How to set up a replica for replication in 6 simple steps with Percona XtraBackup¶
Data is, by far, the most valuable part of a system. Having a backup done systematically and available for a rapid recovery in case of failure is admittedly essential to a system. However, it is not common practice because of its costs, infrastructure needed or even the boredom associated to the task. Percona XtraBackup is designed to solve this problem.
You can have almost real-time backups in 6 simple steps by setting up a replication environment with Percona XtraBackup.
All the things you will need¶
Setting up a replica for replication with Percona XtraBackup is a straightforward procedure. In order to keep it simple, here is a list of the things you need to follow the steps without hassles:
Source
A system with a MySQL-based server installed, configured and running. This
system will be called Source, as it is where your data is stored and the one to be replicated. We will assume the following about this system:
-
the MySQL server is able to communicate with others by the standard TCP/IP port;
-
the SSH server is installed and configured;
-
you have a user account in the system with the appropriate permissions;
-
you have a MySQL’s user account with appropriate privileges.
-
server has binlogs enabled and server-id set up to 1.
Replica
Another system, with a MySQL-based server installed on it. We
will refer to this machine as Replica and we will assume the same things
we did about Source, except that the server-id on Replica is 2.
Percona XtraBackup
The backup tool we will use. It should be installed in both computers for convenience.
Note
It is not recommended to mix MySQL variants (Percona Server, MySQL) in your replication setup. This may produce incorrect xtrabackup_slave_info file when adding a new replica.
Version updates¶
The 8.0.23 version deprecates the CHANGE_MASTER_TO command. In that
version and later, use
the CHANGE_REPLICATION_SOURCE_TO and the appropriate options instead.
The 8.0.22 version deprecates theSTART SLAVE command. In that version or
later, use START REPLICA instead.
1. Make a backup on the Source and prepare it¶
At the Source, issue the following to a shell:
$ xtrabackup --backup --user=yourDBuser --password=MaGiCdB1 --target-dir=/path/to/backupdir
After this is finished you should get:
Expected output
xtrabackup: completed OK!
This will make a copy of your MySQL data dir
to the /path/to/backupdir directory.
You have told Percona XtraBackup to connect to the database server
using your database user and password,
and do a hot backup of all your data in it
(all MyISAM, InnoDB tables and indexes in them).
In order for snapshot to be consistent you need to prepare the data on the source:
$ xtrabackup --user=yourDBuser --password=MaGiCdB1 \
--prepare --target-dir=/path/to/backupdir
You need to select path where your snapshot has been taken. If everything is ok you should get the same OK message. Now the transaction logs are applied to the data files, and new ones are created: your data files are ready to be used by the MySQL server.
Percona XtraBackup knows where your data is by reading your my.cnf file.
If you have your configuration file in a non-standard place,
you should use the flag --defaults-file =/location/of/my.cnf.
If you want to skip writing the username and password
every time you want to access MySQL,
you can set it up in .mylogin.cnf as follows:
mysql_config_editor set --login-path=client --host=localhost --user=root --password
For more information, see MySQL Configuration Utility.
This statement provides root access to MySQL.
2. Copy backed up data to the Replica¶
On the Source, use rsync or scp to copy the data from the Source to the
Replica. If you are syncing the data directly to replica’s data directory,
we recommend that you stop the mysqld there.
$ rsync -avpP -e ssh /path/to/backupdir Replica:/path/to/mysql/
After data has been copied, you can back up the original or previously installed MySQL datadir (NOTE: Make sure mysqld is shut down before you move the contents of its datadir, or move the snapshot into its datadir.). Run the following commands on the Replica:
$ mv /path/to/mysql/datadir /path/to/mysql/datadir_bak
and move the snapshot from the Source in its place:
$ xtrabackup --move-back --target-dir=/path/to/mysql/backupdir
After you copy data over, make sure the Replica MySQL has the proper permissions to access them.
$ chown mysql:mysql /path/to/mysql/datadir
If the ibdata and iblog files are located in directories outside the datadir, you must put these files in their proper place after the logs have been applied.
3. Configure the Source’s MySQL server¶
On the source, run the following command to add the appropriate grant. This grant allows the replica to be able to connect to source:
mysql> GRANT REPLICATION SLAVE ON *.* TO 'repl'@'$replicaip'
IDENTIFIED BY '$replicapass';
Also make sure that firewall rules are correct and that the Replica can
connect to the Source. Run the following command on the Replica to test
that you can run the mysql client on Replica, connect to the Source,
and authenticate.
mysql> mysql --host=Source --user=repl --password=$replicapass
Verify the privileges.
mysql> SHOW GRANTS;
4. Configure the Replica’s MySQL server¶
Copy the my.cnf file from the Source to the Replica:
$ scp user@Source:/etc/mysql/my.cnf /etc/mysql/my.cnf
and change the following options in /etc/mysql/my.cnf:
server-id=2
and start/restart mysqld on the Replica.
