Mha-Atlas-MySQL高可用方案实践(一)
Mha-Atlas-MySQL高可用方案实践
一,mysql-mha环境准备
1.1 实验环境:
1.2 软件包
用到得所有包
链接:https://pan.baidu.com/s/19tiKXNEW4C6oWi9OFmcDYA
提取码:be07
1) mha管理节点安装包:
mha4mysql-manager-0.56-0.el6.noarch.rpm
mha4mysql-manager-0.56.tar.gz
2) mha node数据节点安装包:
mha4mysql-node-0.56-0.el6.noarch.rpm
mha4mysql-node-0.56.tar.gz
3) mysql中间件:
Atlas-2.2.1.el6.x86_64.rpm
4) mysql源码安装包
mysql-5.6.17-linux-glibc2.5-x86_64.tar
1.3 主机名映射
echo "192.168.200.69 sl-Mysql-Mater" >> /etc/hosts
echo "192.168.200.79 sl-Mysql-Slave1" >> /etc/hosts
echo "192.168.200.80 sl-Mysql-Slave2" >> /etc/hosts
scp /etc/hosts root@192.168.200.79:/etc/
scp /etc/hosts root@192.168.200.80:/etc/
1.4 关闭selinux和iptables(三台都需要关闭)
[root@mysql-db01 ~]# vim /etc/sysconfig/selinux
[root@mysql-db01 ~]# cat /etc/sysconfig/selinux | grep -v "#"
SELINUX=disabled
SELINUXTYPE=targeted
[root@mysql-db01 ~]# setenforce 0
[root@mysql-db01 ~]# service iptables stop
iptables: Setting chains to policy ACCEPT: filter [ OK ]
iptables: Flushing firewall rules: [ OK ]
iptables: Unloading modules: [ OK ]
[root@mysql-db01 ~]# chkconfig iptables off
二,简介
2.1 作者简介
姓名:松信嘉范
MySQL/Linux专家
2001年索尼公司入职
2001年开始使用oracle
2004年开始使用MySQL
2006年9月-2010年8月MySQL从事顾问
2010年-2012年DeNA
2012年至今Facebook
2.2 软件简介
- MHA(Master High Availability)目前在MySQL高可用方面是一个相对成熟的解决方案,是一套优秀的作为MySQL高可用性环境下故障切换和主从提升的高可用软件。在MySQL故障切换过程中,MHA能做到0~30秒之内自动完成数据库的故障切换操作,并且在进行故障切换过程中,MHA能最大程度上保证数据库的一致性,以达到真正意义上的高可用。
- MHA由两部分组成:MHA Manager(管理节点)和MHA Node(数据节点)。MHA Manager可以独立部署在一台独立的机器上管理多个Master-Slave集群,也可以部署在一台Slave上。当Master出现故障时,它可以自动将最新数据的Slave提升为新的Master,然后将所有其他的Slave重新指向新的Master。整个故障转移过程对应程序是完全透明的。
2.3 工作流程
- 从宕机崩溃的master保存二进制日志事件(binlog events);
- 识别含有最新更新的slave;
- 应用差异的中继日志(relay log)到其他的slave;
- 应用从master保存的二进制日志事件(binlog events);
- 提升一个slave为新的master;
- 使其他的slave连接新的master进行复制;
1,复制主库binlog日志出来(因为还有没来的及复制到从库的二进制日志呢)
2,找出relaylog日志最全的从库 (每个从库复制的速度是有差异的)
3,将最全的relaylog日志在所有从库中同步(第一次数据同步)
4,将之前最全的那个从库提升为主库
5,将复制出来的binlog日志,放到新提升的主库里
6,其他所有从库重新指向新提升的主库,继续主从复制。
2.5 MHA工具介绍
MHA软件由两部分组成,Manager工具包和Node工具包,具体的说明如下
#Manager工具包主要包括以下几个工具
masterha_check_ssh #检查MHA的SSH配置状况
masterha_check_repl #检查MySQL复制状况
masterha_check_status #检测当前MHA运行状态
masterha_master_monitor #检测master是否宕机
masterha_manger #启动MHA
masterha_master_switch #控制故障转移(自动或者手动)
masterha_conf_host #添加或删除配置的server信息
masterha_secondary_check #试图建立TCP连接从远程服务器
masterha_stop #停止MHA
#Node工具包主要包括以下几个工具
save_binary_logs #保存和复制master的二进制日志
apply_diff_relay_logs #识别差异的中继日志事件
filter_mysqlbinlog #去除不必要的ROLLBACK事件
purge_relay_logs #清除中继日志
三,mysql环境准备
3.1 环境检查
3.2 安装mysql
3.2.1 安装包准备
上方有所有软件包得下载链接
3.2.2 安装(3台都装)
yum -y install ncurses-devel libaio
tar xf mysql-5.6.17-linux-glibc2.5-x86_64.tar.gz -C /usr/local/
ln -s /usr/local/mysql-5.6.17-linux-glibc2.5-x86_64 /usr/local/mysql
