在安装之前，我们先来做些规划和准备工作

环境

1
2
3
4
5
6
7
Centos: 6.5
manager 10.20.64.209
node1	10.20.64.202
node2	10.20.64.203
node3	10.20.64.204
node4	10.20.64.210

系统依赖包

1
2
3
4
5
6
perl -MCPAN -e 'install "YAML"'
yum -y install perl-CPAN
yum -y install perl-DBD-MySQL
perl -MCPAN -e "install Config::Tiny"
perl -MCPAN -e "install Log::Dispatch"
perl -MCPAN -e "install Parallel::ForkManager"

部署位置
{version}参数部分为自动化预留，手动的同学可参考例子

Key Value
node basedir /opt/mha-node{version} 例如mha-node056
manager basedir /opt/mha-manager{version} 例如mha-manager056
config dir /opt/mha-manager{version}/etc
manager work dir /opt/mha-manager{version}/apps/{app_name}
manager log dir /opt/mha-manager{version}/logs
script log dir /opt/mha-manager{version}/scripts
remote work dir /opt/mha-node{version}/apps/{app_name}

系统软链
（很多属于个人习惯，可以跳过，很多都是为了自动化做铺垫。如果读者觉得没有必要，可以忽略下文中软链部分：））

Key Value
node basedir link /usr/local/mha-node
manager basedir link /usr/local/mha-manager

我们看了官方文档后知道MHA需要在主机间建立信任关系，而且官方做法是使用root帐号建立的，这样是很不安全的

我们知道mha的配置文件是有ssh_user选项的，如下，这里我创建了mha系统帐号，并在主机间使用mha帐号建立信任关系

1
2
3
[server default]
ssh_user=mha
ssh_port=22

创建mysql和mha帐号

1
2
3
4
5
6
useradd mha -u 512 -G mysql -d /home/mha
passwd mha
输入密码
说明：为了统一，我们将mha用户uid固定为512，并隶属于mysql组
   （由于mha在切换过程中需要访问binlog目录，当然你也可以开放binlog目录的其他用户读权限）

建立mha账号信任关系
创建公钥

1
2
3
4
5
ssh-keygen
同时将自己的公钥也添加到授权文件中
cat /home/mha/.ssh/id_rsa.pub > /home/mha/.ssh/authorized_keys
chmod 600 /home/mha/.ssh/authorized_keys

节点间信任

1
2
3
4
推送本地公钥要N个远程
cat /home/mha/.ssh/id_rsa.pub | ssh {nodeN ip} "cat - >> /home/mha/.ssh/authorized_keys"
ssh {nodeN ip} "chmod 600 /home/mha/.ssh/authorized_keys"

管理机与节点间信任

1
2
3
4
5
将已经生成的node节点公钥拿到本地
mha@manager ssh｛nodeN ip} 'cat /home/mha/.ssh/id_rsa.pub' >> /home/mha/.ssh/authorized_keys
也将自己的公钥推送给node节点
cat /home/mha/.ssh/id_rsa.pub | ssh {nodeN ip} "cat - >> /home/mha/.ssh/authorized_keys"

sshd配置

1
2
3
4
是否允许使用基于 GSSAPI 的用户认证：关闭认证
#GSSAPIAuthentication no 
service sshd reload

创建mha的mysql实例账号
MHA用到以下权限，所以不用赋予ALL（DBA洁癖）

1
GRANT SELECT, INSERT, UPDATE, DELETE, CREATE, DROP, RELOAD, SUPER ON *.* TO 'mha'@'%' IDENTIFIED BY 'xxx';

我们可以将编译安装、创建管理目录、配置全局环境变量封装到RPM包中

详细内容如下

编译安装MHA

Node
node basedir
node basedir link 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
tar zxvf mha4mysql-node-0.56.tar.gz
cd mha4mysql-node-0.56
perl Makefile.PL INSTALL_BASE=/opt/mha-node056
说明：这里为选择了INSTALL_BASE选项，它会把node的lib包集中在base目录/lib中
如果不写，或者使用PREFIX的话，lib会散落在系统lib中，卸载时不利于清理
make && make install
chown -R mha:mysql /opt/mha-node056
ln -s /opt/mha-node056 /usr/local/mha-node
ln -s /opt/mha-node056/lib/perl5/MHA /usr/local/share/perl5/MHA-Node
echo '/usr/local/share/perl5/MHA-Node' >> /etc/ld.so.conf

Node安装完成后会有如下脚本

1
2
3
4
apply_diff_relay_logs
filter_mysqlbinlog
purge_relay_logs
save_binary_logs

