Home » Mysql » Too many auto increments with ON DUPLICATE KEY UPDATE

Too many auto increments with ON DUPLICATE KEY UPDATE

Posted by: admin November 30, 2017 Leave a comment

Questions:

I have a basic table with columns:

  • id (primary with AI)
  • name (unique)
  • etc

If the unique column doesn’t exist, INSERT the row, otherwise UPDATE the row….

INSERT INTO pages (name, etc)
VALUES
  'bob',
  'randomness'
ON DUPLICATE KEY UPDATE
 name = VALUES(name),
 etc = VALUES(etc)

The problem is that if it performs an UPDATE, the auto_increment value on the id column goes up. So if a whole bunch of UPDATES are performed, the id auto_increment goes through the roof.

Apparently it was a bug: http://bugs.mysql.com/bug.php?id=28781

…but I’m using InnoDB on mySQL 5.5.8 on shared hosting.

Other people having issues with no solution years ago:
prevent autoincrement on MYSQL duplicate insert and
Why does MySQL autoincrement increase on failed inserts?

Ideas on a fix? Have I maybe structured the database incorrectly somehow?

******EDIT****: It appears adding innodb_autoinc_lock_mode = 0 to your my.ini file fixes the problem but what options do I have for shared hosting?

******EDIT 2******: OK, I think my only option is to change to MyISAM as the storage engine. Being a mega mySQL newbie, I hope that doesn’t cause many issues. Yeah?

Answers:

I don’t think there is a way to bypass this behaviour of INSERT ... ON DUPLICTE KEY UPDATE.

You can however put two statements, one UPDATE and one INSERT, in one transaction:

START TRANSACTION ;

UPDATE pages
SET etc = 'randomness'
WHERE name = 'bob' ;

INSERT INTO pages (name, etc)
SELECT 
      'bob' AS name
    , 'randomness' AS etc 
FROM dual 
WHERE NOT EXISTS
      ( SELECT *
        FROM pages p
        WHERE p.name = 'bob'
      ) ;

COMMIT ;

Questions:
Answers:

The on duplicate key functionality of MySQL is exactly the same as doing two separate queries, one to select, then one to either update the selected record, or insert a new record. Doing so programmatically is just as fast and will prevent this problem in the future as well as make your code more portable.