我有一个存储过程，它将运行20秒并占用大量CPU，我在慢查询日志中发现了它，该存储过程始终存在，将运行15到30秒。我相信此存储过程导致了较高的CPU使用率。因此，我试图使用explain获取存储过程的查询计划。而且我对如何改进此存储过程并不了解。请分享一些有关如何改进存储过程的信息。
CREATE DEFINER=`root`@`localhost` PROCEDURE `SelectUsage4`(
IN p_ids MEDIUMTEXT
, IN p_locationIDs MEDIUMTEXT
, IN p_indicatorIDs MEDIUMTEXT
, IN p_fromDate date
, IN p_toDate date
, IN p_yearly tinyint(4)
, IN p_monthly tinyint(4)
, IN p_halfYear1 tinyint(4)
, IN p_halfYear2 tinyint(4)
, IN p_fourMonths1 tinyint(4)
, IN p_fourMonths2 tinyint(4)
, IN p_fourMonths3 tinyint(4)
, IN p_q1 tinyint(4)
, IN p_q2 tinyint(4)
, IN p_q3 tinyint(4)
, IN p_q4 tinyint(4)
, IN p_biMonthly1 tinyint(4)
, IN p_biMonthly2 tinyint(4)
, IN p_biMonthly3 tinyint(4)
, IN p_biMonthly4 tinyint(4)
, IN p_biMonthly5 tinyint(4)
, IN p_biMonthly6 tinyint(4)
, IN p_approvalStatus int(11)
, IN p_language nvarchar(10)
)
BEGIN
select
case when (select count(lbl.id) from `labels` as lbl where lbl.ObjectID = l.id and lbl.ObjectName = 'locations' and lbl.ColumnName = 'LocationName' and lbl.LanguageCode = p_language) > 0 then
(select content from `labels` as lbl where lbl.ObjectID = l.id and lbl.ObjectName = 'locations' and lbl.ColumnName = 'LocationName' and lbl.LanguageCode = p_language limit 1)
else
l.LocationName
end as LocationName
, l.ParentID as LocationParentID
, l.Active as LocationActive
, l.RegionID
, case when (select count(lbl.id) from `labels` as lbl where lbl.ObjectID = i.id and lbl.ObjectName = 'indicators' and lbl.ColumnName = 'IndicatorName' and lbl.LanguageCode = p_language) > 0 then
(select content from `labels` as lbl where lbl.ObjectID = i.id and lbl.ObjectName = 'indicators' and lbl.ColumnName = 'IndicatorName' and lbl.LanguageCode = p_language limit 1)
else
i.IndicatorName
end as IndicatorName
, i.ParentID as IndicatorParentID
, i.Unit
, i.DecimalPlaces
, i.Active as IndicatorActive
, i.IndicatorType
, u.*
from
`usage` as u
left join `locations` as l on u.LocationID = l.id
left join `Indicators` as i on u.IndicatorID = i.id
where
u.IsDeleted = 0
and (
(p_fromDate is null and p_toDate is null)
or
(
p_fromDate is not null and p_toDate is not null
and
DATE(CONCAT(convert(u.`Year`, char(4)), '-', convert(u.`Month`, char(2)), '-1')) between p_fromDate and p_toDate
)
or
(
p_fromDate is not null and p_toDate is not null
and
u.`Month` is null
and
u.`Year` between Year(p_fromDate) and Year(p_toDate)
)
)
and (p_yearly is null or (p_yearly is not null and p_yearly = 1 and u.`Month` is null) or (p_yearly is not null and p_yearly = 0 and u.`Month` is not null))
and (p_monthly is null or (p_monthly is not null and p_monthly = 1 and u.`Month` is not null))
and (p_ids is null or FIND_IN_SET(u.id, p_ids))
and (p_locationIDs is null or FIND_IN_SET(u.LocationID, p_locationIDs))
and (p_indicatorIDs is null or FIND_IN_SET(u.IndicatorID, p_indicatorIDs))
and
(
(p_halfYear1 is null or u.HalfYear1 = p_halfYear1)
or (p_halfYear2 is null or u.HalfYear2 = p_halfYear2)
)
and
(
(p_fourMonths1 is null or u.FourMonths1 = p_fourMonths1)