useradd mysql -s /sbin/nologin -M
/usr/local/mysql/scripts/mysql_install_db --user=mysql --basedir=/usr/local/mysql --datadir=/usr/local/mysql/data/
/bin/cp /usr/local/mysql/support-files/my-default.cnf /etc/my.cnf
/bin/cp /usr/local/mysql/support-files/mysql.server /etc/init.d/mysqld
ln -s /usr/local/mysql/bin/* /usr/local/bin/
which mysqladmin
3.2.3 加入开机自启动并启动mysql
chkconfig mysqld on
chkconfig mysqld --list
/etc/init.d/mysqld start
3.2.4 配置密码
mysqladmin -uroot password 'linyaonie'
四,配置基于GTID的主从复制
4.1 先决条件
- 主库和从库都要开启binlog
- 主库和从库server-id不同
- 要有主从复制用户
4.2 主库操作(mysql-db01)
4.2.1 修改配置文件
#修改主库配置文件/etc/my.cnf
[root@mysql-db01 mysql]# cat /etc/my.cnf
[client]
socket = /usr/local/mysql/data/mysql.sock
[mysqld]
lower_case_table_names = 1
default-storage-engine = InnoDB
port = 3306
datadir = /usr/local/mysql/data
character-set-server = utf8
socket = /usr/local/mysql/data/mysql.sock
log_bin = mysql-bin #开启binlog日志
server_id = 1 #设置server_id
innodb_buffer_pool_size = 200M
slave-parallel-workers = 8
thread_cache_size = 600
back_log = 600
slave_net_timeout = 60
max_binlog_size = 512M
key_buffer_size = 8M
query_cache_size = 64M
join_buffer_size = 2M
sort_buffer_size = 2M
query_cache_type = 1
thread_stack = 192K
#重启动MySQL服务
[root@mysql-db01 mysql]# /etc/init.d/mysqld restart
Shutting down MySQL.. SUCCESS!
Starting MySQL. SUCCESS!
4.2.2 登陆MySQL删除不必要的用户并创建主从复制用户
1)删除不必要的用户
mysql> select user,host from mysql.user;
+------+------------+
| user | host |
+------+------------+
| root | 127.0.0.1 |
| root | ::1 |
| | localhost |
| root | localhost |
| | mysql-db01 |
| root | mysql-db01 |
+------+------------+
6 rows in set (0.00 sec)
mysql> drop user root@'127.0.0.1';
Query OK, 0 rows affected (0.00 sec)
mysql> drop user root@'::1';
Query OK, 0 rows affected (0.00 sec)
mysql> drop user ' '@'localhost';
Query OK, 0 rows affected (0.00 sec)
mysql> drop user ' '@'mysql-db01';
Query OK, 0 rows affected (0.00 sec)
mysql> select user,host from mysql.user;
+------+------------+
| user | host |
+------+------------+
| root | localhost |
| root | mysql-db01 |
+------+------------+
2 rows in set (0.00 sec)
2)创建主从复制用户
mysql> grant replication slave on *.* to rep@'192.168.0.%' identified by '123123';
Query OK, 0 rows affected (0.00 sec)
mysql> select user,host from mysql.user;
+------+-------------+
| user | host |
+------+-------------+
| rep | 192.168.0.% |
| root | localhost |
| root | mysql-db01 |
+------+-------------+
3 rows in set (0.00 sec)
mysql> show grants for rep@'192.168.0.%';
+--------------------------------------------------------------------------------------------------------------------------+
| Grants for rep@192.168.0.% |
+--------------------------------------------------------------------------------------------------------------------------+
