Huge ibdata1 files getting created during incremental restores

Asked by Mike Siekkinen on 2013-02-15

I've been testing incremental backups/restors with xtrabackup/innobackupex. I have one DB that is all innodb, about 18G total.

After copying the first full and running w/ apply-logs redo only, that seems fine. Then when I try to apply the first incremental (which was about ~2G total) it's been sitting there spinning for nearly an hour. The ibdata1 for the restored backup has the ibdata1 growing to over 300G!

This is with file per table turned on.

What could be causing this? This seems very very wrong. I didn't anticipate needing that much space (or who know how much more it would ultimately consume) so I ended up killing the restore as I really didn't know if it was in some infinite loop that would go until space ran out.

Is this expected? Is there a known issue of something I'm doing wrong that could be causing this?

Question information

Language:
English Edit question
Status:
Solved
For:
Percona XtraBackup moved to https://jira.percona.com/projects/PXB Edit question
Assignee:
No assignee Edit question
Solved by:
Mike Siekkinen
Solved:
2013-02-27
Last query:
2013-02-27
Last reply:
2013-02-15
Alexey Kopytov (akopytov) said : #1

Hi Mike,

No, it doesn't look like a known problem. What's the size of ibdata1 before applying the first incremental? Does this growth to 300MB occur at the time .delta files are applied or when applying the redo log from incremental (that should be clear from the xtrabackup output). Having the full log from xtrabackup might also be helpful. Thanks.

Mike Siekkinen (9-mikes) said : #2

Hi,

This was a result of me using "xtrabackup" instead of the "xtrabackup_55" binary in my tests. Switching that looks to have resolved this problem.