• Blog
  • Get Help Now!
  • Customer Portal

BlueGecko Denmark

  • Why you need us
    • Why act now
    • Why are we unique
  • What we do
    • Products Supported 24/7
      • MySQL
      • Postgres
      • SQL Server
      • Open Source
      • OS & Storage
  • How we do it
    • Consulting
    • System Monitoring
    • Security
  • Support
    • Blog
      • Oracle
      • MySQL
      • News & Events
      • Oracle Applications
      • Amazon Web Services
      • Jeremiah Wilton’s Oradeblog
      • OurSQL Community Podcast
      • Remote DBA
    • Team viewer
    • Whitepapers
  • About Us
    • Philosophy
    • The Blue Gecko Teams
    • Employment Opportunities
    • Partners
    • News
    • Press
  • Contact Us

October 22, 2010 by mhamrick

InnoDB tablespace fragmentation (find it and fix it!)

If you frequently delete rows (or update rows with variable-length data types), you can end up with a lot of wasted space in your data file(s), similar to filesystem fragmentation.

If you’re not using the innodb_file_per_table option, the only thing you can do about it is export and import the database, a time-and-disk-intensive procedure.

But if you are using innodb_file_per_table, you can identify and reclaim this space!

Prior to 5.1.21, the free space counter is available from the table_comment column of information_schema.tables. Here is some SQL to identify tables with at least 100M (actually 97.65M) of free space:

SELECT table_schema, table_name, table_comment FROM information_schema.tables WHERE engine LIKE 'InnoDB' AND table_comment RLIKE 'InnoDB free: ([0-9]{6,}).*';

Starting with 5.1.21, this was moved to the data_free column (a much more appropriate place):

SELECT table_schema, table_name, data_free/1024/1024 AS data_free_MB FROM information_schema.tables WHERE engine LIKE 'InnoDB' AND data_free > 100*1024*1024;

You can reclaim the lost space by rebuilding the table. The best way to do this is using ‘alter table’ without actually changing anything:

ALTER TABLE foo ENGINE=InnoDB;

This is what MySQL does behind the scenes if you run ‘optimize table’ on an InnoDB table. It will result in a read lock, but not a full table lock. How long it takes is completely dependent on the amount of data in the table (but not the size of the data file). If you have a table with a high volume of deletes or updates, you may want to run this monthly, or even weekly.

Filed Under: MySQL Tagged With: MySQL

Categories

  • Amazon Web Services
  • Configuration Management
  • downtime
  • Drizzle
  • ebs
  • Education
  • elastic block store
  • elastic compute cloud
  • hosting
  • hot backup
  • ignorance
  • Infrastructure
  • IOUG
  • Jeremiah Wilton's Oradeblog
  • misconception
  • misconceptions
  • Monocle
  • MySQL
  • MySQL Council
  • News
  • News & Events
  • Oracle
  • Oracle APEX
  • Oracle Applications
  • Oracle Data Guard
  • Oracle Database
  • Oracle Database Appliance
  • Oracle Linux
  • Our People
  • OurSQL Community Podcast
  • outage
  • parallel
  • performance
  • podcast
  • proof
  • Remote DBA
  • replication
  • S3
  • Security
  • SQL Server
  • System monitoring
  • Tomcat
  • Uncategorized

Contact Blue Gecko

  • This field is for validation purposes and should be left unchanged.

Get Our Newsletter

  • This field is for validation purposes and should be left unchanged.

The Fine Print

  • Privacy Policy
  • Terms of Use
  • When is remote DBA the right solution
  • What are remote DBA services
  • Get Help

Virtual runs

Blog Categories

  • 24/7 support
  • Blue Gecko 360° Services
  • Products Supported 24/7

Contact Blue Gecko

Blue Gecko A/S
Slotsgade 21
DK – 4200 Slagelse
Denmark
Phone: (DK) +45 70 60 51 20

Networks

  • Email
  • Twitter

© Blue Gecko Group