| GRANT REPLICATION SLAVE ON *.* TO 'rep'@'192.168.0.%' IDENTIFIED BY PASSWORD '*E56A114692FE0DE073F9A1DD68A00EEB9703F3F1' |
+--------------------------------------------------------------------------------------------------------------------------+
1 row in set (0.00 sec)
4.3 从库操作(mysql-db02和mysql-db03)
4.3.1 修改配置文件
#修改mysql-db02配置文件(和mysql-db01配置文件一致)
#只需要修改server-id = 5选项
[root@mysql-db02 ~]# cat /etc/my.cnf
[client]
socket = /usr/local/mysql/data/mysql.sock
[mysqld]
lower_case_table_names = 1
default-storage-engine = InnoDB
port = 3306
datadir = /usr/local/mysql/data
character-set-server = utf8
socket = /usr/local/mysql/data/mysql.sock
log_bin = mysql-bin #从binlog也要打开
server_id = 5 #仅需修改此项
innodb_buffer_pool_size = 200M
slave-parallel-workers = 8
thread_cache_size = 600
back_log = 600
slave_net_timeout = 60
max_binlog_size = 512M
key_buffer_size = 8M
query_cache_size = 64M
join_buffer_size = 2M
sort_buffer_size = 2M
query_cache_type = 1
thread_stack = 192K
[root@mysql-db02 ~]# /etc/init.d/mysqld restart #重启mysql
Shutting down MySQL.. SUCCESS!
Starting MySQL. SUCCESS!
#修改mysql-db03配置文件(和mysql-db01配置文件一致)
#只需要修改server-id = 10选项
[root@mysql-db03 ~]# cat /etc/my.cnf
[client]
socket = /usr/local/mysql/data/mysql.sock
[mysqld]
lower_case_table_names = 1
default-storage-engine = InnoDB
port = 3306
datadir = /usr/local/mysql/data
character-set-server = utf8
socket = /usr/local/mysql/data/mysql.sock
log_bin = mysql-bin #从binlog也要打开
server_id = 10 #只需修改此项
innodb_buffer_pool_size = 200M
slave-parallel-workers = 8
thread_cache_size = 600
back_log = 600
slave_net_timeout = 60
max_binlog_size = 512M
key_buffer_size = 8M
query_cache_size = 64M
join_buffer_size = 2M
sort_buffer_size = 2M
query_cache_type = 1
thread_stack = 192K
[root@mysql-db03 ~]# /etc/init.d/mysqld restart #重启mysql
Shutting down MySQL.. SUCCESS!
Starting MySQL. SUCCESS!
特别提示:
在以往如果是基于binlog日志的主从复制,则必须要记住主库的master状态信息。
但是在MySQL5.6版本里多了一个Gtid的功能,可以自动记录主从复制位置点的信息,并在日志中输出出来。
4.4 开启GTID
#没开启之前先看一下GTID状态
mysql> show global variables like '%gtid%';
+--------------------------+-------+
| Variable_name | Value |
+--------------------------+-------+
| enforce_gtid_consistency | OFF |
| gtid_executed | |
| gtid_mode | OFF |
| gtid_owned | |
| gtid_purged | |
+--------------------------+-------+
5 rows in set (0.00 sec)
编辑mysql配置文件(主库从库都需要修改)
mysql-db01,mysql-db02,mysql-db03都需要加入上图的上行代码
修改完配置文件以后重启动数据库
[root@mysql-db01 ~]# /etc/init.d/mysqld restart
Shutting down MySQL.. SUCCESS!
Starting MySQL. SUCCESS!
[root@mysql-db02 ~]# /etc/init.d/mysqld restart
Shutting down MySQL.. SUCCESS!
Starting MySQL. SUCCESS!
[root@mysql-db03 ~]# /etc/init.d/mysqld restart
Shutting down MySQL.. SUCCESS!
