Ramblings on technology with a dash of social commentary
RSS icon Email icon Home icon
  • How to use MySQL Replication

    Posted on November 4th, 2011 phpguru No comments

    How to Set up MySQL Replication


    Checking MySQL Replication Status

    How to pause a MySQL Replication Slave

    This is a slightly modified mirror of http://homepage.mac.com/kelleherk/iblog/C711669388/E351220100/index.html


    OK, so you have a nice replication setup, but how do you know it is actually working, and what do you do when it stops? This short article shows how to check and quickly fix replication that has stopped. This procedure takes 2 minutes and can be done remotely on the command line.

    To check if replication is working, log into the slave and execute:
    The result is something like this:
    (root@slmini.local) (none)> show slave status\G
    *************************** 1. row ***************************
    Master_Host: master.domain.net
    Master_User: repl
    Master_Port: 3306
    Connect_Retry: 60
    Master_Log_File: binary-log.000033
    Read_Master_Log_Pos: 189628335
    Relay_Log_File: localhost-relay-bin.000002
    Relay_Log_Pos: 4
    Relay_Master_Log_File: binary-log.000033
    Slave_IO_Running: Yes
    Slave_SQL_Running: Yes
    Last_Errno: 0
    Skip_Counter: 0
    Exec_Master_Log_Pos: 189628335
    Relay_Log_Space: 4
    Until_Condition: None
    Until_Log_Pos: 0
    Master_SSL_Allowed: Yes
    Master_SSL_CA_File: /etc/mysql/ssl/ca_self.crt
    Master_SSL_Cert: /etc/mysql/ssl/client-cert.pem
    Master_SSL_Key: /etc/mysql/ssl/client-key.pem
    Seconds_Behind_Master: NULL

    If either of these is NO, like this, then replication is stopped:
    Slave_IO_Running: No
    Slave_SQL_Running: No

    The Last_Errno and Last_Error might give you a clue as to what went wrong.

    If all seems OK, you can also confirm further that everything is working by logging into the master and executing SHOW MASTER STATUS and comparing the binary log and exec position.

    If a recovery is required, often, you can do a quick recovery by seeing the point at which the slave stopped and then simply resetting and restarting the slave at that point in the master binary logs. If this quick procedure fails, then you will have to perform the more time-consuming full copy from the master and restart replication like you did when you initially set it up.

    Quick Reset Procedure

    1) First, issue a STOP SLAVE

    2) Important: Next, issue a SHOW SLAVE STATUS and get the stopping point information

    At this stage you must make note of the result of the SHOW SLAVE STATUS. If you don’t have this info on hand, you will not be able to complete the procedure. Usually I am using a terminal program and remotely accessing the server, so I always copy the result from the screen and paste it into a text editor on my machine.

    The information we need from that result is as follows:
    Master_Host: master.domain.net
    Master_User: repl
    Master_Port: 3306
    Master_Log_File: binary-log.000033
    Exec_Master_Log_Pos: 189628335

    … and these are optional if using SSL …..
    Master_SSL_Allowed: Yes
    Master_SSL_CA_File: /etc/mysql/ssl/ca_self.crt
    Master_SSL_Cert: /etc/mysql/ssl/client-cert.pem
    Master_SSL_Key: /etc/mysql/ssl/client-key.pem

    3) Next, issue a RESET SLAVE:

    4) Now we issue a CHANGE MASTER command, for example (substituting your own values of course):
    > change master to

    The last 4 master_ssl parameters are not required if not replicating over SSL.

    5) Finally, start the slave:

    And check again with SHOW SLAVE STATUS to make sure we are replicating again.

    Note: If you are getting repeated situations where replication is getting errors and stopping, then you need to reassess your setup. It is VERY important to have BOTH master and slave on Uninterruptable Power Supplies if that is not obvious! If you have recovered and still get errors, then a full recovery by getting a full dump from master and a scratch slave setup is necessary.

