How to move InnoDB table to another drive? - mysql

Question

I have MySQL running on SSDs, SSDs that I'm about to run out of space on. My webhost overcharges for SSDs and the majority of the data in MySQL is "archived" data (i.e. data that isn't actively used). I have larger HDDs that can hold this data. As such, I want to be able to move specific InnoDB tables from the SSDs to the HDDs.
One solution I've thought about and researched is moving the individual .ibd files (I have innodb_file_per_table enabled) for the specific tables in question to the HDDs and then symlink. However, researching this, it looks like that is a bad idea for InnoDB.
I've also seen that since 5.6, MySQL supports the DATA DIRECTORY command:
To create a new InnoDB file-per-table tablespace in a specific
location outside the MySQL data directory, use the DATA DIRECTORY =
absolute_path_to_directory clause of the CREATE TABLE statement.
Plan the location in advance, because you cannot use the DATA
DIRECTORY clause with the ALTER TABLE statement. The directory you
specify could be on another storage device with particular performance
or capacity characteristics, such as a fast SSD or a high-capacity
HDD.
The problem is, it looks like this is only supported for new tables. I want to do it for existing tables. Any tips on how? I'm running Percona MySQL, if it helps.
Thanks!
UPDATE: Here is what I tried, but I'm getting a syntax error:
CREATE TABLE abc_2 LIKE abc ENGINE=InnoDB DATA DIRECTORY='/xxx/mysql/archive/'

Answer 1

Apparently CREATE ... LIKE ... DATA DIRECTORY ... is a combination that is not supported.
Do SHOW CREATE TABLE to get the current definition. Edit it to add DATA DIRECTORY and INDEX_DIRECTORY. Then use the edited text to create the new table.
Then INSERT INTO new_tbl SELECT * FROM real_tbl; and shuffle the names: RENAME TABLE real_tbl TO old_tbl, new_tbl TO real_tbl;.
Verify the results and finally DROP old_tbl;

Answer 2

I have dealt with this problem myself, and eventually found a more elegant solution than 'create new - copy - switch': detaching, moving and re-importing the underlying tablespace files. This is much more efficient on large and/or heavily indexed tables as MySQL does not have to redo work it has already done.
In short, it comes down to the following steps:
FLUSH TABLES `table_name` FOR EXPORT;
While keeping the connection open, move the tablespace files in a shell:
$ mv /var/lib/mysql/database_name/table_name.{ibd,cfg} ~
Now back in MySQL release the lock, drop the table, re-create it with the correct DATA DIRECTORY and discard its tablespace:
UNLOCK TABLES;
SHOW CREATE TABLE `table_name`;
DROP TABLE `table_name`;
CREATE TABLE `table_name` /* ... */ DATA DIRECTORY='/path/to/desired/location';
ALTER TABLE `table_name` DISCARD TABLESPACE;
Now copy the moved tablespace files to the desired location:
$ cp -a ~/table_name.{ibd,cfg} /path/to/desired/location
And import them:
ALTER TABLE `table_name` IMPORT TABLESPACE;
More background and motivation for why 'create new - copy - switch' is inefficient can be found in a blogpost I wrote on this topic: https://www.moxio.com/blog/28/moving-individual-mysql-tables-on-disk.