Starting MySQL. SUCCESS!
再次查看GTID状态
[root@mysql-db01 ~]# mysql -uroot -p123123
Warning: Using a password on the command line interface can be insecure.
Welcome to the MySQL monitor. Commands end with ; or g.
Your MySQL connection id is 1
Server version: 5.6.17-log MySQL Community Server (GPL)
Copyright (c) 2000, 2014, Oracle and/or its affiliates. All rights reserved.
Oracle is a registered trademark of Oracle Corporation and/or its
affiliates. Other names may be trademarks of their respective
owners.
Type 'help;' or 'h' for help. Type 'c' to clear the current input statement.
mysql> show global variables like '%gtid%';
+--------------------------+-------+
| Variable_name | Value |
+--------------------------+-------+
| enforce_gtid_consistency | ON | #执行GTID一致
| gtid_executed | |
| gtid_mode | ON | #开启GTID模块
| gtid_owned | |
| gtid_purged | |
+--------------------------+-------+
5 rows in set (0.00 sec)
mysql>
再次提示:
主库从库都必须要开启GTID,否则在做主从复制的时候就会报错
4.5 配置主从复制(mysql-db02,mysql-db03)
mysql> change master to
-> master_host='192.168.0.51', #主库IP
-> master_user='rep', #主库复制用户
-> master_password='123123', #主库复制用密码
-> master_auto_position=1; #GTID位置点(自动追踪需要同步的position)
Query OK, 0 rows affected, 2 warnings (0.00 sec)
4.6 开启从库的主从复制功能(mysql-db02,mysql-db03)
mysql> start slave; #开启主从同步功能
Query OK, 0 rows affected, 1 warning (0.01 sec)
mysql> show slave statusG
*************************** 1. row ***************************
Slave_IO_State: Waiting for master to send event
Master_Host: 192.168.0.51
Master_User: rep
Master_Port: 3306
Connect_Retry: 60
Master_Log_File: mysql-bin.000003
Read_Master_Log_Pos: 151
Relay_Log_File: mysql-db02-relay-bin.000002
Relay_Log_Pos: 361
Relay_Master_Log_File: mysql-bin.000003
Slave_IO_Running: Yes #此项yes代表成功
Slave_SQL_Running: Yes #此项yes代表成功
Replicate_Do_DB:
Replicate_Ignore_DB:
Replicate_Do_Table:
Replicate_Ignore_Table:
#....以下省略若干行....
两个从库mysql-db02和mysql-db03都执行以上步骤。
4.7 什么是GTID
- GTID(Global Transaction)全局事务标识符:是一个唯一的标识符,它创建并与源服务器(主)上提交的每个事务相关联。此标识符不仅对其发起的服务器是唯一的,而且在给定复制设置中的所有服务器上都是唯一的。所有交易和所有GTID之间都有1对1的映射。
- GTID实际上是由UUID+TID组成的。其中UUID是一个MySQL实例的唯一标识。TID代表了该实例上已经提交的事务数量,并且随着事务提交单调递增。
- 下面是一个GTID的具体形式:
3E11FA47-71CA-11E1-9E33-C80AA9429562:23
4.8 GTID的新特性
(1)支持多线程复制:事实上是针对每个database开启相应的独立线程,即每个库有一个单独的(sql thread)
(2)支持启用GTID,在配置主从复制,传统的方式里,你需要找到binlog和POS点,然后change master to 指向。在mysql5.6里,无须再知道binlog和POS点,只需要知道master的IP/端口/账号密码即可,因为同步复制是自动的,MySQL通过内部机制GTID自动找点同步。
(3)基于Row复制只保存改变的列,大大节省磁盘空间,网络,内存等
(4)支持把Master和Slave的相关信息记录在Table中;原来是记录在文件里,现在则记录在表里,增强可用性
(5)支持延迟复制
4.9 开启方法
#mysql配置文件:
[mysqld]
gtid_mode=ON
enforce_gtid_consistency
#查看
show global variables like ‘%gtid%’;
4.10 从库设置(mysql-db02,mysql-db03)
#登陆从库
[root@mysql-db02 ~]# mysql -uroot -p123123
#临时禁用自动删除relay log功能
mysql> set global relay_log_purge = 0;
Query OK, 0 rows affected (0.00 sec)
#设置只读
mysql> set global read_only=1;
Query OK, 0 rows affected (0.00 sec)
编辑配置文件/etc/my.cnf
修改完配置文件,别忘了重启动mysql服务
root@mysql-db02 ~]# /etc/init.d/mysqld restart
Shutting down MySQL.. SUCCESS!