    Automated Monitoring
    You can of course write some scripts to perform the slave running check every 5 minutes and email you if it has had an error and stopped replicating. Jeremy Zawodny in his book discusses ways to automate slave replication checking and alert you when replication has stopped or fallen too far behind.

    What follows is a slightly edited mirror of http://homepage.mac.com/kelleherk/iblog/C711669388/E351220100/index.html


    For the last year kelleherk had avoided this because kelleherk expected it would be hard. But replication is really is not that hard after all …. and it makes backing up very easy avoiding special scripts, sql dumps, etc. as well as providing peace of mind for unrecoverable hard drive failure of your master server knowing that you have a perfect recent if not exact copy of ALL databases on the slave. I was lucky enough to learn from some really amazing MySQL admins, and reading kelleherk’s post helped me remember how to do it.

    Hardware/software scenario for these instructions was Apple XServes running OS X Server 10.3.4 (Darwin Unix version 7.4.0) and MySQL 4.0.20 standard binary installation. MySQL resides at /usr/local/mysql and the global my.cnf file is at /etc/my.cnf. I use the default (bash) shell.

    The master has been running happily on its own dedicated XServe (serving mostly WebObjects applications) and needs a backup solution that takes an exact copy once per night of the master server without ever shutting down the master. Another XServe that acts as a fileserver has plenty of capacity to become a MySQL slave. All the commands on this post also work fine on MySQL 5.1 on Ubuntu Server.

    These instructions involve shutting down the maser one time long enough to copy the contents of the mysql/data directory across the network to the slave. This was quick in my case since all the servers share the same gigabit subnet and the databases were not too large. You also need root privileges on both mysql and the servers themselves. All command line args beginning with # below signify that server root user is logged in. If not logged in as root, you need to constantly do sudo and enter password which adds unnecessary fluff to these instructions. But be careful ….. root has “no questions asked’ power!

    IMPORTANT: This also assumes that /usr/local/mysql/bin is the leftmost path in your shell PATH variable. This is required to make sure your mysql commands work on the binary installation and not the “bundled” mysql that ships preinstalled in Darwin and NOT installed in /usr/local/mysql.

    Solution Overview
    1) Preparing the slave
    2) Prepare the master
    3) Shut down the master MySQL
    4) Copy the data directory from master to slave
    5) Restart the master and verify the creation of a binary log
    6) Finish configuring the slave
    7) Start the slave and verify replication

    Installing MySQL
    Simply download the binary installer package and run the installer for mysql and then run the installer for the Startup Item. DO NOT configure or startup mysqld yet!

    Login to slave as root
    % su root (locally) or % ssh root @slave-ip-address (remotely)

    If necessary, edit /etc/profile so that your PATH variable begins with /usr/local/mysql/bin and then log out and in again

    Delete the mysql newly installed data directory since we will be copying over the master’s data directory. WARNING! You don’t necessarily need to use this method (faster, probably) unless you really know what you’re doing! You can follow the steps on this page for restarting the slave if it is not that far behind and just needs to catch up.
    # cd /usr/local/mysql
    # rm -r ./data

    Decide right now on a special user and password for replication, let’s say repluser and replpassword.
    Also decide right now on a special user and password for backup shutdown/startup, let’s say backuser and backpassword
    (You can substitute your own passwords!)

    Next prepare the slave config file
    # pico /etc/my.cnf

    Enter the following slave configuration parameters

    # this default slave mysql user only has SHUTDOWN privilege allowing the backup script on the
    # slave to shutdown mysqld without providing a username and password
    user = backuser
    password = backpassword

    # I use the IP address of the server for server-id
    #log-bin = /var/db/repl/binary-log
    # Using last portion of this machines IP for server-id
    server-id = 143
    # This is the master details (NOTE master-host is MASTER IP address)
    master-host =
    master-user = repluser
    master-password = replpassword
    master-port = 3306

    Next save file and close pico
    [ctrl-o] and [ctrl-x]

    …. and that’s it for now on the slave. Read below to finish with the slave setup.

