mysql 的事件之 Waiting for table metadata lock详解数据库

os:debian 8.2
mysql:5.5.49

nodeb 为 nodea 的一个slave，在nodea上删除一个无效的索引，发现nodeb上出现长时间的 Waiting for table metadata lock。
分析的过程如下：

环境

# lsb_release -a 
No LSB modules are available. 
Distributor ID: Debian 
Description:    Debian GNU/Linux 8.2 (jessie) 
Release:    8.2 
Codename:   jessie 
 
# mysql 
 
[email protected]> select version(); 
+---------------------+ 
| version()           | 
+---------------------+ 
| 5.5.49-0+deb8u1-log | 
+---------------------+ 
1 row in set (0.00 sec) 

nodea 操作

root@localhost> drop index idx_abc_x1 on abc;

nodeb 操作

[email protected]> show processlist; 
+---------+-------------+-------------------+--------+---------+----------+----------------------------------+------------------------------------------------------------------------------------------------------+ 
| Id      | User        | Host              | db     | Command | Time     | State                            | Info                                                                                                 | 
+---------+-------------+-------------------+--------+---------+----------+----------------------------------+------------------------------------------------------------------------------------------------------+ 
|      48 | system user |                   | NULL   | Connect | 10446048 | Waiting for master to send event | NULL                                                                                                 | 
|      49 | system user |                   | EMdata | Connect |     1543 | Waiting for table metadata lock  | DROP INDEX IDX_ABC_X1 on ABC  
~~~ 
~~~ 
~~~ 
| 2427727 | emsel       | 10.10.2.51:59835  | EMdata | Query   |      134 | Waiting for table metadata lock  | select * from abc 

查询锁信息

[email protected]> SELECT * FROM information_schema.innodb_lock_waits; 
[email protected]> SELECT * FROM information_schema.innodb_locks; 
[email protected]> SELECT * FROM information_schema.innodb_trx;

因为slave都是select操作，和开发人员沟通后可以使用 kill trx_mysql_thread_id 杀掉进程。

所以尽量避免在高并发的时候做DDL操作。

lock_wait_timeout

还有一个参数 lock_wait_timeout 用来控制metadata lock的超时（单位为秒），允许的值范围为1到31536000（1年）。 默认值为31536000。

[email protected]> show global variables like 'lock_wait_timeout'; 
+-------------------+----------+ 
| Variable_name     | Value    | 
+-------------------+----------+ 
| lock_wait_timeout | 31536000 | 
+-------------------+----------+ 
1 row in set (0.00 sec) 

参考：
IT虾米网