Node类库

1
2
3
4
5
6
7
8
9
BinlogHeaderParser.pm
BinlogManager.pm
BinlogPosFinderElp.pm
BinlogPosFinder.pm
BinlogPosFinderXid.pm
BinlogPosFindManager.pm
NodeConst.pm
NodeUtil.pm
SlaveUtil.pm
Manager
manager basedir
manager basedir link 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
tar zxvf mha4mysql-manager-0.56.tar.gz
cd mha4mysql-manager-0.56
perl Makefile.PL INSTALL_BASE=/opt/mha-manager056
make && make install
chown -R mha:mysql /opt/mha-manager056
ln -s /opt/mha-node056 /usr/local/mha-manager
ln -s /opt/mha-node056/lib/perl5/MHA /usr/local/share/perl5/MHA-Manager
echo '/usr/local/share/perl5/MHA-Manager' >> /etc/ld.so.conf

Manager安装完成后会有如下脚本

1
2
3
4
5
6
7
8
9
masterha_stop
masterha_conf_host
masterha_check_status
masterha_check_repl
masterha_secondary_check
masterha_master_switch
masterha_master_monitor
masterha_manager
masterha_check_ssh

Manager类库

1
2
3
4
5
6
7
8
9
10
11
12
13
14
Config.pm
DBHelper.pm
FileStatus.pm
HealthCheck.pm
ManagerAdmin.pm
ManagerAdminWrapper.pm
ManagerConst.pm
ManagerUtil.pm
MasterFailover.pm
MasterMonitor.pm
MasterRotate.pm
ServerManager.pm
Server.pm
SSHCheck.pm

创建管理目录
一定要保证/opt/mha-node056与/opt/mha-manager056目录权限为mha:mysql

Node
remote work dir 
1
2
3
mkdir -p /opt/mha-node056/apps
chown -R mha:mysql /opt/mha-node056
Manager
config dir
manager work dir
manager log dir
script log dir 
1
2
3
mkdir -p /opt/mha-manager056/{etc,apps,logs,scripts}
chown -R mha:mysql /opt/mha-manager056

配置全局环境变量
MHA-Node

1
2
3
4
echo "MHA_NODE_PATH=/usr/local/mha/bin" >> /etc/profile
echo "export PATH=\$PATH:\$MHA_NODE_PATH" >> /etc/profile
source /etc/profile

MHA-Manager

1
2
3
4
echo "MHA_MANAGER_PATH=/usr/local/mha/bin" >> /etc/profile
echo "export PATH=\$PATH:\$MHA_MANAGER_PATH" >> /etc/profile
source /etc/profile

配置文件
manger & node 合并为一个配置文件{app_name}.cnf

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
mha$ vi {app_name}.cnf
# Manager
[server default]
ssh_user=mha
ssh_port=22
user=mha
password=
port=3306
repl_user=repl
repl_password=
log_level=info
# Const
client_bindir=/usr/local/mysql/bin
client_libdir=/usr/local/mysql/lib
# Variable
master_binlog_dir=/opt/mysqldata/{port}/binlog
remote_workdir=/opt/mha-node{version}/apps/{app_name}
manager_workdir=/opt/mha-manager{version}/apps/{app_name}
manager_log=/usr/local/mha-manager/logs/{app_name}.log
# Check
secondary_check_script= masterha_secondary_check -s {node2 ip} -s {node3 ip}
ping_interval=5
# Failover DIY Scripts
# master_ip_failover_script= /scripts/master_ip_failover
这里可以使用你自己的脚本，结合dns或者haproxy完成切换的完整工作
# shutdown_script= scripts/power_manager
# report_script= scripts/send_report
# master_ip_online_change_script= scripts/master_ip_online_change
# Node
[server1]
hostname={node1 ip}
candidate_master=1
node_label={app_name} master
[server2]
hostname={node2 ip}
candidate_master=1
node_label={app_name} standby
[server3]
hostname={node3 ip}
no_master=1
node_label={app_name} slave

用户环境变量
这个属于踩坑后的配置（后续补充上报错内容）
在MHA check的时候会报错，所以需要添加完整的环境变量到到mha用户下

1
2
3
4
5
6
7
8
9
10
mha$
echo $PATH
将环境变量实际内容贴到environment文件中，否则bash失效
echo "PATH=/usr/local/bin:/bin:/usr/bin:/usr/local/sbin:/usr/sbin:/sbin:/usr/local/mysql/bin:/home/mha/bin" > /home/mha/.ssh/environment
echo "PermitUserEnvironment yes" >> /etc/ssh/sshd_config
是否允许sshd处理 ~/.ssh/environment 以及 ~/.ssh/authorized_keys 中的 environment= 选项：允许
service sshd reload

检测MHA
检测SSH通讯

1
mha$ masterha_check_ssh --conf=/usr/local/mha-manager/etc/{app_name}.conf

检测集群同步

1
2
3
mha$ masterha_check_repl --conf=/usr/local/mha-manager/etc/{app_name}.conf
--  /opt/mha-manager056/apps/test masterha_check_repl自动创建

启动MHA

1
mha$ masterha_manager --conf=/usr/local/mha-manager/etc/{app_name}.conf &

MHA Manager在每次切换后，将退出进程。修改配置后，再次检测，启动