Starting MySQL. SUCCESS!
root@mysql-db03 ~]# /etc/init.d/mysqld restart
Shutting down MySQL.. SUCCESS!
Starting MySQL. SUCCESS!
五,部署MHA
5.1 环境准备(所有节点mysql-db01,mysql-db02,mysql-db03)
#光盘安装依赖包
[root@mysql-db01 ~]# yum -y install perl-DBD-MySQL
#安装mha4mysql-node-0.56-0.el6.noarch.rpm
[root@mysql-db01 rpm]# rpm -ivh mha4mysql-node-0.56-0.el6.noarch.rpm
Preparing... ########################################### [100%]
1:mha4mysql-node ########################################### [100%]
[root@mysql-db01 ~]# mysql -uroot -p123123
mysql> grant all privileges on *.* to mha@'192.168.0.%' identified by '123123';
Query OK, 0 rows affected (0.00 sec)
mysql> select user,host from mysql.user where user='mha';
+------+-------------+
| user | host |
+------+-------------+
| mha | 192.168.0.% | #主库上创建从库会自动复制
+------+-------------+
1 row in set (0.00 sec)
#特别提示:3台MySQL都需要安装mha4mysql-node-0.56-0.el6.noarch.rpm
5.2 部署管理节点(mha-manager)
5.2.1 在mysql-db03上部署管理节点
#使用阿里云源+epel源
wget -O /etc/yum.repos.d/CentOS-Base.repo http://mirrors.aliyun.com/repo/Centos-6.repo
wget -O /etc/yum.repos.d/epel-6.repo http://mirrors.aliyun.com/repo/epel-6.repo
#安装manager依赖包(需要公网源)
[root@mysql-db03 ~]# yum -y install perl-Config-Tiny epel-release perl-Log-Dispatch perl-Parallel-ForkManager perl-Time-HiRes
#安装manager包
[root@mysql-db03 rpm]# rpm -ivh mha4mysql-manager-0.56-0.el6.noarch.rpm
Preparing... ########################################### [100%]
1:mha4mysql-manager ########################################### [100%]
5.2.2 编辑配置文件
#创建配置文件目录
[root@mysql-db03 ~]# mkdir -p /etc/mha
#创建日志目录
[root@mysql-db03 ~]# mkdir -p /var/log/mha/mha1
#创建配置文件(默认没有)
[root@mysql-db03 ~]# cd /etc/mha/
[root@mysql-db03 mha]# ls
[root@mysql-db03 mha]# vim /etc/mha/mha1.cnf
[root@mysql-db03 mha]# cat /etc/mha/mha1.cnf
[server default]
manager_log=/var/log/mha/mha1/manager #manager管理日志存放路径
manager_workdir=/var/log/mha/mha1 #manager管理日志的目录路径
master_binlog_dir=/usr/local/mysql/data #binlog日志的存放路径
user=mha #管理账户
password=123123 #管理账户密码
ping_interval=2 #存活检查的间隔时间
repl_user=rep #主从复制的授权账户
repl_password=123123 #主从复制的授权账户密码
ssh_user=root #用于ssh连接的账户
[server1]
hostname=192.168.0.51
port=3306
[server2]
#candidate_master=1 #此条暂时注释掉(后面解释)
#check_repl_delay=0 #此条暂时注释掉(后面解释)
hostname=192.168.0.52
port=3306
[server3]
hostname=192.168.0.53
port=3306
#**特别提示:**
#以上配置文件内容里每行的最后不要留有空格,因此,不能复制的呦
特别说明:
参数:candidate_master=1
解释:设置为候选master,如果设置该参数以后,发生主从切换以后会将此从库提升为主库,即使这个主库不是集群中事件最新的slave
参数:check_repl_delay=0
解释:默认情况下如果一个slave落后master 100M的relay logs 的话,MHA将不会选择该slave作为一个新的master,因为对于这个slave的恢复需要花费很长时间,通过设置check_repl_delay=0,MHA触发切换在选择一个新的master的时候将会忽略复制延时,这个参数对于设置了candidate_master=1的主机非常有用,因为这个候选主在切换的过程中一定是新的master
5.3 配置ssh信任(所有节点mysql-db01,mysql-db02,mysql-db03)
#创建密钥对
[root@mysql-db03 ~]# ssh-keygen -t dsa -P "" -f ~/.ssh/id_dsa >/dev/null 2>&1
