Database mysql master-slave scheme

Let's talk about the concept of dual hot standby , It is to keep the state of two databases synchronized automatically . Operations on any database are automatically applied to another database , Always keep the data in both databases consistent . There are many advantages to this . 1. We can do disaster recovery , If one of them is broken, you can switch to the other . 2. Load balancing can be done , The request can be allocated to any one of them , Improve website throughput .  For remote hot standby , Especially suitable for disaster recovery . No more nonsense . We go straight to the subject . We will mainly introduce two parts ：

One , MySQL  How backup works

Two , Back up the actual combat

 

We began to .

I'm using mysql 5.5.34,

 

 

One , mysql How backup works

To put it simply is to put Executed on a server sql The statement is also repeated on other servers , In this way, as long as the initial state of the two databases is the same , Then they can be synchronized all the time .

 

Of course, this duplication and repetition are mysql Automatically , We just need to configure it .

 

Let's go into more detail about the principle , Here is a picture ：

There are two servers in the figure above , Demonstrated from a master server （master） Synchronize the data to the slave server （slave） The process of .

This is a master - From the copied example . Lord - Master copy each other, just do the above example in reverse again . So we introduce the principle with this example .

 

For one mysql The server , There are usually two threads responsible for replication and being replicated . When replication is turned on .

 

1. As the primary server Master,  Will record every change to Binary log Binarylog in . （ The slave server will be responsible for reading this log, Then do it again at your own place .）

 

2. As a slave Slave, Will use master Log in to master On , Read master Of Binarylog,  Write to your own relay log Relaylog, And then my own sql The thread will be responsible for reading the relay log , And do it again .  Here, the changes on the master server are synchronized to the slave server .

 

stay mysql You can view the master of the current server on , From the State . In fact, it is the current server Binary（ As the master server ） Status and location . And its RelayLog（ As a slave ） Replication progress of .

 

For example, we view the master status on the master server ：

 

mysql> show master status\G
*************************** 1. row ***************************
            File: mysql-bin.000014
        Position: 107
    Binlog_Do_DB: 
Binlog_Ignore_DB: mysql,information_schema,performance_schema,amh
1 row in set (0.00 sec)

Explain the meaning of these lines a little ：

1. The first line indicates Currently recording binarylog The file name is ： mysql-bin.000014.

We can do it in mysql Data directory , Find this file ：

 

2.  The second line , 107. Indicates the current file offset , Is written in mysql-bin.000014 Record location of the file .

These two points constitute Status of the primary server .  When configuring slave servers , These two values are required . Tell the slave server where to read data from the master server . （ After logging in from the server , Find this log file , And start copying after this offset .）

 

3. The third line , And the fourth line , Indicates the database to be recorded and the database to be ignored . Only databases that need to be recorded , Its changes will be written into mysql-bin.000014 Log file .  These two parameters will be introduced again later .

 

We can also use the slave server , View the replication status of the slave server .

 

   1:  mysql> show slave status\G

   2:  *************************** 1. row ***************************

   3:                 Slave_IO_State: Waiting for master to send event

   4:                    Master_Host: 198.**.***.***

   5:                    Master_User: r*******

   6:                    Master_Port: 3306

   7:                  Connect_Retry: 60

   8:                Master_Log_File: mysql-bin.000014

   9:            Read_Master_Log_Pos: 107

  10:                 Relay_Log_File: mysqld-relay-bin.000013

  11:                  Relay_Log_Pos: 253

  12:          Relay_Master_Log_File: mysql-bin.000014

  13:               Slave_IO_Running: Yes

  14:              Slave_SQL_Running: Yes

  15:                Replicate_Do_DB: 

  16:            Replicate_Ignore_DB: mysql,information_schema,amh,performance_schema

  17:             Replicate_Do_Table: 

  18:         Replicate_Ignore_Table: 

  19:        Replicate_Wild_Do_Table: 

  20:    Replicate_Wild_Ignore_Table: 

  21:                     Last_Errno: 0

  22:                     Last_Error: 

  23:                   Skip_Counter: 0

  24:            Exec_Master_Log_Pos: 107

  25:                Relay_Log_Space: 556

  26:                Until_Condition: None

  27:                 Until_Log_File: 

  28:                  Until_Log_Pos: 0

  29:             Master_SSL_Allowed: No

 

Let's focus on the part of the red circle on the way ：

1.  Master_host refer to The address of the primary server . 

2. Master_user It refers to the users used for replication on the primary server .  The slave server will use this account to log in to the master service . replicate .