In case you’re using init script on Debian-based system to start mysqld, be
sure that the password for debian-sys-maint user has been updated, and
it’s the same as that user’s password on the Source. Password can be seen
and updated in /etc/mysql/debian.cnf.
5. Start the replication¶
On the Replica, review the content of the file xtrabackup_binlog_info,
it will be something like:
$ cat /var/lib/mysql/xtrabackup_binlog_info
The results should resemble the following:
Expected output
Source-bin.000001 481
The term master is deprecated. Do the following on a
MySQL console and use the username and
password you’ve set up in STEP 3 :
-
Version 8.0.23 or later, use the
CHANGE_REPLICATION_SOURCE_TOstatement -
Before 8.0.23, use the
CHANGE MASTERstatement
CHANGE REPLICATION SOURCE TO
SOURCE_HOST='$sourceip',
SOURCE_USER='repl',
SOURCE_PASSWORD='$replicapass',
SOURCE_LOG_FILE='Source-bin.000001',
SOURCE_LOG_POS=481;
Start the replica:
START REPLICA;
The term slave is deprecated. Do the following:
- Version 8.0.22 or later, use
START REPLICA - Before version 8.0.22, use
START SLAVE
6. Check¶
On the Replica, check that everything went OK with:
SHOW REPLICA STATUS\G
The result shows the status:
Expected output
Slave_IO_Running: Yes
Slave_SQL_Running: Yes
Seconds_Behind_Master: 13
Both IO and SQL threads need to be running. The Seconds_Behind_Master
means the SQL currently being executed has a current_timestamp of 13
seconds ago. It is an estimation of the lag between the Source and
the Replica. Note that at the beginning, a high value could be shown
because the Replica has to “catch up” with the Source.
Adding more replicas to the Source¶
You can use this procedure with slight variation to add new replicas to a
source. We will use Percona XtraBackup to clone an already configured
replica. We will continue using the previous scenario for convenience, but
we will add a NewReplica to the plot.
At the Replica, do a full backup:
$ xtrabackup --user=yourDBuser --password=MaGiCiGaM \
--backup --slave-info --target-dir=/path/to/backupdir
By using the --slave-info Percona XtraBackup creates additional file
called xtrabackup_slave_info.
Apply the logs:
$ xtrabackup --prepare --use-memory=2G --target-dir=/path/to/backupdir/
Note
In the prepare phase, the --use-memory parameter speeds up the process if the amount of RAM assigned to the option is available. Use the parameter only in the prepare phase. In the other phases the parameter makes the application lazy allocate this memory (reserve) but does not affect database pages.
Copy the directory from the Replica to the NewReplica (NOTE: Make
sure mysqld is shut down on the NewReplica before you copy the contents
the snapshot into its datadir.):
rsync -avprP -e ssh /path/to/backupdir NewReplica:/path/to/mysql/datadir
For example, to set up a new user, user2, you add another grant on
the Source:
> GRANT REPLICATION SLAVE ON *.* TO 'user2'@'$newreplicaip'
IDENTIFIED BY '$replicapass';
On the NewReplica, copy the configuration file from the Replica:
$ scp user@Replica:/etc/mysql/my.cnf /etc/mysql/my.cnf
Make sure you change the server-id variable in /etc/mysql/my.cnf to 3 and
disable the replication on start:
skip-slave-start
server-id=3
After setting server_id, start mysqld.
Fetch the master_log_file and master_log_pos from the
file xtrabackup_slave_info, execute the statement for setting up the
source and the log file for the NewReplica:
> CHANGE MASTER TO
MASTER_HOST='$Sourceip',
MASTER_USER='repl',
MASTER_PASSWORD='$replicapass',
MASTER_LOG_FILE='Source-bin.000001',
MASTER_LOG_POS=481;
The term master is deprecated. Do the following
and then start the replica:
-
Version 8.0.23 or later, use the
CHANGE_REPLICATION_SOURCE_TOstatement -
Before 8.0.23, use the
CHANGE MASTERstatement
The term slave is deprecated. Do the following:
- Version 8.0.22 or later, use
START REPLICA. - Before version 8.0.22, use
START SLAVE
> START REPLICA;
If both IO and SQL threads are running when you check the NewReplica,
server is replicating the Source.