#发送mysql-db03公钥,包括自己
[root@mysql-db03 ~]# ssh-copy-id -i /root/.ssh/id_dsa.pub root@192.168.0.51
[root@mysql-db03 ~]# ssh-copy-id -i /root/.ssh/id_dsa.pub root@192.168.0.52
[root@mysql-db03 ~]# ssh-copy-id -i /root/.ssh/id_dsa.pub root@192.168.0.53
#发送mysql-db02公钥,包括自己
[root@mysql-db02 ~]# ssh-copy-id -i /root/.ssh/id_dsa.pub root@192.168.0.51
[root@mysql-db02 ~]# ssh-copy-id -i /root/.ssh/id_dsa.pub root@192.168.0.52
[root@mysql-db02 ~]# ssh-copy-id -i /root/.ssh/id_dsa.pub root@192.168.0.53
#发送mysql-db01公钥,包括自己
[root@mysql-db01 ~]# ssh-copy-id -i /root/.ssh/id_dsa.pub root@192.168.0.51
[root@mysql-db01 ~]# ssh-copy-id -i /root/.ssh/id_dsa.pub root@192.168.0.52
[root@mysql-db01 ~]# ssh-copy-id -i /root/.ssh/id_dsa.pub root@192.168.0.53
5.4 启动测试
5.4.1 ssh检查检测
[root@mysql-db03 ~]# masterha_check_ssh --conf=/etc/mha/mha1.cnf #ssh检查命令
Tue Sep 5 03:01:38 2017 - [warning] Global configuration file /etc/masterha_default.cnf not found. Skipping.
Tue Sep 5 03:01:38 2017 - [info] Reading application default configuration from /etc/mha/mha1.cnf..
Tue Sep 5 03:01:38 2017 - [info] Reading server configuration from /etc/mha/mha1.cnf..
Tue Sep 5 03:01:38 2017 - [info] Starting SSH connection tests..
..中间省略若干行..
Tue Sep 5 03:01:40 2017 - [debug] Connecting via SSH from root@192.168.0.53(192.168.0.53:22) to root@192.168.0.52(192.168.0.52:22)..
Tue Sep 5 03:01:40 2017 - [debug] ok.
Tue Sep 5 03:01:40 2017 - [info] All SSH connection tests passed successfully. #出现这个就表示成功
5.4.2 主从复制检测
(1)错误的主从复制检测 [root@mysql-db03 ~]# masterha_check_repl --conf=/etc/mha/mha1.cnf
如果不出意外,检测结果会是下面的样子
因此在mysql-db02和mysql-db03上添加主从复制的用户即可。
grant replication slave on *.* to rep@'192.168.0.%' identified by '123123';
再次检查如下图所示:
5.5 启动MHA
#启动
[root@mysql-db03 ~]# nohup masterha_manager --conf=/etc/mha/mha1.cnf --remove_dead_master_conf --ignore_last_failover < /dev/null >/var/log/mha/mha1/manager.log 2>&1 &
[root@mysql-db03 ~]# ps -ef | grep perl | grep -v grep
root 4961 4690 0 06:33 pts/2 00:00:00 perl /usr/bin/masterha_manager --conf=/etc/mha/mha1.cnf --remove_dead_master_conf --ignore_last_failover
#说明:
nohup:启动命令
--conf:指定配置文件位置
--remove_dead_master_conf:如果有master down了,就去掉配置文件里该master的部分。
5.6 进行mha自动切换master的测试
初始状态:
(1)登陆mysql-db02(192.168.0.53)查看信息状态
#登陆数据库mysql-db02(192.168.0.53)
[root@mysql-db03 ~]# mysql -uroot -p123123
mysql> show slave statusG
*************************** 1. row ***************************
Slave_IO_State: Waiting for master to send event
Master_Host: 192.168.0.51 #这是主库IP地址
Master_User: rep
Master_Port: 3306
Connect_Retry: 60
Master_Log_File: mysql-bin.000003
Read_Master_Log_Pos: 1656
Relay_Log_File: mysql-db02-relay-bin.000004
Relay_Log_Pos: 1796
Relay_Master_Log_File: mysql-bin.000003
Slave_IO_Running: Yes
Slave_SQL_Running: Yes
..以下省略若干内容..
