编程

故障分析 | 从慢日志问题看 MySQL 半一致性读的应用场景

763 2023-07-16 15:48:00

背景

某系统执行更新操作发现很慢,发现有大量慢日志,其中 Lock time 时间占比很高,MySQL 版本为 5.7.25,隔离级别为 RR。

分析

查看表结构以及 UPDATE 语句的执行计划:

mysql> show create table test;


+-------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| Table | Create Table |
+-------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| test | CREATE TABLE `test` (
`id` int(11) NOT NULL AUTO_INCREMENT,
`name` varchar(30) COLLATE utf8mb4_bin DEFAULT NULL,
PRIMARY KEY (`id`)
) ENGINE=InnoDB AUTO_INCREMENT=2621401 DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_bin |
+-------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
1 row in set (0.00 sec)


mysql> explain update test set name ='test' where name='a';
+----+-------------+-------+------------+-------+---------------+---------+---------+------+---------+----------+-------------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+-------------+-------+------------+-------+---------------+---------+---------+------+---------+----------+-------------+
| 1  | UPDATE      | test  | NULL       | index | NULL   | PRIMARY | 4 | NULL | 2355988 | 100.00 | Using where |
+----+-------------+-------+------------+-------+---------------+---------+---------+------+---------+----------+-------------+
1 row in set (0.00 sec)

通过执行计划发现,该 SQL 是走的主键全索引扫描,并且对于 name 列未加索引,当多个事务同时执行时,就会观察到有阻塞出现。

若 name 列的重复值不多,那么可以对 name 列添加索引即可解决该问题。因为 InnoDB 的行锁机制是基于索引列来实现的,如果 UPDATE 语句能使用到 name 列的索引,那么就不会产生阻塞,导致业务卡顿。

但若是 name 列的值的区分度很低,就会导致 SQL 不会走 name 列的索引,示例如下:

先添加索引

mysql> alter table test add index tt(name);
Query OK, 0 rows affected (2.74 sec)
Records: 0 Duplicates: 0 Warnings: 0

然后查看执行计划,发现可能用到的索引有 tt,但是实际情况依然走的主键全索引扫描。

mysql> explain update test set name ='test' where name='a';
+----+-------------+-------+------------+-------+---------------+---------+---------+------+---------+----------+-------------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+-------------+-------+------------+-------+---------------+---------+---------+------+---------+----------+-------------+
| 1 | UPDATE | test | NULL | index | tt | PRIMARY | 4 | NULL | 2355988 | 100.00 | Using where |
+----+-------------+-------+------------+-------+---------------+---------+---------+------+---------+----------+-------------+
1 row in set (0.00 sec)

因为 MySQL 的优化器是基于代价来评估的,我们可以通过 optimizer trace 来观察。

mysql> show variables like 'optimizer_trace';
+-----------------+--------------------------+
| Variable_name | Value |
+-----------------+--------------------------+
| optimizer_trace | enabled=off,one_line=off |
+-----------------+--------------------------+
1 row in set (0.01 sec)

可以看到值为 enabled=off,表明这个功能默认是关闭的。

如果想打开这个功能,必须⾸先把 enabled 的值改为 on

mysql> set optimizer_trace="enabled=on";
Query OK, 0 rows affected (0.00 sec)

然后执行该 SQL,查看详细的信息,这里我们主要关注的是 PREPARE 阶段的成本计算。

mysql> update test set name ='test' where name='a';
Query OK, 262144 rows affected (5.97 sec)
Rows matched: 262144 Changed: 262144 Warnings: 0

mysql> SELECT * FROM information_schema.OPTIMIZER_TRACE\G 

详细结果如下。

mysql> SELECT * FROM information_schema.OPTIMIZER_TRACE\G
*************************** 1. row ***************************
QUERY: update test set name ='test' where name='a'
TRACE: {
"steps": [
{
"substitute_generated_columns": {
}
},
{
"condition_processing": {
"condition": "WHERE",
"original_condition": "(`test`.`name` = 'a')",
"steps": [
{
"transformation": "equality_propagation",
"resulting_condition": "multiple equal('a', `test`.`name`)"
},
{
"transformation": "constant_propagation",
"resulting_condition": "multiple equal('a', `test`.`name`)"
},
{
"transformation": "trivial_condition_removal",
"resulting_condition": "multiple equal('a', `test`.`name`)"
}
]
}
},
{
"table": "`test`",
"range_analysis": {
"table_scan": {
"rows": 2355988,
"cost": 475206
},
"potential_range_indexes": [
{
"index": "PRIMARY",
"usable": true,
"key_parts": [
"id"
]
},
{
"index": "tt",
"usable": true,
"key_parts": [
"name",
"id"
]
}
],
"setup_range_conditions": [
],
"group_index_range": {
"chosen": false,
"cause": "no_join"
},
"analyzing_range_alternatives": {
"range_scan_alternatives": [
{
"index": "tt",
"ranges": [
"0x0100610000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000 <= name <= 0x0100610000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000"
],
"index_dives_for_eq_ranges": true,
"rowid_ordered": true,
"using_mrr": false,
"index_only": false,
"rows": 553720,
"cost": 664465,
"chosen": false,
"cause": "cost"
}
],
"analyzing_roworder_intersect": {
"usable": false,
"cause": "too_few_roworder_scans"
}
}
}
}
]
}
MISSING_BYTES_BEYOND_MAX_MEM_SIZE: 0
INSUFFICIENT_PRIVILEGES: 0
1 row in set (0.00 sec)

可以发现执行全表扫描的成本为 475206,走索引 tt 的成本为 664465,所以 MySQL 选择了全表扫描

那么如果是这种情况改怎么处理呢?

如果 InnoDB 隔离级别是 RR,数据库层面没有太好的方式,推荐应用端进行改造。

如果数据库隔离级别可以更改,那么可以改为 RC 来解决阻塞的问题。因为 RC 模式下支持半一致性读。

什么是半一致性读呢?

简单来说就是当要对行进行加锁时,会多一步判断该行是不是真的需要上锁。比如全表扫描更新的时候,我们只需要更新 WHERE 匹配到的行,如果是没有半一致性读就会把所有数据进行加锁,但是有了半一致性读,那么会判断是否满足 WHERE 条件,若不满足则不会加锁(提前释放锁)。

那么对于区分度低的字段就可以使用半一致性读特性来优化,这样更新不同的值就不会互相等待,导致业务卡顿。

结论

行锁机制是基于索引列实现的,若没有使用到索引,则会进行全表扫描。

半一致性读是基于 RC 隔离级别的优化,可以减少锁冲突以及锁等待,提升并发。