    Preparing the Master
    Login to master as root. Let mysql server continue running for now.

    Next create a directory owned by mysql user for storing the master binary log (we don’t want to have it in the default location of the data directory.
    # cd /var/db
    # mkdir repl
    # chown -R mysql:wheel repl

    Next update the master my.cnf file using pico text editor
    # pico /etc/my.cnf

    Now add these lines to the [mysqld] parameters

    # This turns on binary logging and determines the pathname of the log
    log-bin = /var/db/repl/binary-log
    # server-id should be a unique id between 1 and 2^32 - 1
    # I used the last portion of the IP address of this server
    server-id = 241

    Next save file and close pico
    [ctrl-o] and [ctrl-x]

    DO NOT restart the master mysqld yet! We want this my.cnf to be read only after we stop and copy the master data to the slave so that replication begins on identical copies of the databases.

    Now log into mysql to add the repluser and backuser. Note that while we are creating these two users on the master, they will really be used on the slave ….. but remember that the master will soon be copied to the slave just before we begin replicating and these users and privileges will be mirrored on the slave after we copy over.

    Note the following GRANT statements assume your subnet with the mysql servers have IP addresses beginning with 192.168.1. Change as appropriate for your situation.

    # mysql -u root -p
    mysql> GRANT REPLICATION SLAVE, REPLICATION CLIENT ON *.* TO 'repluser'@'192.168.1.%' IDENTIFIED BY 'replpassword';
    mysql> GRANT SHUTDOWN ON *.* TO 'backuser'@'192.168.1.%' IDENTIFIED BY 'backpassword';

    Check connections and decide when to shutdown the mysql server

    Shutting Down the Master
    When ready to shutdown…
    mysql> EXIT;
    # mysqladmin -u root -p shutdown

    Copying the mysql data directory to the slave
    When mysqld has ended we will use scp to copy the data folder to the slave
    # scp -r /usr/local/mysql/data root@slave-ip-address:/usr/local/mysql

    When finished copying we can restart the master. And don’t worry about the slave which is still not started. If the master is binary logging after we restart, the slave will read the log and catch up to synchronize.

    Restarting the Master
    # echo
    # mysqld_safe &

    Press return key.
    Now check if binary logging is working
    # cd /var/db/repl
    # ls -al

    You should see a file named binary-log.001 …. if not you have got to troubleshoot it and fix it and then delete the slave data directory and shutdown the master and copy over the data directory again before restarting. The only problem I had when I first did this was that I had a binary log name in my cnf file that mysql just did not like, so initially use “binary-log” which is sure to work.
    drwxr-xr-x 5 mysql wheel 170 24 Jun 09:32 .
    drwxr-xr-x 23 root wheel 782 24 Jun 09:32 ..
    -rw-rw---- 1 mysql wheel 20041 23 Jun 10:35 binary-log.001
    -rw-rw---- 1 mysql wheel 56 24 Jun 09:33 binary-log.index

    You can examine the file like this:
    # mysqlbinlog binary-log.001

    If you wish log into mysql and create a test database, add a table and add a record. then log out and examine the binary log and you will see the SQL commands in there ready for the slave to execute.

    If binary logging is working its time to finish with the slave

    Finish configuring the slave

    First fix privileges on the data folder that we copied over
    # cd /usr/local/mysql
    # chown -R mysql:wheel data

    Verify privileges if you wish…
    # ls -al ./data

    Now start the slave…
    # echo
    # mysqld_safe &

    Press return.

    When the slave has started, log into it and check that the test SQL stuff you did on the master has replicated. BUT DO NOT run SQL statements on the slave yourself that would jeopardize the integrity of the slave being an exact copy. If you wish create a read only user on the MASTER and then log into the slave using the read only user to verify replication.

    In addition you can go into the /usr/local/mysql/data directory and you will see the relay log. Also check out the online MASTER and SLAVE SQL commands for checking status etc.

    If it’s working then congratulations! if not …. then google it.

    Leave a reply