MySQL-MMM的读写分离及高可用

青衣

　　用途
  mmm是基于信息探测方式进行mysql主从复制架构的监测与故障转移
  mmm可以做到负载均衡,100%的数据可用性
  mmm所涉及的检查项   服务器可达性,服务可达性,复制线程可控性
　　如图: 当 master1在宕机时, mmm可以将之前分摊的流量进行转移,甚至于将从服务器提升为主
　　延续双主模型.

mmm-agent端状态一览
    online 节点可用
    replication_delay 复制延迟或无法进行(检查req_backlog文件)
    replication_fail 复制失败
    awating_recovery 等待恢复
    hard_offline  主机离线
    admin_offline  主控端离线
    unknown  未知错误

规划如下
172.16.43.200  主控机器负责监测与资源的管理
172.16.43.1  vip(172.16.43.11)  master1 主主复制第一台(可读写)
172.16.43.2  vip(172.16.43.12)  master2 主主复制第二台(只读)
172.16.43.3  vip(172.16.43.13)  slave   主从方式复制 1 , 2 的信息(只读)

　　

　　实验过程如下
　　主控安装安装: ansible, mysql客户端, mysql-mmm
集群节点安装如下: mariadb, mysql-agent
　　

　　i)  主机互信,ansible部署的关键

yum -y install ansible-1.5.4-1.el6.noarch.rpm
vim /etc/ansible/hosts
# 主控节点完成
[masterserver]
master1.king.com
master2.king.com
.
[slaveserver]
slave.king.com
# 主控节点完成互信多个被管理节点不重复了
ssh-keygen -t rsa
ssh-copy-id -i .ssh/id_rsa.pub root@slave.king.com
.
# 主控节点完成安装mysql-mmm*
yum -y install mysql mysql-mmm*
.
# 实现mmm配置文件有两种
# 1. 主控端   mmm_common, mmm_mon
# 2. 集群端   mmm_common, mmm_agent
# 主控端配置common   /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/sbin
replication_user    repl
replication_password  repl
agent_user        mmm_agent
agent_password      agent_password
</host>
.
<host master1>
ip 172.16.43.1
mode master
peer master2
</host>
.
<host master2>
ip 172.16.43.2
mode master
peer master1
</host>
.
<host slave>
ip 172.16.43.3
mode slave
</host>
.
# 谁能写  vip是多少(参照本文配置)
<role writer>
hosts master1, master2
ips 172.16.43.11
mode exclusive
</role>
.
# 谁能读  vip是多少(参照本文配置)
<role reader>
hosts master1, master2
ips 172.16.43.11, 172.16.43.12, 172.16.43.13
mode balanced
</role>
.
# 主控节点完成
vim /etc/mysql-mmm/mmm_mon.conf
# 需要改 ping_ips, 把你想监控的集群加入进入就可以了
# 172.16.0.1 是网关ip
ping_ips 172.16.0.1, 172.16.43.1, 172.16.43.2, 172.16.43.3
.
<host default>
  monitor_user    mmm_monitor
  monitor_password  monitor_password
</host>

