
五.配置四臺mysql服務器的mysql-mmm代理
1.剛才在配置文件中有設置監控用戶和代理用戶，需要在數據庫中進行授權,因為是主從復制，只需要在主上進行授權即可
grant super, replication client, process on *.* to 'mmm_agent'@'192.168.6.%' identified by 'RepAgent'; grant replication client on *.* to 'mmm_monitor'@'192.168.6.%' identified by 'RepMonitor'; flush privleges;
2.分別在4臺服務器上編輯mmm_agent文件
vim /etc/mysql-mmm/mmm_agent.conf
this db3 #是哪個服務器就寫哪個host，跟mmm_common.conf對應
3.開啟對應防火墻端口
iptables -I INPUT -p tcp --dport 9989 -j ACCEPT #代理服務的端口，需要與Monitor端進行通訊
4.啟動代理服務
systemctl start mysql-mmm-agent #Centos7 service mysql-mmm-agent start #Centos6
六.查看Mysql集群狀態及故障測試
1.在Monitor端執行mmm_control show
db1(192.168.6.30) master/ONLINE. Roles: writer(192.168.6.35)
db2(192.168.6.31) master/ONLINE. Roles:
db3(192.168.6.107) slave/ONLINE. Roles: reader(192.168.6.36)
db4(192.168.6.108) slave/ONLINE. Roles: reader(192.168.6.37)
出現如上顯示則表示配置正確
2.故障測試
①在master端執行 service mysql stop
db1(192.168.6.30) master/HARD_OFFLINE. Roles:
db2(192.168.6.31) master/ONLINE. Roles: writer(192.168.6.35)
db3(192.168.6.107) slave/ONLINE. Roles: reader(192.168.6.36)
db4(192.168.6.108) slave/ONLINE. Roles: reader(192.168.6.37)
VIP成功轉移
②在slave端執行 service mysql stop
db1(192.168.6.30) master/HARD_OFFLINE. Roles:
db2(192.168.6.31) master/ONLINE. Roles: writer(192.168.6.35)
db3(192.168.6.107) slave/ONLINE. Roles: reader(192.168.6.36),reader(192.168.6.37)
db4(192.168.6.108) slave/HARD_OFFLINE. Roles:
VIP成功轉移,當重新啟動mysql服務時，狀態就會恢復到ONLINE，到此MMM架構的MySQL集群就已經配置完成了
第二部分 利用Amoeba實現單點登錄
由於MMM提供3個出口，1個寫2個讀，要想連接這個架構需要把應用也分成讀和寫兩部分，這裏就可以使用中間件來整合這個架構
Amoeba是什麽？
Amoeba(變形蟲)項目,該開源框架於2008年 開始發布一款 Amoeba for Mysql軟件。這個軟件致力於MySQL的分布式數據庫前端代理層，它主要在應用層訪問MySQL的時候充當SQL路由功能，專註於分布式數據庫代理層（Database Proxy）開發。座落與 Client、DB Server(s)之間,對客戶端透明。具有負載均衡、高可用性、SQL 過濾、讀寫分離、可路由相關的到目標數據庫、可並發請求多臺數據庫合並結果。 通過Amoeba你能夠完成多數據源的高可用、負載均衡、數據切片的功能，目前Amoeba已在很多 企業的生產線上面使用。
開始部署
1.下載amoeba
wget https://sourceforge.net/projects/amoeba/files/Amoeba%20for%20mysql/3.x/amoeba-mysql-3.0.5-RC-distribution.zip
2.安裝
unzip amoeba-mysql-3.0.5-RC-distribution.zip mv amoeba-mysql-3.0.5-RC /usr/local/amoeba/
3.Java環境安裝
tar xvf jdk-8u111-linux-x64.tar.gz mv jdk1.8.0_111/ /usr/local/jdk
vim /etc/profile
export JAVA_HOME=/usr/local/jdk
export PATH=${PATH}:${JAVA_HOME}/bin
export CLASSPATH=${CLASSPATH}:${JAVA_HOME}/libsource /etc/profile
4.配置amoeba
vim /usr/local/amoeba/conf/dbServers.xml
<?xml version="1.0" encoding="gbk"?>
<!DOCTYPE amoeba:dbServers SYSTEM "dbserver.dtd">
<amoeba:dbServers xmlns:amoeba="http://amoeba.meidusa.com/">
<!--
Each dbServer needs to be configured into a Pool,
If you need to configure multiple dbServer with load balancing that can be simplified by the following configuration:
add attribute with name virtual = "true" in dbServer, but the configuration does not allow the element with name factoryConfig
such as 'multiPool' dbServer
-->
<dbServer name="abstractServer" abstractive="true">
<factoryConfig class="com.meidusa.amoeba.mysql.net.MysqlServerConnectionFactory">
<property name="connectionManager">${defaultManager}</property>
<property name="sendBufferSize">64</property>
<property name="receiveBufferSize">128</property>
<!-- mysql port --
<property name="port">3306</property> <!--設置Amoeba要連接的mysql數據庫的端口，默認是3306-->
<!-- mysql schema -->
<property name="schema">amoeba</property> <!--設置缺省的數據庫-->
<!-- mysql user -->
<property name="user">amoeba</property> <!--設置amoeba連接後端數據庫服務器的賬號和密碼,需要在Mysql中授權-->
<property name="password">12345</property>
</factoryConfig>
<poolConfig class="com.meidusa.toolkit.common.poolable.PoolableObjectPool">
<property name="maxActive">500</property>
<property name="maxIdle">500</property>
<property name="minIdle">1</property>
<property name="minEvictableIdleTimeMillis">600000</property>
<property name="timeBetweenEvictionRunsMillis">600000</property>
<property name="testOnBorrow">true</property>
<property name="testOnReturn">true</property>
<property name="testWhileIdle">true</property>
</poolConfig>
</dbServer>
<dbServer name="writedb" parent="abstractServer"> <!--設置寫服務器名稱-->
<factoryConfig>
<!-- mysql ip -->
<property name="ipAddress">192.168.6.35</property> <!--寫服務器VIP-->
</factoryConfig>
</dbServer>
