[teacher Zhao Yuqiang] MySQL high availability architecture: MHA

MHA（Master HA） It's an open source MySQL High availability program for , It's for MySQL The master-slave replication architecture provides automating master failover function .MHA It's monitoring master Node failure , It's going to improve the ones with the latest data slave Nodes become new master node , in the meantime ,MHA Consistency issues can be avoided by getting additional information from other nodes .MHA It also provides master Online switching function of nodes , Switch on demand master/slave node .
MHA It was by the Japanese yoshinorim（ I used to work in DeNA Now I work in FaceBook） The development is relatively mature MySQL High availability solution .MHA In the 30 Fail over in seconds , And it can be used in the process of fail over , Ensure data consistency as much as possible . At present, Taobao is also developing similar products TMHA, At present, it supports one master and one subordinate .

One 、MHA The composition of

（ One ）MHA The role of

MHA There are two roles of service , MHA Manager( The management node ) and MHA Node( Data nodes )：

• MHA Manager： Usually deployed separately on a single machine to manage multiple master/slave colony ( Group ), Every master/slave A cluster is called a application, It is used to manage and coordinate the whole cluster .
• MHA Node： Run on each MySQL Server (master/slave/manager), It has parsing and cleaning through monitoring logs Feature scripts to speed failover . It is mainly the agent that receives instructions from the management node , Agents need to run on every mysql Node . In short node It is used to collect the generated data from the node server bin-log. Compare whether the slave node that intends to be promoted to the new master node owns and completes the operation , If it is not sent to the new master node, it will be promoted to the master node after local application .

（ Two ）MHA Tools provided

MHA There are many tools available , The common ones are as follows ：

• Manager node

• Node node （ These tools are usually made of MHA Manager The script triggers , There is no need for human operation ）

Two 、MHA How it works

• From the crash of master Save binary log events （binlog events）;
• Identify those with the latest updates slave ;
• Apply differential relay logs (relay log) To other slave ;
• Application from master Saved binary log events (binlog events);
• Upgrade one slave As new master ;
• Use other slave Connect new master replicate .

3、 ... and 、 Deploy MHA

（ One ） Prepare the experimental environment

• Three hosts ：mysql111、mysql112、mysql113
• operating system ：CentOS
• Configure host name （/etc/hosts）
• Configure password free login
• Each host is installed with MySQL

（ Two ） Build master-slave environment

• Each station MySQL Turn on binlog, Set up server-id

vi /etc/my.cnf
log-bin=mysql-binlog
server-id=1           （ Be careful ： Each station MySQL Set up different server-id）

show variables like '%log_bin%';

• On all machines , Create a master-slave copy account

create user 'repl'@'192.168.79.%' identified by 'Welcome_1';
grant replication slave on *.* to 'repl'@'192.168.79.%';
flush privileges;

• On all machines , Create an administrative account

create user 'myadmin'@'192.168.79.%' identified by 'Welcome_1';
grant all privileges on *.* to 'myadmin'@'192.168.79.%';
flush privileges;

• On the master and slave Libraries , Enable GTID

set @@GLOBAL.ENFORCE_GTID_CONSISTENCY = ON;
set @@GLOBAL.GTID_MODE = OFF_PERMISSIVE;
set @@GLOBAL.GTID_MODE = ON_PERMISSIVE;
set @@GLOBAL.GTID_MODE = ON;

Parameter description ：
GTID yes MySQL 5.6 New features , Its full name is Global Transaction Identifier, Can be simplified MySQL Master slave switching and Failover.GTID Used in binlog Uniquely identifies a transaction in . When the transaction commits ,MySQL Server Writing binlog When , Will write a special Binlog Event, The type is GTID_Event, Specifies the next transaction GTID, And then write the transaction Binlog. Master slave synchronization GTID_Event And business Binlog Will be passed to the slave Library , The same is used in the execution of the slave library GTID Write binlog, After the master-slave synchronization , Can be through GTID Confirm the location of synchronization from the library to . in other words , Whether it's cascading , It's still one master and many subordinates , Both can pass GTID Find something automatically , You don't have to go through it like before File_name and File_position I have some .

• Configure the master-slave copy command on the slave database and turn on the master-slave synchronization

change master to master_host='mysql111',master_user='repl',\
master_password='Welcome_1',master_auto_position=1;

start slave;

• Check the status of master-slave replication on the slave library

show slave status\G;

• Test master-slave replication
  

（ 3、 ... and ） install MHA

• All nodes are installed node Software dependency packages and node Software

yum -y install perl-DBD-MySQL 
rpm -ivh mha4mysql-node-0.58-0.el7.centos.noarch.rpm

• stay mysql111 Install on manager Software dependency packages

yum install -y perl-Config-Tiny 
yum install -y epel-release 
yum install -y perl-Log-Dispatch 
yum install -y perl-Parallel-ForkManager 
yum install -y perl-Time-HiRes

• stay mysql111 Install on manager Software

rpm -ivh mha4mysql-manager-0.58-0.el7.centos.noarch.rpm

（ Four ） establish Manager Configuration file for

# Create a directory of configuration files 、 Log directory 
mkdir -p /etc/mha
mkdir -p /var/log/mha/log

# edit mha The configuration file  vi /etc/mha/mha.cnf
[server default]
manager_log=/var/log/mha/log/manager
manager_workdir=/var/log/mha/log
master_binlog_dir=/var/lib/mysql
user=myadmin
password=Welcome_1
ping_interval=2
repl_user=repl
repl_password=Welcome_1
ssh_user=root
[server1]
hostname=mysql111
port=3306
[server2]
hostname=mysql112
port=3306
[server3]
hostname=mysql113
port=3306

（ 5、 ... and ） Check status , And open MHA

• Check mutual trust

masterha_check_ssh --conf=/etc/mha/mha.cnf

• Check master-slave replication status

masterha_check_repl --conf=/etc/mha/mha.cnf

• Turn on MHA-manager

nohup masterha_manager --conf=/etc/mha/mha.cnf > /var/log/mha/log/manager.log < /dev/null 2>&1 &

• see MHA state

masterha_check_status --conf=/etc/mha/mha.cnf

• test HA Highly available automatic switching

 stay mysql111 On , perform shutdown operation 
mysqladmin -uroot -pWelcome_1 shutdown

MHA It will automatically switch between master and slave . After switching ,MHA The process will automatically stop running .

 stay mysql112 and mysql113 Upper observation , Execute the following statement ：
show slave status\G;