为什么 DELETE 比 SELECT 慢那么多，然后按 id 删除？

Question

我有一个相当繁忙的 InnoDB 表（200,000 行，我猜大概每秒有几十个查询）。由于一个错误，我得到了 14 行（相同的）无效电子邮件地址，并想删除它们。

我只是尝试DELETE FROM table WHERE email='invalid address'并在大约 50 秒后得到“超出锁定等待超时”。这并不奇怪，因为行列没有编入索引。

但是，然后我做到了SELECT id FROM table WHERE email='invalid address'，这花了 1.25 秒。运行DELETE FROM table WHERE id in (...)，从 SELECT 结果中复制粘贴 id，花费了 0.02 秒。

到底是怎么回事？有人可以解释为什么带有条件的 DELETE 如此缓慢以至于超时，但是执行 SELECT 然后按 id 删除却如此之快？

谢谢。

编辑：根据要求，我发布了表结构以及一些explain结果。我还应该注意，没有引用此表的外键。

但是，情况对我来说似乎很简单：我选择了一个未编入索引的字段。这需要扫描整个表，但它并不是很大。id是主键，因此按 id 删除非常快，应该如此。

mysql> show create table ThreadNotification2 \G
*************************** 1. row ***************************
       Table: ThreadNotification2
Create Table: CREATE TABLE `ThreadNotification2` (
  `id` bigint(20) NOT NULL AUTO_INCREMENT,
  `alertId` bigint(20) DEFAULT NULL,
  `day` int(11) NOT NULL,
  `frequency` int(11) DEFAULT NULL,
  `hour` int(11) NOT NULL,
  `email` varchar(255) DEFAULT NULL,
  `highlightedTitle` longtext,
  `newReplies` bit(1) NOT NULL,
  `numReplies` int(11) NOT NULL,
  `postUrl` longtext,
  `sendTime` datetime DEFAULT NULL,
  `sent` bit(1) NOT NULL,
  `snippet` longtext,
  `label_id` bigint(20) DEFAULT NULL,
  `organization_id` bigint(20) DEFAULT NULL,
  `threadEntity_hash` varchar(255) DEFAULT NULL,
  `user_uid` bigint(20) DEFAULT NULL,
  PRIMARY KEY (`id`),
  KEY `FK3991E9D279251FE` (`organization_id`),
  KEY `FK3991E9D35FC0C96` (`label_id`),
  KEY `FK3991E9D3FFC22CB` (`user_uid`),
  KEY `FK3991E9D5376B351` (`threadEntity_hash`),
  KEY `scheduleSentReplies` (`day`,`frequency`,`hour`,`sent`,`numReplies`),
  KEY `sendTime` (`sendTime`),
  CONSTRAINT `FK3991E9D279251FE` FOREIGN KEY (`organization_id`) REFERENCES `Organization` (`id`),
  CONSTRAINT `FK3991E9D35FC0C96` FOREIGN KEY (`label_id`) REFERENCES `Label` (`id`),
  CONSTRAINT `FK3991E9D3FFC22CB` FOREIGN KEY (`user_uid`) REFERENCES `User` (`uid`),
  CONSTRAINT `FK3991E9D5376B351` FOREIGN KEY (`threadEntity_hash`) REFERENCES `ThreadEntity` (`hash`)
) ENGINE=InnoDB AUTO_INCREMENT=4461945 DEFAULT CHARSET=utf8
1 row in set (0.08 sec)

mysql> explain SELECT * FROM ThreadNotification2 WHERE email='invalid address';
+----+-------------+---------------------+------+---------------+------+---------+------+--------+-------------+
| id | select_type | table               | type | possible_keys | key  | key_len | ref  | rows   | Extra       |
+----+-------------+---------------------+------+---------------+------+---------+------+--------+-------------+
|  1 | SIMPLE      | ThreadNotification2 | ALL  | NULL          | NULL | NULL    | NULL | 197414 | Using where |
+----+-------------+---------------------+------+---------------+------+---------+------+--------+-------------+
1 row in set (0.03 sec)


mysql> explain select * from ThreadNotification2 where id in (3940042,3940237,3941132,3941255,3941362,3942535,3943064,3944134,3944228,3948122,3953081,3957876,3963849,3966951);
+----+-------------+---------------------+-------+---------------+---------+---------+------+------+-------------+
| id | select_type | table               | type  | possible_keys | key     | key_len | ref  | rows | Extra       |
+----+-------------+---------------------+-------+---------------+---------+---------+------+------+-------------+
|  1 | SIMPLE      | ThreadNotification2 | range | PRIMARY       | PRIMARY | 8       | NULL |   14 | Using where |
+----+-------------+---------------------+-------+---------------+---------+---------+------+------+-------------+
1 row in set (0.00 sec)



mysql> delete from ThreadNotification2 where email='invalid address';
ERROR 1205 (HY000): Lock wait timeout exceeded; try restarting transaction
mysql> select id from ThreadNotification2 where email='invalid address';
+---------+
| id      |
+---------+
| 3940042 |
| 3940237 |
| 3941132 |
| 3941255 |
| 3941362 |
| 3942535 |
| 3943064 |
| 3944134 |
| 3944228 |
| 3948122 |
| 3953081 |
| 3957876 |
| 3963849 |
| 3966951 |
+---------+
14 rows in set (1.25 sec)

mysql> delete from ThreadNotification2 where id in (3940042,3940237,3941132,3941255,3941362,3942535,3943064,3944134,3944228,3948122,3953081,3957876,3963849,3966951);
Query OK, 14 rows affected (0.02 sec)

Answer 1

如果该字段email是未编入索引，那么这两个DELETE和SELECT应该一样慢。

我能想到的唯一可能性是：您说该表被大量访问。也许其他人在您尝试执行DELETE.

我想也许你应该在那里插入一些模拟行并尝试删除它们。这样做2或3次。如果 的持续时间有很大差异DELETE，那么可能是 DB 负载的原因。

PS：只有当人们不会被那些模拟行惹恼时才这样做：D。