(2)停掉mysql-db01(192.168.0.51)上的MySQL服务
[root@mysql-db01 ~]# /etc/init.d/mysqld stop
Shutting down MySQL..... SUCCESS!
(3)查看mysql-db03上的MySQL从库同步状态
[root@mysql-db03 ~]# mysql -uroot -p123123 -e 'show slave statusG'
Warning: Using a password on the command line interface can be insecure.
*************************** 1. row ***************************
Slave_IO_State: Waiting for master to send event
Master_Host: 192.168.0.52 #现在的主库IP
Master_User: rep
Master_Port: 3306
Connect_Retry: 60
Master_Log_File: mysql-bin.000006 #binlog日志
Read_Master_Log_Pos: 777 #binlog日志位置
Relay_Log_File: mysql-db03-relay-bin.000002
Relay_Log_Pos: 408
Relay_Master_Log_File: mysql-bin.000006
Slave_IO_Running: Yes
Slave_SQL_Running: Yes
..以下省略若干内容..
(4)查看mysql-db02上的MySQL,主库同步状态。
(5)查看mysql-db03上的mha进程状态
[root@mysql-db03 ~]# ps -ef | grep perl | grep -v grep #查询发现mha进程已经没了
[root@mysql-db03 ~]#
(6)查看mha配置文件信息
说明:
当作为主库的mysql-db01上的MySQL宕机以后,mha通过检测发现mysql-db01宕机,那么会将binlog日志最全的从库立刻提升为主库,而其他的从库会指向新的主库进行再次同步。
此处需要进行简单的mha日志记录的讲解:/var/log/mha/mha1/manager
5.7 进行mha的故障还原测试
由于mysql-db01的MySQL服务宕机,因此mha将mysql-db02提升为了主库。因此,我们需要将宕机的mysql-db01的MySQL服务启动,然后作为主库mysql-db02的从库。
初始状态:
(1)将故障宕机的mysql-db01的MySQL服务启动并授权进行从同步
[root@mysql-db01 ~]# /etc/init.d/mysqld start
Starting MySQL. SUCCESS!
[root@mysql-db01 ~]# mysql -uroot -p123123
mysql> CHANGE MASTER TO MASTER_HOST='192.168.0.52', MASTER_PORT=3306, MASTER_AUTO_POSITION=1, MASTER_USER='rep', MASTER_PASSWORD='123123';
mysql> start slave;
mysql> show slave statusG #查看从同步状态
(2)将mha配置文件里缺失的部分补全
[root@mysql-db03 ~]# cat /etc/mha/mha1.cnf
[server default]
manager_log=/var/log/mha/mha1/manager
manager_workdir=/var/log/mha/mha1
master_binlog_dir=/usr/local/mysql/data
password=123123
ping_interval=2
repl_password=123123
repl_user=rep
ssh_user=root
user=mha
[server1]
hostname=192.168.0.51
port=3306
[server2]
hostname=12.168.0.52
port=3306
[server3]
hostname=192.168.0.53
port=3306
(3)启动mha进程
[root@mysql-db03 ~]# nohup masterha_manager --conf=/etc/mha/mha1.cnf --remove_dead_master_conf --ignore_last_failover < /dev/null >/var/log/mha/mha1/manager.log 2>&1 &
[root@mysql-db03 ~]# ps -ef | grep perl | grep -v grep
root 5226 4690 0 09:42 pts/2 00:00:00 perl /usr/bin/masterha_manager --conf=/etc/mha/mha1.cnf --remove_dead_master_conf --ignore_last_failover
(4)停掉mysql-db02上的MySQL服务
[root@mysql-db02 ~]# /etc/init.d/mysqld stop
Shutting down MySQL..... SUCCESS!