<dbServer name="slave1" parent="abstractServer"> <!--設置讀服務器名稱-->
<factoryConfig>
<!-- mysql ip -->
<property name="ipAddress">192.168.6.36</property> <!--讀服務器VIP-->
</factoryConfig>
</dbServer>
<dbServer name="slave2" parent="abstractServer">
<factoryConfig>
<!-- mysql ip -->
<property name="ipAddress">192.168.6.37</property>
</factoryConfig>
</dbServer>
<dbServer name="myslaves" virtual="true"> <!--讀服務器池的名稱-->
<poolConfig class="com.meidusa.amoeba.server.MultipleServerPool">
<!-- Load balancing strategy: 1=ROUNDROBIN , 2=WEIGHTBASED , 3=HA-->
<property name="loadbalance">1</property> <!--調度算法選擇輪詢-->
<!-- Separated by commas,such as: server1,server2,server1 -->
<property name="poolNames">slave1,slave2</property> <!--讀服務器池的服務器-->
</poolConfig>
</dbServer>
</amoeba:dbServers>vim /usr/local/amoeba/conf/amoeba.xml
<?xml version="1.0" encoding="gbk"?>
<!DOCTYPE amoeba:configuration SYSTEM "amoeba.dtd">
<amoeba:configuration xmlns:amoeba="http://amoeba.meidusa.com/">
<proxy>
<!-- service class must implements com.meidusa.amoeba.service.Service -->
<service name="Amoeba for Mysql" class="com.meidusa.amoeba.mysql.server.MySQLService">
<!-- port -->
<property name="port">8066</property> <!--登錄數據庫時連接的端口-->
<!-- bind ipAddress -->
<!--
<property name="ipAddress">127.0.0.1</property> <!--登錄數據庫時連接的IP-->
-->
<property name="connectionFactory">
<bean class="com.meidusa.amoeba.mysql.net.MysqlClientConnectionFactory">
<property name="sendBufferSize">128</property>
<property name="receiveBufferSize">64</property>
</bean>
</property>
<property name="authenticateProvider">
<bean class="com.meidusa.amoeba.mysql.server.MysqlClientAuthenticator">
<property name="user">root</property> <!--連接數據庫時使用的用戶名-->
<property name="password">123456</property> <!--連接數據庫時使用的密碼-->
<property name="filter">
<bean class="com.meidusa.toolkit.net.authenticate.server.IPAccessController">
<property name="ipFile">${amoeba.home}/conf/access_list.conf</property>
</bean>
</property>
</bean>
</property>
</service>
<runtime class="com.meidusa.amoeba.mysql.context.MysqlRuntimeContext">
<!-- proxy server client process thread size -->
<property name="executeThreadSize">128</property>
<!-- per connection cache prepared statement size -->
<property name="statementCacheSize">500</property>
<!-- default charset -->
<property name="serverCharset">utf8</property>
<!-- query timeout( default: 60 second , TimeUnit:second) -->
<property name="queryTimeout">60</property>
</runtime>
</proxy>
<!--
Each ConnectionManager will start as thread
manager responsible for the Connection IO read , Death Detection
-->
<connectionManagerList>
<connectionManager name="defaultManager" class="com.meidusa.toolkit.net.MultiConnectionManagerWrapper">
<property name="subManagerClassName">com.meidusa.toolkit.net.AuthingableConnectionManager</property>
</connectionManager>
</connectionManagerList>
<!-- default using file loader -->
<dbServerLoader class="com.meidusa.amoeba.context.DBServerConfigFileLoader">
<property name="configFile">${amoeba.home}/conf/dbServers.xml</property>
</dbServerLoader>
<queryRouter class="com.meidusa.amoeba.mysql.parser.MysqlQueryRouter">
<property name="ruleLoader">
<bean class="com.meidusa.amoeba.route.TableRuleFileLoader">
<property name="ruleFile">${amoeba.home}/conf/rule.xml</property>
<property name="functionFile">${amoeba.home}/conf/ruleFunctionMap.xml</property>
</bean>
</property>
<property name="sqlFunctionFile">${amoeba.home}/conf/functionMap.xml</property>
<property name="LRUMapSize">1500</property>
<property name="defaultPool">writedb</property> <!--默認使用的服務器名-->
<property name="writePool">writedb</property> <!--寫服務器池的名稱-->
<property name="readPool">myslaves</property> <!--讀服務器池的名稱-->
<property name="needParse">true</property>
</queryRouter>
</amoeba:configuration>5.授權amoeba用戶
在master1上執行
create database amoeba; grant all on *.* to amoeba@"192.168.6.109"; flush privilegs;
6.修改JVM數值
vim /usr/local/amoeba/jvm.properties
JVM_OPTIONS="-server -Xms256m -Xmx1024m -Xss256k #將Xss數值改成256k以上，否則服務無法啟動
7.啟動服務
/usr/local/amoeba/bin/launcher &
8.在Monitor端登錄Amoeba
mysql -uroot -p123456 -h127.0.0.1 -P8066
出現如上顯示登錄成功。
附上最終架構圖：
MMM架構實現MySQL高可用讀寫分離（進階版,包含Amoeba）