mysql update的优化案例

1.今天接到开发人员   ：mysql update 50w和30w的表关联慢，等了2小时也没有出来

mysql> UPDATE DIM_DEVICE_WIR_E_UTRANCELL_F a, temp_vipcell b SET a.vip_bts=b.vip_bts where a.cgi=b.cgi;
ERROR 1205 (HY000): Lock wait timeout exceeded; try restarting transaction

2.那么我的处理方案

先看关联的数据cgi是否有重复，基数有多大？          言外之意-->适合建索引吗，查询，更新，就是减少io，减少数据扫描的块，避免笛卡尔积等



3.OK

那肯定适合建立索引

4.查看索引

  show index from 2 个表

no rows ;

无索引

5.建立索引

   所以create index idx_xxxx_01 on 表1（cgi）;

卡了5分钟，察。。。

6.不对，再慢也得有响应啊

于是怀疑锁表

mysql> show processlist  默认100行，哇塞，太多

7.show processlist  where info like '*xx表*' ;

 报错

8. show prcoesslist 应该是默认列100个，并且不支持带参数

     show full processlist;   --列出全部进程信息

9.查询某个表被锁，应该是这样

mysql>SELECT id,user, host, time, command,info    FROM information_schema.processlist  WHERE  state IS NOT NULL  and info like '%xxx表%';

mysql>kill id号;

10.我们打算用10000条做测试， 如果不行，我们就1w一次

mysql> create table temp_vipcell_2 as select * from temp_vipcell limit 10000;
Query OK, 10000 rows affected (0.26 sec)
Records: 10000 Duplicates: 0 Warnings: 0

mysql> create index idx_temp_01 on temp_vipcell_2 (cgi);
Query OK, 0 rows affected (0.04 sec)
Records: 0  Duplicates: 0  Warnings: 0

mysql> show index from temp_vipcell_2;
+----------------+------------+-------------+--------------+-------------+-----------+-------------+----------+--------+------+------------+---------+---------------+
| Table          | Non_unique | Key_name    | Seq_in_index | Column_name | Collation | Cardinality | Sub_part | Packed | Null | Index_type | Comment | Index_comment |
+----------------+------------+-------------+--------------+-------------+-----------+-------------+----------+--------+------+------------+---------+---------------+
| temp_vipcell_2 |          1 | idx_temp_01 |            1 | cgi         | A         |        9985 |     NULL | NULL   |      | BTREE      |         |               |
+----------------+------------+-------------+--------------+-------------+-----------+-------------+----------+--------+------+------------+---------+---------------+
1 row in set (0.00 sec)

11.更新试试，4秒搞定50w+1w的更新，按理推算，34w应该是34秒搞定。

mysql> UPDATE DIM_DEVICE_WIR_E_UTRANCELL_F a, temp_vipcell_2 b SET a.vip_bts=b.vip_bts where a.cgi=b.cgi;
Query OK, 0 rows affected (4.13 sec)
Rows matched: 10000 Changed: 0 Warnings: 0

12.我们在主表建立索引，果然和推算一样34秒
mysql> create index idx_dim_device_wir_e_001 on DIM_DEVICE_W (cgi);
Query OK, 0 rows affected (2.26 sec)
Records: 0 Duplicates: 0 Warnings: 0

mysql> UPDATE DIM_DEVICE_W a, temp_vipcell_2 b SET a.vip_bts=b.vip_bts where a.cgi=b.cgi;
Query OK, 0 rows affected (4.13 sec)
Rows matched: 10000 Changed: 0 Warnings: 0

mysql> UPDATE DIM_DEVICE_W a, temp_vipcell b SET a.vip_bts=b.vip_bts where a.cgi=b.cgi;
Query OK, 331262 rows affected (34.84 sec)
Rows matched: 341262 Changed: 331262 Warnings: 0

13.算了不折腾，一晚一次，34秒够了。

     别想歪了，兄弟们。

数据库的东西是越快越好， 而某些东西是越慢越好