前几天,在所有数据库服务器部署了监控磁盘空间的存储过程和作业后(MS SQL 监控磁盘空间告警),今天突然收到了两封告警邮件,好吧,存储规划是一方面,但是,是不是要分析一下是什么原因造成磁盘空间不足的呢?会不会是因为突然暴增的日志文件,抑或是系统业务猛增导致数据量暴增,还是历史数据累计原因....分析总得有数据来支撑吧,但是现在只有那些数据文件的当前大小信息,没有数据文件的历史增长变化信息,所以,今天就想实现这么一个功能,每天(频率可以调整)去收集一下数据文件的信息,放到一个表里面,这样方便我们分析数据文件的增长演变例程,甚至你可以将数据文件的增长幅度和业务变化关联起来分析....
那么接下来就是我的设计思路和实现代码,目前只是简单实现,以后将继续优化,丰富一些功能。
首先我们创建一个表DiskCapacityHistory,用来保存数据库文件的历史增长变化信息:
USE msdb; GO IF EXISTS (SELECT 1 FROM dbo.sysobjects WHERE id = OBJECT_ID(N‘‘) AND xtype=‘U‘) DROP TABLE DiskCapacityHistory; GO CREATE TABLE dbo.DiskCapacityHistory ( [Date_CD] INT , [DataBaseID] INT , [FileID] INT , [DataBaseName] sysname , [LogicalName] VARCHAR(32) , [FileTypeDesc] NVARCHAR(60) , [PhysicalName] NVARCHAR(260) , [StateDesc] NVARCHAR(60) , [MaxSize] NVARCHAR(32) , [GrowthType] NVARCHAR(8) , [IsReadOnly] INT , [IsPercentGrowth] SMALLINT , [Size] FLOAT , [Growth_MOM_RAT] FLOAT , [Growth_YOY_RAT] FLOAT , CONSTRAINT PK_DiskCapacityHistory PRIMARY KEY(Date_CD, DataBaseID, FileID) );
接下来,我们创建存储过程,负责来收集、统计这些数据库的文件的相关信息。关于环比/同比,正常情况一般是:
环比: (指标当前值 - 指标值(上个月同一天))/ 指标值(上个月同一天) 。
同比: (指标当前值 - 指标值(去年月同一天))/ 指标值(去年月同一天) 。
其实如果关注每天的数据文件变化情况,这个代码里面的环比、同比其实意义不大,其实我们可以这样定义环比、同比:
环比: (指标当前值 - 指标值(昨天))/指标值(昨天)。
同比: (指标当前值 - 指标值 (上个月))/指标值(上个月)
当然,你也可以把这四个指标都加上,对比参考,侧重点不同而已。
存储过程
- IF OBJECT_ID(N‘sp_diskcapacity_cal‘)IS NOT NULL
- DROP PROCEDURE sp_diskcapacity_cal;
- GO
- CREATE PROCEDURE dbo.sp_diskcapacity_cal
- AS
- BEGIN
- INSERT INTO dbo.DiskCapacityHistory
- (
- [Date_CD] ,
- [DataBaseID] ,
- [FileID] ,
- [DataBaseName] ,
- [LogicalName] ,
- [FileTypeDesc] ,
- [PhysicalName] ,
- [StateDesc] ,
- [MaxSize] ,
- [GrowthType] ,
- [IsReadOnly] ,
- [IsPercentGrowth] ,
- [Size]
- )
- SELECT CAST(REPLACE(CONVERT(varchar(10),GETDATE(),120),‘-‘,‘‘) AS INT)
- AS DateCD ,
- database_id AS DataBaseId ,
- file_id AS FileID ,
- DB_NAME(database_id) AS DataBaseName ,
- name AS LogicalName ,
- type_desc AS FileTypeDesc ,
- physical_name AS PhysicalName ,
- state_desc AS StateDesc ,
- CASE WHEN max_size = 0 THEN N‘不允许增长‘
- WHEN max_size = -1 THEN N‘自动增长‘
- ELSE LTRIM(STR(max_size * 8.0 / 1024 / 1024, 14, 2)) + ‘G‘
- END AS MaxSize ,
- CASE WHEN is_percent_growth = 1
- THEN RTRIM(CAST(Growth AS CHAR(10))) + ‘%‘
- ELSE RTRIM(CAST(Growth AS CHAR(10))) + ‘M‘