or (p_fourMonths2 is null or u.FourMonths2 = p_fourMonths2)
or (p_fourMonths3 is null or u.FourMonths3 = p_fourMonths3)
)
and
(
(p_q1 is null or u.Q1 = p_q1)
or (p_q2 is null or u.Q2 = p_q2)
or (p_q3 is null or u.Q3 = p_q3)
or (p_q4 is null or u.Q4 = p_q4)
)
and
(
(p_biMonthly1 is null or u.BiMonthly1 = p_biMonthly1)
or (p_biMonthly2 is null or u.BiMonthly2 = p_biMonthly2)
or (p_biMonthly3 is null or u.BiMonthly3 = p_biMonthly3)
or (p_biMonthly4 is null or u.BiMonthly4 = p_biMonthly4)
or (p_biMonthly5 is null or u.BiMonthly5 = p_biMonthly5)
or (p_biMonthly6 is null or u.BiMonthly6 = p_biMonthly6)
)
and (
p_approvalStatus is null
or
(
select ara.ApprovalStatus
from `tasks_details` as t
inner join `approval_request_tasks` as art on t.TaskID = art.TaskID
inner join `approval_request_approvers` as ara on art.ApprovalRequestID = ara.ApprovalRequestID
where
t.IsDeleted = 0
and
t.ObjectID = u.id
and t.ObjectType = 'Usage'
order by
ara.ModifiedDate desc limit 1
) = p_approvalStatus
)
order by
i.IndicatorName, l.LocationName
;
END
存储过程的解释计划如下：
+----+--------------------+-------+--------+----------------+---------+---------+-----------------------------------+-------+--------+----------------------------------------------+
| id | select_type | TABLE | TYPE | possible_keys | KEY | key_len | ref | ROWS | Extra | |
+----+--------------------+-------+--------+----------------+---------+---------+-----------------------------------+-------+--------+----------------------------------------------+
| 1 | PRIMARY | u | ALL | (NULL) | (NULL) | (NULL) | (NULL) | 75095 | 10.00 | USING WHERE; USING TEMPORARY; USING filesort |
| 1 | PRIMARY | l | eq_ref | PRIMARY,Index1 | PRIMARY | 4 | pg.u.LocationID | 1 | 100.00 | |
| 1 | PRIMARY | i | eq_ref | PRIMARY | PRIMARY | 4 | pg.u.IndicatorID | 1 | 100.00 | |
| 6 | DEPENDENT SUBQUERY | ara | INDEX | Index1 | Index1 | 28 | (NULL) | 1384 | 100.00 | USING INDEX; USING filesort |
| 6 | DEPENDENT SUBQUERY | art | ref | Index1 | Index1 | 4 | pg.ara.ApprovalRequestID | 1 | 100.00 | USING INDEX |
| 6 | DEPENDENT SUBQUERY | t | ref | Index1 | Index1 | 161 | pg.art.TaskID,pg.u.id,const,const | 1 | 100.00 | USING INDEX |
| 5 | DEPENDENT SUBQUERY | lbl | ref | Index1 | Index1 | 644 | const,pg.i.id,const,const | 1 | 100.00 | |
| 4 | DEPENDENT SUBQUERY | lbl | ref | Index1 | Index1 | 644 | const,pg.i.id,const,const | 1 | 100.00 | USING INDEX |
| 3 | DEPENDENT SUBQUERY | lbl | ref | Index1 | Index1 | 644 | const,pg.l.id,const,const | 1 | 100.00 | |
| 2 | DEPENDENT SUBQUERY | lbl | ref | Index1 | Index1 | 644 | const,pg.l.id,const,const | 1 | 100.00 | USING INDEX |
+----+--------------------+-------+--------+----------------+---------+---------+-----------------------------------+-------+--------+----------------------------------------------+
usage的表结构
CREATE TABLE `usage` (
`id` int(11) NOT NULL AUTO_INCREMENT,
`LocationID` int(11) NOT NULL,
`IndicatorID` int(11) NOT NULL,
`Year` int(11) DEFAULT NULL,
`Month` int(11) DEFAULT NULL,
`HalfYear1` tinyint(4) DEFAULT NULL,
`HalfYear2` tinyint(4) DEFAULT NULL,
`FourMonths1` tinyint(4) DEFAULT NULL,
`FourMonths2` tinyint(4) DEFAULT NULL,
`FourMonths3` tinyint(4) DEFAULT NULL,
`Q1` tinyint(4) DEFAULT NULL,
`Q2` tinyint(4) DEFAULT NULL,
`Q3` tinyint(4) DEFAULT NULL,
`Q4` tinyint(4) DEFAULT NULL,
`BiMonthly1` tinyint(4) DEFAULT NULL,
`BiMonthly2` tinyint(4) DEFAULT NULL,
`BiMonthly3` tinyint(4) DEFAULT NULL,
`BiMonthly4` tinyint(4) DEFAULT NULL,
`BiMonthly5` tinyint(4) DEFAULT NULL,
`BiMonthly6` tinyint(4) DEFAULT NULL,
`DateOfUsage` date DEFAULT NULL,
`Price` decimal(24,10) DEFAULT NULL,
`PriceUnit` int(11) DEFAULT NULL,
`ExchangeRate` decimal(24,10) DEFAULT NULL,
`Value` decimal(24,10) DEFAULT NULL,
`ValueUnit` varchar(100) CHARACTER SET utf8 DEFAULT NULL,
`Remarks` varchar(1000) CHARACTER SET utf8 DEFAULT NULL,
`CreatedBy` varchar(100) CHARACTER SET utf8 DEFAULT NULL,
`CreatedDate` datetime DEFAULT NULL,
`ModifiedBy` varchar(100) CHARACTER SET utf8 DEFAULT NULL,
`ModifiedDate` datetime DEFAULT NULL,
`IsDeleted` tinyint(1) NOT NULL DEFAULT '0',
`IsHeatRecovery` tinyint(1) NOT NULL DEFAULT '0',
PRIMARY KEY (`id`),
KEY `Index1` (`LocationID`,`IndicatorID`,`Year`,`Month`,`HalfYear1`,`HalfYear2`,`FourMonths1`,`FourMonths2`,`FourMonths3`,`Q1`,`Q2`,`Q3`,`Q4`,`IsDeleted`,`CreatedDate`,`ModifiedDate`),
KEY `Index2` (`LocationID`,`IndicatorID`,`Year`,`BiMonthly1`,`BiMonthly2`,`BiMonthly3`,`BiMonthly4`,`BiMonthly5`,`BiMonthly6`,`CreatedDate`,`ModifiedDate`,`IsDeleted`),
KEY `Index3` (`LocationID`,`IndicatorID`,`DateOfUsage`,`IsDeleted`)
) ENGINE=InnoDB AUTO_INCREMENT=79273 DEFAULT CHARSET=utf8 COLLATE=utf8_unicode_ci;
locations的表结构
CREATE TABLE `locations` (
`id` int(11) NOT NULL AUTO_INCREMENT,
`ParentID` int(11) DEFAULT NULL,
`LocationName` varchar(100) CHARACTER SET utf8 NOT NULL,
`Active` tinyint(1) NOT NULL DEFAULT '1',
`CreatedBy` varchar(100) CHARACTER SET utf8 DEFAULT NULL,
`CreatedDate` datetime DEFAULT NULL,
`ModifiedBy` varchar(100) CHARACTER SET utf8 DEFAULT NULL,
`ModifiedDate` datetime DEFAULT NULL,
`IsDeleted` tinyint(1) NOT NULL DEFAULT '0',
`RegionID` int(11) DEFAULT NULL,
`IsRegionComingFromParent` tinyint(1) DEFAULT NULL,
PRIMARY KEY (`id`),
KEY `Index1` (`id`,`ParentID`,`Active`,`IsDeleted`,`RegionID`)
) ENGINE=InnoDB AUTO_INCREMENT=445 DEFAULT CHARSET=utf8 COLLATE=utf8_unicode_ci;
indicators的表结构
CREATE TABLE `indicators` (
`id` int(11) NOT NULL AUTO_INCREMENT,
`ParentID` int(11) DEFAULT NULL,
`IndicatorName` varchar(255) CHARACTER SET utf8 NOT NULL,
`IndicatorType` int(11) DEFAULT NULL,
`Unit` varchar(100) CHARACTER SET utf8 NOT NULL,
`DecimalPlaces` int(11) DEFAULT NULL,
`Active` tinyint(4) NOT NULL DEFAULT '1',
`CreatedBy` varchar(100) CHARACTER SET utf8 DEFAULT NULL,
`CreatedDate` datetime DEFAULT NULL,
`ModifiedBy` varchar(100) CHARACTER SET utf8 DEFAULT NULL,
`ModifiedDate` datetime DEFAULT NULL,
`IsDeleted` tinyint(4) NOT NULL DEFAULT '0',
