MySQL JOIN very slow for no apparent reason

Question

MySQL version 8.0.17:

SELECT b.ColB
FROM Table2 b
JOIN Table1 a ON a.ColA = b.ColA
WHERE b.ColB = 1234 -- some number

With the JOIN , this query takes about 2 seconds to run, without the JOIN it is almost instant. This is just a minimal example, the actual case requires the WHERE condition to cover multiple values (eg WHERE b.ColB BETWEEN 1000 AND 2000 ) - which takes a very very long time...

Structure of Table1 (irrelevant columns and indexes removed for brevity):

CREATE TABLE `Table1` (
  `ColA` int(11) NOT NULL AUTO_INCREMENT,
  PRIMARY KEY (`ColA`)
) ENGINE=InnoDB AUTO_INCREMENT=138221783 DEFAULT CHARSET=utf8 COLLATE=utf8_bin;

Structure of Table2 (irrelevant columns and indexes removed for brevity):

CREATE TABLE `Table2` (
  `ColA` int(11) NOT NULL,
  `ColB` int(11) NOT NULL,
  PRIMARY KEY (`ColA`,`ColB`),
  KEY `FK_Table2_Table1_idx` (`ColA`),
  KEY `FK_Table2_Table3_idx` (`ColB`),
  CONSTRAINT `FK_Table2_Table1` FOREIGN KEY (`ColA`) REFERENCES `Table1` (`ColA`) ON DELETE CASCADE ON UPDATE CASCADE
) ENGINE=InnoDB DEFAULT CHARSET=utf8 COLLATE=utf8_bin;

This is the CSV of the explain on this query:

id,select_type,table,partitions,type,possible_keys,key,key_len,ref,rows,filtered,Extra
1,SIMPLE,rc,NULL,ref,"PRIMARY,FK_Table2_Table1_idx,FK_Table2_Table3_idx",FK_Table2_Table3_idx,4,const,607,100.00,"Using index"
1,SIMPLE,r,NULL,eq_ref,PRIMARY,PRIMARY,4,Table2.ColA,1,100.00,"Using index"

Some extra info running this SELECT COUNT(*) FROM Table1 took 8.735 seconds and returned 1947948.

Running this SELECT COUNT(*) FROM Table2 took 168.422 seconds and returned 19486319.

Why is the JOIN causing the query to run so very slowly?

Answer 1

If the only index is PRIMARY KEY(colA, colB) and the WHERE clause is testing colB=constant , then there is no usable index.

If this is a many-to-many mapping table you need to add INDEX(colB, colA) to allow for efficiently going from B to A.

Answer 2

To reduce the number of Rows that have to be joined reduce the number of rows as much as you can.

For your qiery

Use

SELECT b.ColB
FROM (SELECT * FROM Table2 WHERE ColB = 1234) b
JOIN Table1 a ON a.ColA = b.ColA

Instead of the SELECT * FROM Table2, you must use only the column that you actually need.