　　ii) 安装 mariadb, mysql-agent安装与配置

# 编写 install.yaml, 使用ansible-playbook执行按成安装配置
- hosts: masterserver:slaveserver
  remote_user: root
  tasks:
  - name: mariadb install
    copy: src=/root/mariadb-10.0.10-linux-x86_64.tar.gz dest=/tmp
  - name: conf service
    command: tar xf /tmp/mariadb-10.0.10-linux-x86_64.tar.gz -C /usr/local
  ###############################################################
  # 见谅下面的 name: x , 写了半天才知道不能在一个name中包含多个command,所以..  #
  ###############################################################
  - name: 1
    command: ln -sv /usr/local/mariadb-10.0.10-linux-x86_64 /usr/local/mysql
  - name: 2
    command: cp /usr/local/mysql/support-files/mysql.server /etc/rc.d/init.d/mysqld
  - name: 3
    command: chmod +x /etc/rc.d/init.d/mysqld
  - name: 4
    command: mkdir /mydata/data -p
  - name: 5
    command: mkdir /mydata/binlogs -p
  - name: 6
    command: mkdir /mydata/relaylogs -p
  - name: 7
    command: useradd -r -s /sbin/nologin mysql -U
  - name: 8
    command: chown mysql.mysql /mydata -R
  - name: 9
    command: chown mysql.mysql /usr/local/mysql -R
  - name: 9-1
    # 将本地配置好的安装文件移植到集群节点, 解压的配置文件可能不符合要求
    copy: src=/usr/local/mysql/scripts/mysql_install_db dest=/usr/local/mysql/scripts/
  - name: 10
    command: /usr/local/mysql/scripts/mysql_install_db --user=mysql --datadir=/mydata/data
  - name: mariadb conf
    # 此处配置文件 /etc/my.cnf 配置好
    # log-bin=/mydata/binlogs/mysql-bin
    # relay-log=/mydata/relaylogs/relay-bin
    # datadir=/mydata/data
    copy: src=/etc/my.cnf dest=/etc/my.cnf
    notify:
    - restart mariadb
  - name: install mysql-mmm-agent
    yum: name=mysql-mmm-agent state=present
  - name: conf mysql-mmm-agent
    copy: src=/etc/mysql-mmm/mmm_common.conf dest=/etc/mysql-mmm
  - name: modify mmm_agent.conf
    command: sed -i 's@^\(this[[:space:]]\).*@\1` item `@' /etc/mysql-mmm/mmm_agent.conf
    with_items:
    - master1
    - master2
    - slave
  - name: 11
    command: sed -i 's@^\(ENABLED=\).*@\11@' /etc/default/mysql-mmm-agent
  handlers:
  - name: restart mariadb
    service: name=mysqld state=restarted

　　不熟yaml的童鞋可以 yaml.org

　　

　　iii) 实现双主复制与主从复制

1. 双主配置 (master1.king.com , master2.king.com)
vim /etc/my.cnf
service-id  = 1 | 2 (不可相同分别设置)
log-slave-updates = 1
*** 自动增长列的配置要隔开
auto-increment-offset = 1
auto-increment-increment = 2
2. 均授权复制账号给对方, 进入命令行
mysql> grant replication client, replication slave on *.* to repl@'172.16.%.%' identified by 'repl';
mysql> flush privileges;
3. 均授权监控账号
mysql> grant replication client on *.* to 'mmm_monitor'@'172.16.%.%' identified by 'monitor_password';
mysql> grant super, replication client, process on *.* to 'mmm_agent'@'172.16.%.%' identified by 'agent_password';
4. 均连接对方服务器
mysql> change master to master_host='172.16.43.1',master_user='repl', master_password='repl', master_log_file='mysql-bin.000001', master_log_pos=312;
5. 均启动复制线程
mysql> start salve;
.
.
# 主从复制 (slave.king.com)
1、改server-id 配置文件中
vim /etc/my.cnf
server-id     = 3
2, 授权监控账号,进入命令行 mysql
mysql> grant replication client on *.* to 'mmm_monitor'@'172.16.%.%' identified by 'monitor_password';
mysql> grant super, replication client, process on *.* to 'mmm_agent'@'172.16.%.%' identified by 'agent_password';
3、连接主服务器
mysql> change master to master_host='172.16.43.1',master_user='repl', master_password='repl', master_log_file='mysql-bin.000001', master_log_pos=312;
4、启动复制线程
mysql> start slave;

　　

　　iv) 启动测试

# mmm的停止与启动
/etc/init.d/mysql-mmm-agent stop | start
/etc/init.d/mysql-mmm-monitor stop | start
.
# 离线或上线一个节点
mmm_control set_offline master1
mmm_control set_online master1
.
# 查看节点状态
mmm_control show

　　测试1
　　如图: 让 master2 离线,读写均不受影响 (抱歉此图没有截图下来,,但最后结果图中有所显示  : )

　　

　　测试2:
　　再来一次测试,将master1主节点下线,我们观察一下情况, 在图中我们可以看到读写资源已经迁移到了master2节点

　　总结
　　经过测试可以看出MMM在mysql的高可用表现非常良好,无论是只剩一主还是一主一从,数据的同步确实比较及时准确
　　生产环境还需观察...