Improving MySQL with Tarantool

Replicating MySQL is one of the in-memory-database Tarantool’s killer functions. It allows you to keep your existing MySQL database while at the same time accelerating it and scaling it out horizontally. Even if you aren’t interested in extensive expansion, simply replacing existing replicas with Tarantool can save you money, because Tarantool is more efficient per core than MySQL. To read a testimonial of a company that implemented Tarantool replication on a large scale, please see here, as well as here.

I wanted to point out at the outset that if you run into any trouble with regards to the basics of Tarantool, you may wish to consult the first two tutorials in this Tarantool 101 series, which can be found here and here.

And please note that these instructions are for CentOS 7.5 and MySQL 5.7. They also assume that you have systemd installed and are working with an existing MySQL installation. Finally, a helpful log for troubleshooting during this tutorial is replicatord.log in /var/log. You can also have a look at the instance’s log example.log in /var/log/tarantool.

So let’s proceed.

  1. First we’ll install the necessary packages in CentOS:
yum -y install git ncurses-devel cmake gcc-c++ boost boost-devel wget unzip nano bzip2 mysql-devel mysql-lib

2. Next we’ll clone the Tarantool-MySQL replication package from GitHub:

git clone

3. Now we can build the replicator with cmake:

cd mysql-tarantool-replication
git submodule update --init --recursive
cmake .

4. Our replicator will run as a systemd daemon called replicatord, so let’s edit its systemd service file, replicatord.service, in the mysql-tarantool-replication repo.

nano replicatord.service

Change the following line:

ExecStart=/usr/local/sbin/replicatord -c /usr/local/etc/replicatord.cfg

It should read:

ExecStart=/usr/local/sbin/replicatord -c /usr/local/etc/replicatord.yml

5. Next let’s copy some files from our replicatord repo to other necessary locations:

cp replicatord /usr/local/sbin/replicatord
cp replicatord.service /etc/systemd/system

6. Now let’s enter the MySQL console and create a sample database (depending on your existing installation, you may of course be a user other than root):

mysql -u root -p

7. Next we’ll get some sample data from MySQL, which we’ll pull into our root directory, then install from the terminal:

cd menagerie-db
mysql -u root -p menagerie < cr_pet_tbl.sql
mysql -u root -p menagerie < load_pet_tbl.sql
mysql menagerie -u root -p < ins_puff_rec.sql
mysql menagerie -u root -p < cr_event_tbl.sql

8. Let’s enter the MySQL console now and massage the data for use with the Tarantool replicator (we are adding an ID, changing a field name to avoid conflict, and cutting down the number of fields; note that with real data, this is the step that will involve the most tweaking):

mysql -u root -p
USE menagerie;
ALTER TABLE pet CHANGE COLUMN `name` `name2` VARCHAR(255);
ALTER TABLE pet DROP sex, DROP birth, DROP death;

9. Now that we have the sample data set up, we’ll need to edit MySQL’s configuration file for use with the replicator. Everything will go directly under “[mysqld]” in the file. (Note that your my.cnf for MySQL could be in a slightly different location.)

cd nano /etc/my.cnf


binlog_format = ROW server_id = 1 log-bin=mysql-bin interactive_timeout=3600 wait_timeout=3600 max_allowed_packet=32M

10. After exiting nano, we’ll restart mysqld:

systemctl restart mysqld

11. ext, let’s install Tarantool and set up spaces for replication. Copy the entire script at and paste it into your terminal. You may have to press “enter” for the last line (“yum -y install tarantool”).

12. Next, let’s install Tarantool and set up spaces for replication. Copy the entire script at arantool:1

nano /etc/tarantool/instances.available/example.lua

13. Replace the entire contents of the file with the following:

box.cfg {
    listen = 3301;
    memtx_memory = 128 * 1024 * 1024; -- 128Mb
    memtx_min_tuple_size = 16;
    memtx_max_tuple_size = 128 * 1024 * 1024; -- 128Mb
    vinyl_memory = 128 * 1024 * 1024; -- 128Mb
    vinyl_cache = 128 * 1024 * 1024; -- 128Mb
    vinyl_max_tuple_size = 128 * 1024 * 1024; -- 128Mb
    vinyl_write_threads = 2;
    wal_mode = "none";
    wal_max_size = 256 * 1024 * 1024;
    checkpoint_interval = 60 * 60; -- one hour
    checkpoint_count = 6;
    force_recovery = true;
     -- 1 – SYSERROR
     -- 2 – ERROR
     -- 3 – CRITICAL
     -- 4 – WARNING
     -- 5 – INFO
     -- 6 – VERBOSE
     -- 7 – DEBUG
     log_level = 7;
     log_nonblock = true;
     too_long_threshold = 0.5;
 local function bootstrap()
     if not then
         s ='mysqldaemon')
         {type = 'tree', parts = {1, 'unsigned'}})
     if not then
         t ='mysqldata')
         {type = 'tree', parts = {1, 'unsigned'}})

To understand more of what’s happening here, it would be best to have a look back at the earlier articles in our Tarantool 101 series.

14. Now we need to create a symlink from instances.available to a directory named instances.enabled (similar to NGINX). So in /etc/tarantool run the following:

mkdir instances.enabled
ln -s /instances.available/example.lua instances.enabled

15. Next we can start up our Lua program with tarantoolctl, a wrapper for systemd:

tarantoolctl start example.lua

16. Now let’s enter our Tarantool instance, where we can check that our target spaces were successfully created:

tarantoolctl enter example.lua

17. Now that we have MySQL and Tarantool set up, we can proceed to configure our replicator. First let’s work with replicator.yml in the main tarantool-mysql-replication directory.

nano replicatord.yml

Change the entire file as follows, making sure to add your MySQL password and to set the appropriate user:

     port: 3306
     user: root
     connect_retry: 15 # seconds
     binlog_pos_space: 512
     binlog_pos_key: 0
     connect_retry: 15 # seconds
     sync_retry: 1000 # milliseconds
     - database: menagerie
     table: pet
     columns: [ id, name2, owner, species]
     space: 513
     key_fields:  [0]
     # insert_call: function_name
     # update_call: function_name
     # delete_call: function_name

18. Now we need to copy replicatord.yml to the location where systemd looks for it:

cp replicatord.yml /usr/local/etc/replicatord.yml

19. Next we can start up the replicator:

systemctl start replicatord

Now we can enter our Tarantool instance and do a select on the “mysqldata” space. We will see the replicated content from MySQL:

tarantoolctl enter example.lua

The results will be:

- - [1, 'Fluffy', 'Harold', 'cat']
   - [2, 'Claws', 'Gwen', 'cat']
   - [3, 'Buffy', 'Harold', 'dog']
   - [4, 'Fang', 'Benny', 'dog']
   - [5, 'Bowser', 'Diane', 'dog']
   - [6, 'Chirpy', 'Gwen', 'bird']
   - [7, 'Whistler', 'Gwen', 'bird']
   - [8, 'Slim', 'Benny', 'snake']
   - [9, 'Puffball', 'Diane', 'hamster']

21. Finally let’s enter a record into MySQL and then go back to Tarantool to make sure it’s replicated. So first we’ll exit our Tarantool instance:



mysql -u root -p
 USE menagerie;
 INSERT INTO pet(name2, owner, species) VALUES (‘Spot’, ‘Brad’, ‘dog’);

(Note that for the insert statement above, you may need to change the quotes to straight quotes if you are having issues.)
Once back in the terminal enter:

tarantoolctl enter example.lua

You should see the replicated data in Tarantool!