MySQL 如何批量kill 慢查询

Slow query 慢查询 是指 执行很慢的SQL语句。一般会设置一个阈值，例如，100ms，执行时间超过100ms的都会判定为慢查询。 慢查询是一种危险的信号，MySQL 服务可能很快不可用。当大量出现的时候，应该立即kill。

下文主要介绍如何批量kill 慢查询。

首先，查看有哪些慢查询的SQL：

select *  from information_schema.processlist where command='query' and time > 20;

其中，限定条件中加了command 和 time。

有时候，也不会限制command。

因为，对于慢查询，原因可能是空闲连接仍然占有锁，也需要注意。

例如，下面的语句，空闲连接占用锁，导致慢查一直在增加：

mysql >select *  from information_schema.processlist where time > 20;
+------------+---------------------+--------------------+--------------+---------+------+----------+------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+---------+-----------+---------------+
| ID         | USER                | HOST               | DB           | COMMAND | TIME | STATE    | INFO                                                                                                                                                                                                                                                       | TIME_MS | ROWS_SENT | ROWS_EXAMINED |
+------------+---------------------+--------------------+--------------+---------+------+----------+------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+---------+-----------+---------------+
| 1073085839 | orange_server | 10.88.6.26:46595  | orchestrator | Query   |   24 | updating | 
			delete from
				database_instance_maintenance
			where
				maintenance_active is null
				and end_timestamp < NOW() - INTERVAL 7 DAY |   23850 |         0 |             0 |
| 1073085838 | orange_server | 10.88.6.26:46600  | orchestrator | Query   |   27 | updating | 
			delete from
				database_instance_maintenance
			where
				maintenance_active is null
				and end_timestamp < NOW() - INTERVAL 7 DAY |   26850 |         0 |             0 |
| 1073084469 | orange_manager         | 10.1.10.68:32852 | orchestrator | Sleep   | 1806 |          | NULL                                                                                                                                                                                                                                                       | 1805814 |        16 |            16 |
| 1073085824 | orange_server | 10.88.6.26:46594  | orchestrator | Query   |   21 | updating | 
			delete from
				database_instance_maintenance
			where
				maintenance_active is null
				and end_timestamp < NOW() - INTERVAL 7 DAY |   20850 |         0 |             0 |
| 1073085828 | orange_server | 10.88.6.26:46615  | orchestrator | Query   |   25 | updating | 
			delete from
				database_instance_maintenance
			where
				maintenance_active is null
				and end_timestamp < NOW() - INTERVAL 7 DAY |   24850 |         0 |             0 |
| 1073085829 | orange_server | 10.88.6.26:46605  | orchestrator | Sleep   |   30 |          | NULL                                                                                                                                                                                                                                                       |   29850 |         0 |             0 |
| 1073085769 | orange_server | 10.88.6.26:46547  | orchestrator | Query   |   26 | updating | 
			delete from
				database_instance_maintenance
			where
				maintenance_active is null
				and end_timestamp < NOW() - INTERVAL 7 DAY |   25850 |         0 |             0 |
| 1073085833 | orange_server | 10.88.6.26:46614  | orchestrator | Query   |   23 | updating | 
			delete from
				database_instance_maintenance
			where
				maintenance_active is null
				and end_timestamp < NOW() - INTERVAL 7 DAY |   22850 |         0 |             0 |
| 1073085782 | orange_server | 10.88.6.26:46545  | orchestrator | Query   |   22 | updating | 
			delete from
				database_instance_maintenance
			where
				maintenance_active is null
				and end_timestamp < NOW() - INTERVAL 7 DAY |   21850 |         0 |             0 |
| 1073085825 | orange_server | 10.88.6.26:46601  | orchestrator | Query   |   29 | updating | 
			delete from
				database_instance_maintenance
			where
				maintenance_active is null
				and end_timestamp < NOW() - INTERVAL 7 DAY |   28851 |         0 |             0 |
| 1073085785 | orange_server | 10.88.6.26:46553  | orchestrator | Query   |   28 | updating | 
			delete from
				database_instance_maintenance
			where
				maintenance_active is null
				and end_timestamp < NOW() - INTERVAL 7 DAY |   27851 |         0 |             0 |
+------------+---------------------+--------------------+--------------+---------+------+----------+------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+---------+-----------+---------------+
11 rows in set (0.02 sec)

当出现大量慢查的时候，很可能将MySQL 服务拖垮。当机立断，应该赶紧kill 慢查询。

批量kill 慢查的方式如下，首先找到慢查询id，然后进行kill。

以下SQL，可以组合成kill 语句，直接拷贝执行即可。

mysql >select concat('kill ', id,';') from information_schema.processlist where time > 20;
+-------------------------+
| concat('kill ', id,';') |
+-------------------------+
| kill 1073080655;        |
| kill 1073085823;        |
| kill 1073084469;        |
| kill 1073085837;        |
| kill 1073085778;        |
| kill 1073085869;        |
| kill 1073085772;        |
| kill 1073085787;        |
| kill 1073085786;        |
| kill 1073085780;        |
| kill 1073085779;        |
| kill 1073085858;        |
+-------------------------+
12 rows in set (0.03 sec)

以上，介绍了批量kill 慢查询的方式。