背景
在一次日常测试中发现,kill 一个会话后,SQL语句依然在运行并没终止;被kill的会话重新连接并继续执行原来的SQL语句。
测试
本次测试基于MySQL 8.0.27
1.创建测试表
create table t1 (id int, name varchar(30)); insert into t1 values (1,'a'),(2,'b');
2.开启3个会话
session1 | session2 | session3 |
begin; | ||
select * from t1; | ||
rename table t1 to t2; 【由于锁等待,hang住】 | ||
show processlist; 【查看 processlist_id】 | ||
kill session2; | ||
【session2 重新连接并且继续执行语句,处于锁等待状态】 | ||
show processlist; 【可以看到session2重新连接并继续执行SQL】 | ||
commit; | ||
【rename 执行成功】 | ||
show tables; 【t1 被 rename 为 t2】 |
session1:开启一个事务不提交
mysql> use test Database changed mysql> mysql> mysql> begin; Query OK, 0 rows affected (0.00 sec) mysql> select * from t1; +------+------+ | id | name | +------+------+ | 1 | a | | 2 | b | +------+------+ 2 rows in set (0.00 sec)
session2:执行DDL语句
mysql> use test Database changed mysql> mysql> mysql> rename table t1 to t2;
session3:kill session2
mysql> show processlist; +-----+------+---------+---------+--------------------------------+----------------------+ | Id | db | Command | Time | State | Info | +-----+------+---------+---------+--------------------------------+----------------------+ | 6 | NULL | Daemon | 4399013 | Waiting on empty queue | NULL | | 132 | test | Sleep | 232 | | NULL | | 134 | test | Query | 123 | Waiting for table metadata lock| rename table t1 to t2| | 135 | test | Query | 0 | init | show processlist | +-----+------+---------+---------+--------------------------------+----------------------+ 4 rows in set (0.00 sec) mysql> kill 134; Query OK, 0 rows affected (0.01 sec) #为了排版,表格字段略有删减,具体信息请看图片
session2:session2重新连接,并且继续执行DDL语句,仍处于锁等待状态
mysql> rename table t1 to t2; ERROR 2013 (HY000): Lost connection to MySQL server during query No connection. Trying to reconnect... Connection id: 136 Current database: test
session3:查看会话信息
mysql> show processlist; +-----+------+---------+---------+--------------------------------+----------------------+ | Id | db | Command | Time | State | Info | +-----+------+---------+---------+--------------------------------+----------------------+ | 6 | NULL | Daemon | 4399260 | Waiting on empty queue | NULL | | 132 | test | Sleep | 479 | | NULL | | 135 | test | Query | 0 | init | show processlist | | 136 | test | Query | 193 | Waiting for table metadata lock| rename table t1 to t2| +-----+------+---------+---------+--------------------------------+----------------------+ 4 rows in set (0.00 sec) #为了排版,表格字段略有删减,具体信息请看图片
可以看到, kill session2 后,session2 重新连接并且继续执行SQL
session1:提交事务
mysql> commit; Query OK, 0 rows affected (0.01 sec)
session2:执行成功
mysql> use test Database changed mysql> mysql> mysql> rename table t1 to t2; ERROR 2013 (HY000): Lost connection to MySQL server during query No connection. Trying to reconnect... Connection id: 136 Current database: test Query OK, 0 rows affected (8 min 38.00 sec)
通过上述测试,可以看到明明执行了 kill 命令,但是依然没有达到我们想要的效果,似乎 kill 命令没有生效一样。
经过查询资料发现,由于通过MySQL客户端登录,--reconnect
重新连接选项默认是开启的,该选项在每次连接丢失时都会进行一次重新连接尝试;因此在kill session2 后,session2重新连接并再次执行之前的SQL语句,导致感觉 kill 命令没有生效。
--reconnect Reconnect if the connection is lost. Disable with --disable-reconnect. This option is enabled by default. (Defaults to on; use --skip-reconnect to disable.)
解决
可以通过以下2种方式避免上述问题的发生:
1.执行kill query 命令
KILL QUERY
终止连接当前正在执行的语句,但保持连接本身不变
session3:执行 KILL QUERY
命令
mysql> show processlist; +-----+------+---------+---------+--------------------------------+----------------------+ | Id | db | Command | Time | State | Info | +-----+------+---------+---------+--------------------------------+----------------------+ | 6 | NULL | Daemon | 4401560 | Waiting on empty queue | NULL | | 132 | test | Sleep | 11 | | NULL | | 135 | test | Query | 0 | init | show processlist | | 137 | test | Query | 3 | Waiting for table metadata lock| rename table t1 to t2| +-----+------+---------+---------+--------------------------------+----------------------+ 4 rows in set (0.00 sec) mysql> mysql> kill query 137; Query OK, 0 rows affected (0.00 sec) #为了排版,表格字段略有删减,具体信息请看图片
session2:
mysql> rename table t1 to t2; ERROR 1317 (70100): Query execution was interrupted
可以看到session2执行的语句已经被终止了,达到了我们想要的效果。
2.登录mysql客户端时加--skip-reconnect选项
--skip-reconnect
表示当连接丢失时不会进行重新连接的尝试
session2:登录时加 --skip-reconnect
选项
shell> mysql -uroot -p -h127.0.0.1 -P3306 --skip-reconnect
session3:执行 kill 命令
mysql> show processlist; +-----+------+---------+---------+--------------------------------+----------------------+ | Id | db | Command | Time | State | Info | +-----+------+---------+---------+--------------------------------+----------------------+ | 6 | NULL | Daemon | 4402073 | Waiting on empty queue | NULL | | 132 | test | Sleep | 524 | | NULL | | 135 | test | Query | 0 | init | show processlist | | 139 | test | Query | 4 | Waiting for table metadata lock| rename table t1 to t2| +-----+------+---------+---------+--------------------------------+----------------------+ 4 rows in set (0.00 sec) mysql> kill 139; Query OK, 0 rows affected (0.00 sec)
session2:
mysql> rename table t1 to t2; ERROR 2013 (HY000): Lost connection to MySQL server during query
可以看到session2的会话连接已经被终止,并且没有自动重新连接,达到了我们想要的效果。
总结
- 通过MySQL客户端登录时,会话重新连接的选项
--reconnect
默认是开启的,如果要禁止重新连接可在登录时添加 --skip-reconnect KILL CONNECTION
与KILL
相同,它在终止连接正在执行的任何语句后,再终止会话连接。KILL QUERY
终止连接当前正在执行的语句,但保持连接本身不变。