One way to create a snapshot of the data in an existing master database is to use the mysqldump tool. Once the data dump has been completed, you then import this data into the slave before starting the replication process.
To obtain a snapshot of the data using mysqldump:
If you have not already locked the tables on the server to prevent statements that update data from executing:
Start a session on the server by connecting to it with the
command-line client, and flush all tables and block write
statements by executing the
FLUSH TABLES WITH
READ LOCK statement:
mysql> FLUSH TABLES WITH READ LOCK;
Remember to use SHOW MASTER
STATUS and record the binary log details for use
when starting up the slave. The point in time of your
snapshot and the binary log position must match. See
Section 16.1.1.4, “Obtaining the Replication Master Binary Log Coordinates”.
In another session, use mysqldump to create a dump either of all the databases you want to replicate, or of selected individual databases. For example:
shell> mysqldump --all-databases --lock-all-tables >dbdump.db
An alternative to using a bare dump, is to use the
--master-data option, which automatically
appends the CHANGE MASTER TO
statement required on the slave to start the replication
process.
shell> mysqldump --all-databases --master-data >dbdump.db
In the client where you acquired the read lock, release the lock:
mysql> UNLOCK TABLES;
Alternatively, exit the first session to release the read lock.
When choosing databases to include in the dump, remember that you will need to filter out databases on each slave that you do not want to include in the replication process.
You will need either to copy the dump file to the slave, or to use the file from the master when connecting remotely to the slave to import the data.
© 2013, Oracle Corporation and/or its affiliates
User Comments
When your master is overloaded or is short on disk space, execute mysqldump on the slave, connecting remotely to the master: this saves master from extra disk writes when creating the dump file.
You may also pipe the output to gzip:
mysqldump --user=abc --password=abc | gzip > backup.sql.gz
To execute on the other servers, without uncompressing, use the following command:
zcat backup.sql.gz | mysql --user=abc --password=abc
Note that the --master-data option only includes MASTER_LOG_FILE and MASTER_LOG_POS arguments of the "CHANGE MASTER TO" statement; you'll likely want to edit the dump to include MASTER_HOST, MASTER_USER and MASTER_PASSWORD values when importing to a fresh slave server (or realize that you'll need to re-issue a CHANGE MASTER TO statement after the import).
The mysqldump process locked up for me if I first did "FLUSH TABLES WITH READ LOCK;" as described above. I later discovered this would only happen for databases that contained an InnoDB table.
If you are having this issue you can fix it my adding the --single-transaction argument to your mysqldump call.
There is an interesting script sample here: http://www.docplanet.org/linux/backing-up-linux-web-server-live-via-ssh/
showing a way to dump mysql databases directly into gzip and then into ssh connection, thus creating a gzipped dump archive that never resided on the server hard drive. This can be a handy way to ensure that backup does not fill up the server hard drive.
"FLUSH TABLES WITH READ LOCK;" seems to be unnecessary if you run mysqldump --master-data. Since it turns on --lock-all-tables automatically, it locks all tables across all databases. This is achieved by acquiring a global read lock for the duration of the whole dump.
So the lock doesn't need to be acquired manually.
@Gianluigi Yes but the point of the FLUSH TABLES WITH READ LOCK here is that you acquire it, get the master log coordinates, and then complete the dump. If the lock is not in place the database will likely be changed between reading the log coordinates and making the dump even just a couple seconds later. If you're just looking to make a database dump that may be fine but it won't work for setting up replication.
@David: But correct coordinates (those recorded after locking all tables) will be added to the dump file automatically if --master-data parameter is used, aren't they? And the corresponding CHANGE MASTER TO MASTER_LOG_* statements will be executed on the slave the time of dump importing. So, I do not see the point to lock tables and record coordinates separately since --master-data parameter already does it.
Restoring a dump table into the MySQL master server can lead to serious replication delay.
To overcome the replication delay, caused by restoring the dump table on the master, there is a need to widespread the massive inserts. This can be done by the MySQL SLEEP command.
See http://www.mysqldiary.com/as-restoring-a-dump-table-into-the-mysql-master-you-better-get-some-sleep/
If get a "Access denied for user 'ODBC'@'localhost'" error message.
1.run "MySQLInstanceConfig.exe".
2.Select "Create An Anonymous Account" at "Please set the security options".
When using --master-data, as Sprout has pointed out, not all the CHANGE MASTER statements required are in the dumpfile.
His suggestion of editing the dumpfile will work, but not trying to do another CHANGE MASTER to set the host, user etc after importing the dump. That will only work if you put in the filename and log co-ordinates again, as I found to my cost.
It won't tell you what's wrong either!
Every execution of CHANGE MASTER blows off the co-ords, as it assumes you are changing to a new master.
I think you can set the user, etc before importing and it should work.
Add your own comment.