Question

我有这样的表用于在mysql中实现队列：

CREATE TABLE `queue` (
  `id` int(10) unsigned NOT NULL AUTO_INCREMENT,
  `queue_name` varchar(255) NOT NULL,
  `inserted` timestamp NOT NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
  `inserted_by` varchar(255) NOT NULL,
  `acquired` timestamp NULL DEFAULT NULL,
  `acquired_by` varchar(255) DEFAULT NULL,
  `delayed_to` timestamp NULL DEFAULT NULL,
  `priority` int(11) NOT NULL DEFAULT '0',
  `value` text NOT NULL,
  `status` varchar(255) NOT NULL DEFAULT 'new',
  PRIMARY KEY (`id`),
  KEY `queue_index` (`acquired`,`queue_name`,`priority`,`id`)
) ENGINE=InnoDB AUTO_INCREMENT=1 DEFAULT CHARSET=utf8

我的问题是mysql在运行update时使用filesort。执行速度非常慢（表中800k行为5秒）。

DESCRIBE UPDATE queue SET acquired = "test" WHERE acquired IS NULL AND queue_name = "q1" ORDER BY priority, id LIMIT 1;

+----+-------------+-------+-------+---------------+-------------+---------+-------------+--------+-----------------------------+
| id | select_type | table | type  | possible_keys | key         | key_len | ref         | rows   | Extra                       |
+----+-------------+-------+-------+---------------+-------------+---------+-------------+--------+-----------------------------+
|  1 | SIMPLE      | queue | range | queue_index   | queue_index | 772     | const,const | 409367 | Using where; Using filesort |
+----+-------------+-------+-------+---------------+-------------+---------+-------------+--------+-----------------------------+

奇怪的是，当我使用相同的WHERE条件运行SELECT查询并且不使用ORDER列filesort时：

DESCRIBE SELECT id FROM queue WHERE acquired IS NULL AND queue_name = "q1" ORDER BY priority, id LIMIT 1;

+----+-------------+-------+------+---------------+-------------+---------+-------------+--------+--------------------------+
| id | select_type | table | type | possible_keys | key         | key_len | ref         | rows   | Extra                    |
+----+-------------+-------+------+---------------+-------------+---------+-------------+--------+--------------------------+
|  1 | SIMPLE      | queue | ref  | queue_index   | queue_index | 772     | const,const | 409367 | Using where; Using index |
+----+-------------+-------+------+---------------+-------------+---------+-------------+--------+--------------------------+

（查询时间0s）

是否有人知道如何避免在更新查询中使用filesort或如何提高其性能？

此致 Matzz

Answer 1

在mysql论坛（http://forums.mysql.com/read.php?24,620908,620908#msg-620908）讨论后，我报告了错误http://bugs.mysql.com/bug.php?id=74049（已经过验证）。可以使用SELECT FOR UPDATE：

绕过问题

SET @update_id := -1;
SELECT (SELECT @update_id := id)
FROM queue
WHERE acquired IS NULL AND queue_name = "q1"
ORDER BY priority, id LIMIT 1;
FOR UPDATE;
UPDATE queue SET acquired = "test" WHERE id = @update_id;

消除更新查询中的mysql文件排序

1 个答案: