为什么MySQL没有负载，但交易却跑不动？

在这里插入图片描述
在MySQL的数据库中，我们有时会发现MySQL数据库明明没有负载，CPU、硬盘、内存和网络等资源都很空闲，但很多SQL都pending在哪儿，MySQL数据库无法处理交易。这是怎么回事呢？

在数据库系统中出现这种情况通常是锁竞争造成的，MySQL数据库更加容易出现这种情况，因为它的存储层和服务层是分开的，我们来看锁竞争在MySQL 5.7和8.0里的表现和解决办法。

5.7版本

查询锁竞争

在 MySQL 8.0 之前，必须SET GLOBAL innodb_status_output_locks=ON后才能在SHOW ENGINE INNODB STATUS中查到数据锁，例如下面这个事务：

mysql> begin;
Query OK, 0 rows affected (0.00 sec)mysql> select * from t1 where intcol1=0 for update;
...
900 rows in set (0.00 sec)

在SHOW ENGINE INNODB STATUS中查到

---TRANSACTION 7827, ACTIVE 11 sec
222 lock struct(s), heap size 24784, 5720 row lock(s)
MySQL thread id 9912, OS thread handle 139967683151616, query id 11123 localhost root
TABLE LOCK table `mysqlslap`.`t1` trx id 7827 lock mode IX
RECORD LOCKS space id 25 page no 37 n bits 1056 index in_1 of table `mysqlslap`.`t1` trx id 7827 lock_mode X...

修改了900行，却锁住了5720行。查询space id为25对应的对象：

mysql> select * from information_schema.INNODB_SYS_DATAFILES where space=25;
+-------+--------------------+
| SPACE | PATH               |
+-------+--------------------+
|    25 | ./mysqlslap/t1.ibd |
+-------+--------------------+
1 row in set (0.00 sec)

在另外一个session里执行

mysql> update t1 set intcol1=1 where intcol1=0;
ERROR 1205 (HY000): Lock wait timeout exceeded; try restarting transaction

查询锁的情况


mysql> select * from information_schema.innodb_lock_waits;
+-------------------+-------------------+-----------------+------------------+
| requesting_trx_id | requested_lock_id | blocking_trx_id | blocking_lock_id |
+-------------------+-------------------+-----------------+------------------+
| 7829              | 7829:25:4:2       | 7827            | 7827:25:4:2      |
+-------------------+-------------------+-----------------+------------------+
1 row in set, 1 warning (0.00 sec)mysql> select * from information_schema.innodb_locks;
+-------------+-------------+-----------+-----------+------------------+-----------------+------------+-----------+----------+----------------+
| lock_id     | lock_trx_id | lock_mode | lock_type | lock_table       | lock_index      | lock_space | lock_page | lock_rec | lock_data      |
+-------------+-------------+-----------+-----------+------------------+-----------------+------------+-----------+----------+----------------+
| 7829:25:4:2 | 7829        | X         | RECORD    | `mysqlslap`.`t1` | GEN_CLUST_INDEX |         25 |         4 |        2 | 0x000000000200 |
| 7827:25:4:2 | 7827        | X         | RECORD    | `mysqlslap`.`t1` | GEN_CLUST_INDEX |         25 |         4 |        2 | 0x000000000200 |
+-------------+-------------+-----------+-----------+------------------+-----------------+------------+-----------+----------+----------------+
2 rows in set, 1 warning (0.00 sec)

查询阻塞的线程：

SELECT b.trx_mysql_thread_id             AS 'blocked_thread_id' ,b.trx_query                      AS 'blocked_sql_text' ,c.trx_mysql_thread_id             AS 'blocker_thread_id',c.trx_query                       AS 'blocker_sql_text',( Unix_timestamp() - Unix_timestamp(c.trx_started) ) AS 'blocked_time' 
FROM   information_schema.innodb_lock_waits a INNER JOIN information_schema.innodb_trx b ON a.requesting_trx_id = b.trx_id INNER JOIN information_schema.innodb_trx c ON a.blocking_trx_id = c.trx_id 
WHERE  ( Unix_timestamp() - Unix_timestamp(c.trx_started) ) > 4; 
+-------------------+-----------------------------------------+-------------------+------------------+--------------+
| blocked_thread_id | blocked_sql_text                        | blocker_thread_id | blocker_sql_text | blocked_time |
+-------------------+-----------------------------------------+-------------------+------------------+--------------+
|              9921 | update t1 set intcol1=1 where intcol1=0 |              9917 | NULL             |          782 |
+-------------------+-----------------------------------------+-------------------+------------------+--------------+
1 row in set, 1 warning (0.00 sec)

根据线程号查询执行的SQL

SELECT a.sql_text, c.id, d.trx_started 
FROM   performance_schema.events_statements_current a join performance_schema.threads b ON a.thread_id = b.thread_id join information_schema.processlist c ON b.processlist_id = c.id join information_schema.innodb_trx d ON c.id = d.trx_mysql_thread_id 
where c.id=9917
ORDER  BY d.trx_started\G
*************************** 1. row ***************************sql_text: select * from t1 where intcol1=0 for updateid: 9917
trx_started: 2023-05-26 13:24:59
1 row in set (0.00 sec)

注意这里查询出的SQL是阻塞事务的最后一条SQL，并不一定是阻塞的源头。

解决锁竞争

解决方法是针对where中的条件增加索引，使MySQL服务层的过滤能在存储层完成，例如

mysql> create index in_1 on t1(intcol1);
Query OK, 0 rows affected (0.04 sec)
Records: 0  Duplicates: 0  Warnings: 0mysql> show create table t1\G
*************************** 1. row ***************************Table: t1
Create Table: CREATE TABLE `t1` (`intcol1` int(32) DEFAULT NULL,`intcol2` int(32) DEFAULT NULL,`charcol1` varchar(128) DEFAULT NULL,`charcol2` varchar(128) DEFAULT NULL,`charcol3` varchar(128) DEFAULT NULL,KEY `in_1` (`intcol1`)
) ENGINE=InnoDB DEFAULT CHARSET=latin1
1 row in set (0.00 sec)

再锁住同样的行

mysql> begin;
Query OK, 0 rows affected (0.00 sec)mysql> select * from t1 where intcol1=0 for update;
...
900 rows in set (0.00 sec)

在SHOW ENGINE INNODB STATUS中查到

---TRANSACTION 7841, ACTIVE 15 sec
155 lock struct(s), heap size 24784, 1801 row lock(s)
MySQL thread id 9917, OS thread handle 139966608627456, query id 11227 localhost root
TABLE LOCK table `mysqlslap`.`t1` trx id 7841 lock mode IX
RECORD LOCKS space id 25 page no 37 n bits 1056 index in_1 of table `mysqlslap`.`t1` trx id 7841 lock_mode X...

被锁住的记录从之前的5720条减少到1801条。

有索引后执行计划也不同，加索引之前的执行计划

mysql> explain select * from t1 where intcol1=0 for update;
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+-------------+
| id | select_type | table | partitions | type | possible_keys | key  | key_len | ref  | rows | filtered | Extra       |
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+-------------+
|  1 | SIMPLE      | t1    | NULL       | ALL  | NULL          | NULL | NULL    | NULL | 6143 |    10.00 | Using where |
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+-------------+
1 row in set, 1 warning (0.00 sec)

加了索引后的执行计划是：

mysql> explain select * from t1 where intcol1=0 for update;
+----+-------------+-------+------------+------+---------------+------+---------+-------+------+----------+-------+
| id | select_type | table | partitions | type | possible_keys | key  | key_len | ref   | rows | filtered | Extra |
+----+-------------+-------+------------+------+---------------+------+---------+-------+------+----------+-------+
|  1 | SIMPLE      | t1    | NULL       | ref  | in_1          | in_1 | 5       | const |  900 |   100.00 | NULL  |
+----+-------------+-------+------------+------+---------------+------+---------+-------+------+----------+-------+
1 row in set, 1 warning (0.00 sec)

对比一下可以发现后者的Extra字段中没有“Using where”，因为过滤再存储层已经完成了。

生产中可以先 select 出主键id，再用主键id 去 update

8.0版本

从 MySQL 8.0 开始，performance_schema.data_locks显示 InnoDB 数据锁。具体参见拙作《MySQL 8.0运维于优化》第18章第3节“优化索引”。