3. Master_log_file That's what I mentioned earlier , Log file name on the primary server .

4. Read_Master_log_pos This is the logging location of the primary server mentioned above , The server selects the copied files and locations according to these two conditions .

5. Slave_IO_Running:  This means that the slave server is responsible for reading the working status of the master server . The slave server uses this special thread to link to the master server , And copy back the log .

6. Slave_SQL_Running: It refers to the special execution of sql The thread of . It is responsible for copying back Relaylog Execute into your own database . Both parameters must be Yes It means that replication is working properly .

 

Other parameters will be introduced later .

 

 

Two , mysql Dual hot standby actual combat

After understanding the above principle , Let's fight . There are two main points , To synchronize the database state , The same initial state is required , Then it makes sense to configure synchronization . Of course, you can avoid the initial state , This is your freedom . Let's configure it from the beginning .

 

 

Let's start with A Server as starting point ,  Configure its database synchronization to B.  This is the Lord - Copied from . Then do it the other way around , You can back up each other .

 

1, First step ,

stay A Created above for backup user ：

 

grant replication slave on *.* to 'repl_user'@'192.***.***.***' identified by 'hj34$%&mnkb';

 

Top handle ip Change the address to B Mechanical ip Address . Only B Sign in . Security .

The user is called ： repl_user

The password for ： hj34$********nkb

This will be in B It's going to use .

 

2. Turn on the main server binarylog.

Many servers are turned on by default , Let's check here ：

open /etc/my.cnf

 

 

Let me explain the configuration in the red box ：

The first three lines , You may already have .

binlog-do-db Used to represent , Only the database changes are recorded in binary In the log . You can write about hello database . But I commented it out . Just to show you . Can write many lines , Show interest in multiple databases .

binlog-ignore-db Express , Which databases need to be ignored . I've ignored the others here 4 A database .

 

The last two are used in Double master （ Multiple main cycles ） Backup each other . Because every database server can insert data into the same table , If the table has an auto growing primary key , Then there will be primary key conflicts on multiple servers .  The solution to this problem is to make the self incrementing primary key of each database discontinuous .  The picture above says yes , I assume that it may be necessary in the future 10 Servers for backup , therefore auto-increment-increment Set to 10.   and auto-increment-offset=1 Indicates the serial number of this server . from 1 Start , No more than auto-increment-increment.

After doing that , The first one I inserted on this server id Namely 1, Second elements id Namely 11 了 , instead of 2.

（ Empathy , The first one inserted on the second server id Namely 2, The second line is 12, This will be introduced later ） In this way, there will be no primary key conflict . We will demonstrate this later id The effect of .

 

3.  Get master server status , And synchronous initial state .

Suppose I now have these databases in A above .

If you're a brand new installation , Then there is no need to synchronize the initial state , Just skip this step , Go to the back and directly check the status of the primary server .

Here we assume that there is a hello Database as initial state .

 

 

Lock first hello database ：

FLUSH TABLES WITH READ LOCK;

 

 

Then export the data ：

I just need to export hello database , If you have multiple databases as the initial state , You need to export all these databases ：

 

 

Then check it out A Server's binary Log location ：

Remember the file name and Location , It will be used on the slave server .

 

 

The main server is done , The lock can be unlocked ：

 

 

 

4.  Set up slave server B The database that needs to be replicated

Open from the server B Of /etc/my.cnf file ：

 

 

 

Explain the above .

server-id Every server must be different . This may be related to cyclic synchronization . Prevent entering the dead loop .

replicate-do-db You can specify the database to be replicated , I've dropped it here . Show me .

replicate-ignore-db Databases that need to be excluded during replication , I use the , This . Apart from a few databases of the system , All databases are replicated .

relay_log Name of the relay log . As mentioned earlier , The replication thread needs to copy the remote changes to the relay log first , In execution .

log-slave-updates intend , After the relay log is executed , Whether these changes need to be included in your own binarylog. When your B The server needs to be opened when it needs to be the primary server of another server .  That is, the two masters backup each other , Or multiple primary and cyclic backups . We need , So open .