- END AS Growth ,
- Is_Read_Only AS IsReadOnly ,
- Is_Percent_Growth AS IsPercentGrowth ,
- CAST(size * 8.0 / 1024 / 1024 AS DECIMAL(8, 4)) AS Size
- FROM sys.master_files;
- MERGE INTO dbo.DiskCapacityHistory DM USING
- (
- SELECT M.Date_CD ,
- M.DataBaseID ,
- M.FileID ,
- CASE WHEN N.SIZE IS NULL OR N.SIZE = 0 THEN 0 ELSE
- (M.SIZE - N.SIZE)/N.SIZE END AS Growth_MOM_RAT
- FROM dbo.DiskCapacityHistory M
- LEFT JOIN dbo.DiskCapacityHistory N ON
- CAST(CAST(M.Date_CD AS CHAR(8)) AS DATE) = DATEADD(MONTH, 1, CAST(CAST(N.Date_CD AS CHAR(8)) AS DATE))
- AND M.DataBaseID = N.DataBaseID AND M.FileID = N.FileID
- WHERE M.Date_CD = CAST(REPLACE(CONVERT(varchar(10),GETDATE(),120),‘-‘,‘‘) AS INT)
- ) TMP
- ON
- (
- DM.Date_CD = TMP.Date_CD AND
- DM.DatabaseId = TMP.DataBaseId AND
- DM.FileId = TMP.FileId
- )
- WHEN MATCHED THEN UPDATE SET
- DM.Growth_MOM_RAT = TMP.Growth_MOM_RAT;
- END
- GO
顺便吐槽一下:由于前两年一直使用ORACLE数据库,很少接触SQL SERVER,在实现上面功能的时候,我深深的体会到了ORACLE和SQL SERVER的巨大差距,如果用PL/SQL实现,那非常方便快捷,但是用T-SQL让我遇到了几个相当痛苦地方,下面顺便记录对比一下吧:
一:由于我采用INT来保存日期数据,那么需要在DATE类型和INT类型之间转换,我们来对比一下两者的差别吧:
1.1 DATE类型转换为整型:
T-SQL:
SELECT CAST(REPLACE(CONVERT(varchar(10),GETDATE(),120),‘-‘,‘‘) AS INT);
PL/SQL:
SELECT TO_CHAR(Date_CD, ‘YYYYMMDD‘) FROM DUAL;
1.2 整型转换为DATE类型(字段DATE_CD)
T-SQL:
SELECT CAST(CAST(DATE_CD AS CHAR(8)) AS DATE) FROM TEST;
PL/SQL:
SELECT TO_DATE(DATE_CD, ‘YYYY-MM-DD‘) FROM TEST;
结论: 纯属个人感受,从上面的脚本的简单性,方便性上,感觉ORACLE完胜SQL SERVER
二:计算数据文件增长同比、环比值
1:SQL SERVER 2005 没有MERGE语句功能,上面的脚本得改写成
Code Snippet
- UPDATEdbo.DiskCapacityHistory
- SET GROWTH_MOM_RAT =( SELECTCASE WHEN N.SIZE IS NULL
- OR N.SIZE = 0 THEN 0
- ELSE ( dbo.DiskCapacityHistory.SIZE
- - N.SIZE ) / N.SIZE
- END AS Growth_MOM_RAT
- FROM dbo.DiskCapacityHistory N
- WHERE CAST(CAST(dbo.DiskCapacityHistory.Date_CD AS CHAR(8)) AS DATE) = DATEADD(MONTH,
- 1,
- CAST(CAST(N.Date_CD AS CHAR(8)) AS DATE))
- AND dbo.DiskCapacityHistory.DataBaseID = N.DataBaseID
- AND dbo.DiskCapacityHistory.FileID = N.FileID
- )
- WHEREdbo.DiskCapacityHistory.Date_CD = CAST(REPLACE(CONVERT(VARCHAR(10), GETDATE(), 120),
- ‘-‘, ‘‘) AS INT)
- UPDATEdbo.DiskCapacityHistory
- SET GROWTH_YOY_RAT =( SELECTCASE WHEN N.SIZE IS NULL
- OR N.SIZE = 0 THEN 0
- ELSE ( dbo.DiskCapacityHistory.SIZE
- - N.SIZE ) / N.SIZE