`SyncID` varchar(50) COLLATE utf8_unicode_ci DEFAULT NULL,
`LastSyncDate` datetime DEFAULT NULL,
`FormulaSummary` int(11) DEFAULT NULL,
`IndicatorCategory` int(11) DEFAULT NULL,
`BreakSync` tinyint(4) DEFAULT NULL,
`IsInteger` tinyint(1) DEFAULT '0',
`ActiveForReporting` tinyint(1) DEFAULT '1',
`BaselineYear` int(11) DEFAULT NULL,
`Ceiling` decimal(24,10) DEFAULT NULL,
`Floor` decimal(24,10) DEFAULT NULL,
`BreakSyncForUnit` tinyint(1) DEFAULT NULL,
PRIMARY KEY (`id`),
KEY `Index1` (`ParentID`,`IndicatorType`,`IsDeleted`,`Active`)
) ENGINE=InnoDB AUTO_INCREMENT=10396 DEFAULT CHARSET=utf8 COLLATE=utf8_unicode_ci;
[当我调用以下存储过程时，Rows_examined为1133668，它运行了7秒钟，我认为这导致了CPU的繁重计算。
CALL `SelectUsage4`(NULL, '65,92,207,93,94,95,66,67,372,57,96,68,70,69,71,378,379,380,97,98,370,250,99,196,100,197,208,63,183,72,51,74,75,101,73,64,395,251,102,103,104,252,106,209,105,210,429,257,107,258,91,46,267,108,211,259,253,261,254,260,255,109,110,79,80,81,437,111,112,427,428,409,113,413,412,425,28,41,249,114,212,333,335,366,334,368,367,318,391,406,43,115,213,263,116,214,215,117,216,118,217,119,120,121,122,124,218,123,125,126,127,128,129,130,131,219,56,220,221,198,132,133,48,134,222,223,224,135,136,137,225,50,138,271,331,417,414,363,226,139,227,315,140,141,229,199,228,142,143,144,230,146,231,147,78,148,149,316,150,151,264,45,268,232,233,152,269,153,154,200,155,443,234,201,156,157,76,265,49,342,235,236,158,159,160,161,237,238,162,77,163,394,390,439,442,389,388,415,416,418,419,420,387,424,410,421,369,426,239,164,240,272,314,202,241,266,273,165,166,167,203,242,47,270,168,444,169,204,86,328,170,274,243,171,87,374,375,376,377,373,244,275,172,205,371,385,386,173,256,42,174,175,176,245,177,178,277,287,279,288,286,291,317,280,289,284,281,282,295,290,283,292,293,294,285,278,179,246,206,180,276,247,181,88,52,89,182,248,184,185,186,187,188,189,190,191,192,193,194,195,308,364,365,300,304,301,302,303,345,306,323,349,400,350,401,320,440,402,392,324,403,321,441,351,362,361,360,393,322,346,399,325,347,348,299,423,307,381,353,397,352,398,382,358,354,357,355,356,359,145,434,435,438,436,297,430,431,432,433,311,83,85,298,58,40,60,319,59,329,296,61,422,305,327,384,396,383,326,330,62,310,309,312,407', '10065', '2017-12-01', '2017-12-31', NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, 202, 'en-US');
更改
WHEN ( SELECT COUNT(x) ... ) > 0
to
WHEN EXISTS( SELECT 1 ... )
此外，更改
CASE WHEN EXISTS (...)
THEN (SELECT ... LIMIT 1)
ELSE LocationName
END
简单
IFNULL( (SELECT ... LIMIT 1), LocationName)
我没看到CREATE TABLE labels；需要
INDEX(ObjectID, ObjectName, ColumnName, LanguageCode) -- in any order
我不确定convert(u. Month , char(2))是否可以正常工作1位数月。
u.Q1中可以是什么值？我可能对ANDs的ORs这么长的混乱进行了优化。
此：
and (
p_approvalStatus is null
or
( big select ) = p_approvalStatus
)
看起来像查询的很大一部分（请参见EXPLAIN）。建议通过执行以下操作来加快通话速度：
IF (p_approvalStatus IS NULL)
THEN
the current query, but without that `and (...)` -- I expect this to be much faster
ELSE
the current query, except for the check for NULL
END IF
也有可能进一步重写ELSE部分。