问题描述
研发同事反馈某应用执行较慢,对应SQL为:
UPDATEbs_serial_tracSET SERIAL_IS_LOCK = 0,
LOCK_VALUE= '',
UPDATE_USER= 'transSys'
WHERE GOODS_NO = '4418095740626'
AND(
PARENT_CODE= 'F9G7S19722001835'
OR SERIAL = 'F9G7S19722001835');
表bs_serial_trac上索引情况为:
PRIMARY KEY(`ID`),UNIQUE KEY`idx_complex_serial_goodsNo` (`SERIAL`,`GOODS_NO`),KEY`idx_update_time` (`UPDATE_TIME`),KEY`idx_serial_goodsNo` (`GOODS_NO`),KEY`idx_parent_code` (`PARENT_CODE`),KEY `idx_lock_value` (`LOCK_VALUE`)
由于使用OR条件,查询只能基于条件GOODS_NO = '4418095740626' 进行数据查找,其执行计划为:
*************************** 1. row ***************************id:1select_type:UPDATE
table: bs_serial_trac
partitions:NULLtype: range
possible_keys: idx_serial_goodsNokey: idx_serial_goodsNo
key_len:93ref: const
rows:404920filtered:100.00Extra: Usingwhere
1 row in set (0.00 sec)
由于GOODS_NO列选择性较差,满足条件的记录较多,导致查询性能较差:
SELECT COUNT(1)FROMbs_serial_tracWHERE GOODS_NO = '4418095740626';+----------+
| COUNT(1) |
+----------+
| 215447 |
+----------+
解决步骤
问题很明显,由于OR语句导致索引无法正常使用,将OR调整为UNION ALL,调整后的SQL语句为:
SELECT *
FROMbs_serial_tracWHERE GOODS_NO = '4418095740626'
AND PARENT_CODE = 'F9G7S19722007485'
UNION ALL
SELECT *
FROMbs_serial_tracWHERE GOODS_NO = '4418095740626'
AND SERIAL = 'F9G7S19722007485'
AND PARENT_CODE <> 'F9G7S19722007485'
查询速度很快,由之前的800ms优化到10ms以下,看起来很完美,但是查询结果没数据。。。
简单定位下,发现PARENT_CODE列类型为varchar(50) DEFAULT '',PARENT_CODE列值为NULL,做不等于判断时存在问题:
SELECT
NOT(PARENT_CODE <=> 'F9G7S19722007485') ASC1,NOT(PARENT_CODE = 'F9G7S19722007485') ASC2,
PARENT_CODE<> 'F9G7S19722007485' ASC3FROMbs_serial_tracWHERE GOODS_NO = '4418095740626'
AND SERIAL = 'F9G7S19722007485';+----+------+------+
| C1 | C2 | C3 |
+----+------+------+
| 1 | NULL | NULL |
+----+------+------+
因此改写为UNION ALL语句时需要改写为:
SELECT *
FROMbs_serial_tracWHERE GOODS_NO = '4418095740626'
AND PARENT_CODE = 'F9G7S19722007485'
UNION ALL
SELECT *
FROMbs_serial_tracWHERE GOODS_NO = '4418095740626'
AND SERIAL = 'F9G7S19722007485'
AND NOT(PARENT_CODE <=> 'F9G7S19722007485')
或改写为:
SELECT *
FROMbs_serial_tracWHERE GOODS_NO = '4418095740626'
AND PARENT_CODE = 'F9G7S19722007485'
UNION ALL
SELECT *
FROMbs_serial_tracWHERE GOODS_NO = '4418095740626'
AND SERIAL = 'F9G7S19722007485'
AND (PARENT_CODE <> 'F9G7S19722007485' OR PARENT_CODE IS NULL)
由于在UNION ALL的第二部分查询中,PARENT_CODE不用于索引查找,只用于数据过滤,因此两种方式都不会影响查询性能。
确认使用UNION ALL性能满足需求后,将UPDATE操作改写为:
UPDATEbs_serial_tracSET SERIAL_IS_LOCK = 0,
LOCK_VALUE= '',
UPDATE_USER= 'transSys'
WHERE ID IN(SELECT ID FROM(SELECTIDFROMbs_serial_tracWHERE GOODS_NO = '4418095740626'
AND PARENT_CODE = 'F9G7S19722007485'
UNION ALL
SELECTIDFROMbs_serial_tracWHERE GOODS_NO = '4418095740626'
AND SERIAL = 'F9G7S19722007485'
AND (PARENT_CODE <> 'F9G7S19722007485' OR PARENT_CODE IS NULL)
)AST1
)
其对于执行计划为:
*************************** 1. row ***************************id:1select_type:UPDATE
table: bs_serial_trac
partitions:NULLtype:indexpossible_keys:NULL
key: PRIMARYkey_len:8ref:NULLrows:13270473filtered:100.00Extra: Usingwhere
*************************** 2. row ***************************id:2select_type: DEPENDENT SUBQUERYtable: partitions:NULLtype: index_subquery
possible_keys:
key: key_len:8ref: func
rows:2filtered:100.00Extra: Usingindex
*************************** 3. row ***************************id:3select_type: DERIVEDtable: bs_serial_trac
partitions:NULLtype: ref
possible_keys: idx_serial_goodsNo,idx_parent_codekey: idx_parent_code
key_len:153ref: const
rows:1filtered:5.00Extra: Usingwhere
*************************** 4. row ***************************id:4select_type:UNION
table: bs_serial_trac
partitions:NULLtype: const
possible_keys: idx_complex_serial_goodsNo,idx_serial_goodsNo,idx_parent_codekey: idx_complex_serial_goodsNo
key_len:695ref: const,const
rows:1filtered:100.00Extra:NULL
4 rows in set (0.00 sec)
虽然按照主键ID去更新,但是由于使用IN语句,仍导致查询走全表扫描,性能极差,需要将IN查询转换成INNER JOIN:
UPDATEbs_serial_tracAST2INNER JOIN(SELECTIDFROMbs_serial_tracWHERE GOODS_NO = '4418095740626'
AND PARENT_CODE = 'F9G7S19722007485'
UNION ALL
SELECTIDFROMbs_serial_tracWHERE GOODS_NO = '4418095740626'
AND SERIAL = 'F9G7S19722007485'
AND (PARENT_CODE <> 'F9G7S19722007485' OR PARENT_CODE IS NULL)
)AST1ON T1.ID=T2.IDSETT2.SERIAL_IS_LOCK= 0,
T2.LOCK_VALUE= '',
T2.UPDATE_USER= 'transSys'
修改后执行计划为:
*************************** 1. row ***************************id:1select_type:PRIMARY
table: partitions:NULLtype:ALLpossible_keys:NULL
key: NULLkey_len:NULLref:NULLrows:3filtered:100.00Extra:NULL
*************************** 2. row ***************************id:1select_type:UPDATE
table: T2
partitions:NULLtype: eq_ref
possible_keys:PRIMARY
key: PRIMARYkey_len:8ref: T1.ID
rows:1filtered:100.00Extra:NULL
*************************** 3. row ***************************id:2select_type: DERIVEDtable: bs_serial_trac
partitions:NULLtype: ref
possible_keys: idx_serial_goodsNo,idx_parent_codekey: idx_parent_code
key_len:153ref: const
rows:1filtered:5.00Extra: Usingwhere
*************************** 4. row ***************************id:3select_type:UNION
table: bs_serial_trac
partitions:NULLtype: const
possible_keys: idx_complex_serial_goodsNo,idx_serial_goodsNo,idx_parent_codekey: idx_complex_serial_goodsNo
key_len:695ref: const,const
rows:1filtered:100.00Extra:NULL
4 rows in set (0.00 sec)
调整能正常按照主键去操作,性能有保障。
<=>操作符
<=> : NULL-safe equal. This operator performs an equality comparison like the = operator, but returns 1 rather than NULL if both operands are NULL, and 0 rather than NULL if one operand is NULL.
参考: https://dev.mysql.com/doc/refman/8.0/en/comparison-operators.html#operator_equal-to