- END AS Growth_YOY_RAT
- FROM dbo.DiskCapacityHistory N
- WHERE CAST(CAST(dbo.DiskCapacityHistory.Date_CD AS CHAR(8)) AS DATE) = DATEADD(MONTH,
- 12,
- CAST(CAST(N.Date_CD AS CHAR(8)) AS DATE))
- AND dbo.DiskCapacityHistory.DataBaseID = N.DataBaseID
- AND dbo.DiskCapacityHistory.FileID = N.FileID
- )
- WHEREdbo.DiskCapacityHistory.Date_CD = CAST(REPLACE(CONVERT(VARCHAR(10), GETDATE(), 120),
- ‘-‘, ‘‘) AS INT)
或
Code Snippet
- CREATE TABLE #DiskCapacityHistory
- (
- DATE_CD INT ,
- DataBaseID INT ,
- FileID INT ,
- Growth_MOM_RAT FLOAT
- ) ;
- INSERTINTO #DiskCapacityHistory
- SELECT M.DATE_CD ,
- M.DataBaseID ,
- M.FileID ,
- CASE WHEN N.SIZE IS NULL
- OR N.SIZE = 0 THEN 0
- ELSE ( M.SIZE - N.SIZE ) / N.SIZE
- END AS Growth_MOM_RAT
- FROM dbo.DiskCapacityHistory M ,
- dbo.DiskCapacityHistory N
- WHERE CAST(CAST(M.Date_CD AS CHAR(8)) AS DATE) = DATEADD(MONTH, 1,
- CAST(CAST(N.Date_CD AS CHAR(8)) AS DATE))
- AND M.DataBaseID = N.DataBaseID
- AND M.FileID = N.FileID
- AND M.Date_CD = CAST(REPLACE(CONVERT(VARCHAR(10), GETDATE()
- - 1, 120), ‘-‘, ‘‘) AS INT)
- UPDATE dbo.DiskCapacityHistory
- SET Growth_MOM_RAT = M.Growth_MOM_RAT
- FROM #DiskCapacityHistory M
- WHERE dbo.DiskCapacityHistory.DATE_CD = M.DATE_CD
- AND dbo.DiskCapacityHistory.DataBaseID = M.DataBaseID
- AND dbo.DiskCapacityHistory.FileID = M.FileID ;
2: 幸好SQL 2008还把ORACLE的MERGE的功能给模仿了过来,但是T-SQL缺少ORACLE数据库强大的分析函数LAG,如果有这个,我计算环比,同比就非常方便了,一个SQL就搞定了,下面是个例子,本想把ORACLE的SQL也做个例子展现,但是又要建表、造数,折腾起来比较麻烦。
Oracle Sample
- MERGE INTO DM.TM_WGGBO_IDCTOBUSVOLDTL_DAY DM
- USING (
- SELECT *
- FROM (
- SELECT DATE_CD,
- CITY_ID,
- IDC_NODE,
- VOL_TYPE,
- LAG(IDC_VOL_RAT ) OVER(PARTITION BY CITY_ID,IDC_NODE,VOL_TYPE,SUBSTR(DATE_CD,7,2) ORDER BY SUBSTR(DATE_CD,0,6)) AS IDC_MOM_RAT ,
- LAG(IDC_VOL_RAT ) OVER(PARTITION BY CITY_ID,IDC_NODE,VOL_TYPE,SUBSTR(DATE_CD,5,4) ORDER BY SUBSTR(DATE_CD,0,4)) AS IDC_YOY_RAT ,
- FROM DM.TM_WGGBO_IDCTOBUSVOLDTL_DAY
- ) T
- WHERE EXISTS(SELECT 1 FROM ETL.T_IDCVOL_DAY_${ssid} WHERE DATE_CD = T.DATE_CD)
- ) TEMP
- ON (
- DM.DATE_CD = TEMP.DATE_CD AND
- DM.CITY_ID = TEMP.CITY_ID AND
- DM.IDC_NODE = TEMP.IDC_NODE AND
- DM.VOL_TYPE = TEMP.VOL_TYPE
- )
- WHEN MATCHED THEN
- UPDATE
- SET DM.IDC_MOM_RAT = TEMP.IDC_MOM_RAT ,
- DM.IDC_YOY_RAT = TEMP.IDC_YOY_RAT
- ;
- COMMIT;