 

preservation , restart mysql.

 

 

5. Import the initial state , Start syncing .

Just now from A Exported on the server hello.sql Import to B Of hello In the database , If B No, not at the moment hello database , Please create a , Then import ：

Create database ：

mysql> create database hello default charset utf8;

hold hello.sql Upload to B On , Then import ：

 

If you just exported multiple databases , You need to upload and import them one by one .

 

Turn on synchronization , stay B Execute on the server ：

 CHANGE MASTER TO 
       MASTER_HOST='192.***.***.***', 
       MASTER_USER='repl_user', 
       MASTER_PASSWORD='hj3****', 
       MASTER_LOG_FILE='mysql-bin.000004', 
       MASTER_LOG_POS=7145;

 

I won't explain the above parameters . As I said before .

 

restart mysql,  Then check it out slave Is the thread started ：

 

 

Notice the red box in the figure , All two are Yes, Indicates that it is successfully opened .

Slave_IO_Running: Yes

Slave_SQL_Running: Yes

If one of them is No, That means it's not successful . Need to view mysql Error log for . I encountered this problem the first time I did it . Sometimes the password is wrong , Sometimes the firewall 3306 Not open .ip Wrong address , wait . Will lead to failure .

 

Let's look at the error log : mysql The error log of is usually in ：

 

The file name should be your machine name , My name is host1.err You change your own .

 

Come here, Lord - Copy from has been opened . Let's experiment first .

We are A In the database Add data ：

 

I am here A Of hello Database test In the table Continuous insertion 3 Data ,  Pay attention to their self growth id, Namely 1,11,21.  Do you know why . I've already said that , If you don't understand, go back to see .

 

Let's go and have a look at B Does the database have these three data ：

 

open B The database of ：

I found it here . The effect here is not intuitive .

 

Don't at this time B Modify data in . We then configure from B To A Copy .  If you only need master-slave replication , It's over here . You can stop looking at the back . all A All changes in can be automatically synchronized to B, But yes. B Your changes cannot be synchronized to A. Because it's one-way . If you need two-way synchronization , Need to do it again from B To A Copy .

 

Basically the same as above ： Let's briefly introduce ：

 

1. stay B Create user ;

 

2. open /etc/my.cnf , Turn on B Of binarylog：

Notice the newly added section in the red box .

 

3. We do not need to export B The initial state of , Because it has just come from A Lead in .  Just remember it master Log status ：

 

Remember these two numbers , Wait a minute A It's going to use .

B The server is set up .

 

4. Log in to A The server . Relay on ：

 

Notice the added part in the center of the box , No explanation. .

 

5. Start the synchronization ：

above ip The address is B Of ip Address , because A hold B treat as master 了 . No explanation. .

 

And then restart mysql service .

Then check it out ,slave Whether the state is normal ：

 

There are two in the figure No.

Slave_IO_Running: No

Slave_SQL_Running: No

explain slave No success , namely , from B To A Synchronization of did not succeed . Let's go check mysql Error log , The position mentioned earlier ：

 

 

find   machine name .err file , Open it and see. ：

 

Look at the... In the picture error Information .  Said that the relay log file could not be found .

This is because we are configuring A The name of the relay file was changed when the relay file of , however mysql There is no synchronization . The solution is simple .

 

 

First stop mysql service .  Find these three files , Delete them . Be sure to stop first mysql service . Otherwise, it will not succeed . You need to restart the machine . Or by hand kill mysqld.

Okay , start-up mysql after . We are here to check slave state ：

 

 

Notice the two large Yes.  ha-ha .

 

Slave_IO_Running: Yes

Slave_SQL_Running: Yes

 

Proof from B To A The replication of was also successful .

 

Now let's go to B Try inserting a few pieces of data into the server ：

 

I am here B Two pieces of data are inserted in the . Pay attention to their id.  Don't explain .

 

And then we , Log in A Look at it ,A Has the database changed .

You can see that it has been automatically synchronized to A 了 .