(5)查看mysql-db03上的主从同步状态:
[root@mysql-db03 ~]# mysql -uroot -p123123 -e 'show slave statusG'
Warning: Using a password on the command line interface can be insecure.
*************************** 1. row ***************************
Slave_IO_State: Waiting for master to send event
Master_Host: 192.168.0.51 #此时的主库IP切换回了mysql-db01
Master_User: rep
Master_Port: 3306
Connect_Retry: 60
Master_Log_File: mysql-bin.000005
Read_Master_Log_Pos: 231
Relay_Log_File: mysql-db03-relay-bin.000002
Relay_Log_Pos: 361
Relay_Master_Log_File: mysql-bin.000005
Slave_IO_Running: Yes
Slave_SQL_Running: Yes
..以下省略若干行..
(6)启动mysql-db02上的MySQL服务
[root@mysql-db02 ~]# /etc/init.d/mysqld start
Starting MySQL. SUCCESS!
[root@mysql-db02 ~]# mysql -uroot -p123123
mysql> CHANGE MASTER TO MASTER_HOST='192.168.0.51', MASTER_PORT=3306, MASTER_AUTO_POSITION=1, MASTER_USER='rep', MASTER_PASSWORD='123123';
mysql> start slave;
mysql> show slave statusG
(7)再次补全mha配置文件后,启动mha进程
[root@mysql-db03 ~]# cat /etc/mha/mha1.cnf
[server default]
manager_log=/var/log/mha/mha1/manager
manager_workdir=/var/log/mha/mha1
master_binlog_dir=/usr/local/mysql/data
password=123123
ping_interval=2
repl_password=123123
repl_user=rep
ssh_user=root
user=mha
[server1]
hostname=192.168.0.51
port=3306
[server2]
hostname=192.168.0.52
port=3306
[server3]
#andidate_master=1
#check_repl_delay=0
hostname=192.168.0.53
port=3306
[root@mysql-db03 ~]# nohup masterha_manager --conf=/etc/mha/mha1.cnf --remove_dead_master_conf --ignore_last_failover < /dev/null >/var/log/mha/mha1/manager.log 2>&1 &
[root@mysql-db03 ~]# ps -ef | grep perl | grep -v grep
root 5226 4690 0 09:42 pts/2 00:00:01 perl /usr/bin/masterha_manager --conf=/etc/mha/mha1.cnf --remove_dead_master_conf --ignore_last_failover
此时的初始状态还原为下图:
mha配置文件内容如下:
附录:源码安装mha的方法
node节点的源码安装方法:
[root@mysql-db01 ~]# yum -y install perl-DBD-MySQL perl-Config-Tiny perl-Params-Validate perl-CPAN perl-devel perl-ExtUtils-CBuilder perl-ExtUtils-MakeMaker
[root@mysql-db01 ~]# tar xf mha4mysql-node-0.56.tar.gz -C /usr/src/
[root@mysql-db01 ~]# cd /usr/src/mha4mysql-node-0.56/
[root@mysql-db01 mha4mysql-node-0.56]# perl Makefile.PL
[root@mysql-db01 mha4mysql-node-0.56]# make && make install
manager节点的源码安装方法:
[root@mysql-db01 ~]# yum -y install perl-DBD-MySQL perl-Config-Tiny perl-Params-Validate perl-CPAN perl-devel perl-ExtUtils-CBuilder perl-ExtUtils-MakeMaker
[root@mysql-db01 ~]# tar xf mha4mysql-manager-0.56.tar.gz -C /usr/src/
[root@mysql-db01 ~]# cd /usr/src/mha4mysql-manager-0.56/
[root@mysql-db01 mha4mysql-manager-0.56]# perl Makefile.PL
[root@mysql-db01 mha4mysql-manager-0.56]# make && make install