MariaDB/MySQL External Replication

In rare cases, it may be useful to maintain a replica instance of your MySQL/MariaDB database outside of Platform.sh.

Normally an automated backup is better for short-term usage and a mysqldump for longer term storage, but in some cases the data set is large enough that mysqldump is prohibitive. In that case, you can enable external replication using an extra permission.

Note that this guide covers the Platform.sh side; you need to set up and maintain your own replica instance. Consult the MySQL or MariaDB documentation for steps to do so.

Create a replication user

To set up replication you need to create a replication-enabled user. For each database that you’d like to replicate, you need to assign a replication permission/role, under a corresponding endpoint:

# The name of the service container. Must be unique within a project.
mariadb:
  type: mariadb:11.4
  disk: 1024
  configuration:
    schemas:
      - main
    endpoints:
      # Restate the default user to be used by your application.
      mysql:
        default_schema: main
        privileges:
          main: admin
      replicator:
        privileges:
          main: replication

This creates a replicator user, and grants read-only and table locking rights on the main database (namely Select_priv, Show_view_priv, Create_tmp_table_priv, Lock_tables_priv privileges) along with global replication rights (namely Repl_slave_priv and Repl_client_priv privileges) and flushing rights (Reload_priv used for flushing before reading the binary log position). If there is at least one replication permission defined, the bin-logging is enabled on the primary server, which is essential for the replication.

Add a relationship for the new endpoint

Even if you won’t be accessing the replication endpoint from your application, you still need to expose it to an application as a relationship so that you can connect to it over SSH. Add a new relationship to your application container:

name: myapp

[...]

# Relationships enable an app container's access to a service.
relationships:
  # Please note: Legacy definition of the relationship is still supported:
  # More information: https://docs.platform.sh/create-apps/app-reference/single-runtime-image.html#relationships
  database:
    service: "mariadb"
    endpoint: "mysql"
  replication:
    service: "mariadb"
    endpoint: "replicator"

Getting the Primary’s Binary Log Co-ordinates

Open the MySQL CLI to the replication relationship, either by accessing the credentials while on the app container or using the following command.

platform sql -r replication

Now you need to prevent any changes to the data while you view the binary log position. You’ll use this to tell the replica at exactly which point it should start replicating from. On the primary server, flush and lock all tables by running FLUSH TABLES WITH READ LOCK. Keep this session running - exiting it releases the lock. Get the current position in the binary log by running SHOW MASTER STATUS:

mysql> FLUSH TABLES WITH READ LOCK;
Query OK, 0 rows affected (0.016 sec)

mysql> SHOW MASTER STATUS;
+-----------------+----------+--------------+------------------+
| File            | Position | Binlog_Do_DB | Binlog_Ignore_DB |
+-----------------+----------+--------------+------------------+
| binlogs.000002  |  1036    | dflt         |                  |
+-----------------+----------+--------------+------------------+

Record the File and Position details. If binary logging has just been enabled, these are blank. Now, with the lock still in place, copy the data from the primary to the replica.

Login to the app container, then run:

# Dump the data from primary. Note that it dumps only the databases, which "replicator" user has access to.
$ mysqldump --all-databases --single-transaction -h database.internal -P 3306 -u replicator -p > /path/to/dump.sql

Download the dump file, then move it to the server where your replica lives to import it.

# Copy the dump to your replica
$ mysql -u root < /path/to/dump.sql

Note for live databases: You just need to make a local copy of the data, you don’t need to keep the primary locked until the replica has imported the data. Once the mysqldump has completed, you can release the lock on the primary by running UNLOCK TABLES.

mysql> UNLOCK TABLES;

Setting up the replica

Configuring the replica

As mentioned above you have to set up a replica on your own. Assuming that you have a running MariaDB/MySQL replica instance, give the replica a unique server_id (distinct from primary). You can find out primary’s server_id by running:

mysql> SHOW VARIABLES LIKE 'server_id';

+---------------+-------+
| Variable_name | Value |
+---------------+-------+
| server_id     | 1     |
+---------------+-------+

Then set a distinct server_id number (e.g. server_id+1) in your replica config (e.g. my.cnf) under:

[mysqld]
server_id=2

And reload the replica instance for the changes to take an effect.

Set up SSH tunneling

You need to set up an SSH tunnel from the replica server to the primary, tunneled through the application. To do so using the Platform.sh CLI, run the following (replacing BRANCH_NAME with the name of your production branch):

platform tunnel:open --project PROJECT_ID --environment BRANCH_NAME

This opens local SSH tunnels to all services accessible from the application. In practice, you may be better served by setting up the tunnel manually using SSH. Consult the SSH documentation for the best way to do so.

Starting the Replica

Once the data has been imported, you are ready to start replicating. Begin by running a CHANGE MASTER TO, making sure that MASTER_LOG_FILE matches the file and MASTER_LOG_POS the position returned by the earlier SHOW MASTER STATUS on the Platform.sh database. For example:

mysql> CHANGE MASTER TO
  MASTER_HOST='HOST',
  MASTER_USER='replicator',
  MASTER_PASSWORD='REPLICATOR_PASSWORD',
  MASTER_PORT=3306,
  MASTER_LOG_FILE='binlogs.000002',
  MASTER_LOG_POS=1036,
  MASTER_CONNECT_RETRY=10;

Where HOST varies depending on the SSH tunneling configuration you have, and the REPLICATOR_PASSWORD can be obtained by running platform relationships.

Now start the replica with the START SLAVE command:

mysql> START SLAVE;

Check that the replication is working by executing the SHOW SLAVE STATUS command:

mysql> SHOW SLAVE STATUS \G

If replication is working correctly, the values of both Slave_IO_Running and Slave_SQL_Running should be Yes:

Slave_IO_Running: Yes
Slave_SQL_Running: Yes

[Optional/Troubleshooting] Skipping invalid binary log queries

In some cases, after applying primary’s dump to the replica and starting the replica, you might experience replication errors (Slave_SQL_Running: No and Error: in the output of SHOW SLAVE STATUS \G above). Each of such errors needs a careful inspection, but you might be able to just skip some of them. For example:

mysql> STOP SLAVE; SET GLOBAL SQL_SLAVE_SKIP_COUNTER = 1; START SLAVE;
mysql> SHOW SLAVE STATUS \G

In case you have multiple errors you would need to repeat the steps above (preferred) or set SQL_SLAVE_SKIP_COUNTER (which corresponds to skipping the next N events from the primary) to something greater.