为什么MySQL查询返回零行，但是在进行碎片整理后仍然有效？

Question

I have an InnoDB, MySQL table and this query returns zero rows:

SELECT id, display FROM ra_table WHERE parent_id=7266 AND display=1;

However, there are actually 17 rows that should match:

SELECT id, display FROM ra_itable1 WHERE parent_id=7266;

ID    display
------------------
1748  1
5645  1
...

There is an index on display (int 1), and ID is the primary key. The table also has several other fields which I'm not pulling in this query.

After noticing this query wasn't working, I defragmented the table and then the first query started working correctly, but only for a time. It seems after a few days, the query stops working again and I have to defragment to fix it.

My question is, why does the fragmented table break this query?

Additional info: MySQL 5.6.27 on Amazon RDS.

CREATE TABLE `ra_table` (
`id` int(11) NOT NULL AUTO_INCREMENT,
`parent_id` int(6) NOT NULL,
`display` int(1) NOT NULL,
PRIMARY KEY (`id`),
KEY `parent_id` (`parent_id`),
KEY `display` (`display`),
) ENGINE=InnoDB AUTO_INCREMENT=13302 DEFAULT CHARSET=latin1 
ROW_FORMAT=DYNAMIC

Answer 1

There may be a bug in the version you are running.

Meanwhile, change

INDEX(parent_id),
INDEX(display)

to

INDEX(parent_id, display)

By combining them, the query will run faster (and hopefully correctly). An index on a flag ( display ) is likely to never be used.