 

thus , AB The introduction of dual master mutual hot standby is over .

Master-slave Master/Slave Maintenance order ：

master End ：
show master status;— Check the status ：
show processlist; – see slave Next MySQL Process information
reset master; # Use with caution , The log will be cleared and synchronized position

slave End ：

CHANGE MASTER TO MASTER_LOG_FILE=’master.000019′;
show slave status;
show slave logs;
show processlist;
reset slave; # Use with caution , Will empty slave Configuration information 、 Logging and synchronization position
Skip error events on the slave server
mysql>stop slave;
mysql>set global sql_slave_skip_counter = n( Skip the next in the primary server n Events . This command is valid for replication termination caused by the statement . Only valid when the slave server thread is not running );
mysql>start slave;

mysql Synchronize users in primary server Must possess  SUPER ,RELOAD,REPLICATION SLAVE  jurisdiction

When a new slave server is added , You need to start with the slave database load data master; Ensure consistency with other slave database data

set global sql_slave_skip_counter=n # Client running , Used to skip several events , Only when the synchronization process stops due to an error can it be executed .

reset master # The host side runs , Clear all logs , This command is the original flush master

reset slave # Slave operation , Clear the log synchronization location flag , And regenerate master.info

Although regenerated master.info, But it doesn't work , best , Will be driven from mysql Restart the process ,

load table tblname from master
# Slave operation , Reread the data of the specified table from the host , Only one... Can be read at a time , suffer timeout The time limit , Need to adjust timeout Time . To execute this command, you need to synchronize the accounts reload and super jurisdiction . And the corresponding library has select jurisdiction . If the watch is bigger , To increase net_read_timeout and net_write_timeout Value

load data from master # Slave execution , Re read all data from the host . To execute this command, you need to synchronize the accounts reload and super jurisdiction . And the corresponding library has select jurisdiction . If the watch is bigger , To increase net_read_timeout and net_write_timeout Value

change master to master_def_list # Change some host settings online , Multiple are separated by commas , such as

change master to

master_host=’master2.mycompany.com’,

master_user=’replication’,

master_password=’bigs3cret’

master_pos_wait() # Slave operation

show master status # Host running , See the log export information

show slave hosts # Host running , Look at the connected slave .

show slave status (slave)

show master logs (master)

show binlog events [ in 'logname' ] [ from pos ] [ limit [offset,] rows ]

purge [master] logs to ‘logname’ ; purge [master] logs before ‘date’

// Show all binary logs on the machine
mysql> SHOW MASTER LOGS;
// Delete all binary logs on this computer
mysql> RESET MASTER;
// Delete all created on binary-log.xxx Previous binary logs
mysql> PURGE MASTER LOGS TO ‘binary-log.xxx’;
// Keep only the recent 6 Day's diary , The previous ones are deleted
find /var/intra -type f -mtime +6 -name “*.log” -exec rm -f {} \;
// Use the upper left corner of the keyboard （ That is to say Esc below ） That key surrounds , Instructions are commands .-1d It was yesterday , And so on -1m Last month, etc
day=`/bin/date -v -1d +%Y%m%d`;
// Rename the file
mv xxx.log xxx-${day}.log;
// Add the user name and password of the database , The function is to update the log （ Including binary log, query log, etc ）
mysqladmin flush-logs

 

show processlist; Check the process
kill proId; Kill the process  

 

 

Reference resources ：

1. mysql-keepalived- Realize the read-write separation of dual master and hot standby

http://gitsea.com/2013/06/16/mysql-keepalived-%E5%AE%9E%E7%8E%B0%E5%8F%8C%E4%B8%BB%E7%83%AD%E5%A4%87%E8%AF%BB%E5%86%99%E5%88%86%E7%A6%BB/

 

2. MySQL Data synchronization 【 Dual master hot standby 】http://www.cnblogs.com/zhongweiv/archive/2013/02/01/mysql_replication_circular.html

 

3. Mysql Dual computer hot standby

http://yunnick.iteye.com/blog/1845301

 

4. High performance Mysql Replication principle and configuration of master slave architecture http://blog.csdn.net/hguisu/article/details/7325124

 

5.   mysql be based on master-master Dual computer hot standby configuration