MySQL Forums :: InnoDB :: Innodb unexplained IOs

This might be a factor: 197333 inserts, 167667 merged recs, 17805 merges. One of the user comments in the manual suggests that the activity will continue with merging the inserts until the number of merged records equals the number of inserts, as the records in the insert buffer are merged into the clustered index nodes. You can read more about the insert buffer at http://dev.mysql.com/doc/mysql/en/innodb-insert-buffering.html .

In addition, there's the "Main thread id 8, state: purging " portion, indicating the presence of deleted record purging activity.

Those suggest that the server had more insert and deletion activity that it could sustain and is still in the process of catching up, saved from being overwhelmed by the insert buffer and delayed delete purging.

If able you might try keeping laod from it and/or allocating more InnoDB cache RAM than usual to help to reduce the disk load as far as practical while this work continues.