监控数据库运行

下面是整理、收集监控数据库运行的一些常用脚本,也是MS SQL 日常维护管理常用脚本(一)的续集,欢迎大家补充、提意见。

 

查看数据库登录名信息
 
Code Snippet
  1. SELECT name                                AS LoginName ,
  2.        dbname                              AS DefaultDB ,
  3.        createdate                          AS CreateDate,
  4.        updatedate                          AS UpdateDate,
  5.        language                            AS Language  ,
  6.        CASE WHEN isntname = 1 THEN 'NT USER'
  7.            ELSE 'SQL USER'    END          AS UserType
  8. FROM syslogins;

 

查看数据库用户信息
 
SELECT * FROM sysusers;
 
查看用户拥有的服务器角色
 

方法1: 用SSMS管理工具查看


方法2: 脚本查询

 
查看用户角色
  1. SELECT name            ,
  2.        CASE WHEN sysadmin     = 1       THEN 'yes'     ELSE '' END AS IsSysadmin        ,
  3.        CASE WHEN dbcreator    = 1       THEN 'yes'     ELSE '' END AS IsDbCreate        ,
  4.        CASE WHEN securityadmin= 1       THEN 'yes'     ELSE '' END AS IsSecurityadmin   ,
  5.        CASE WHEN bulkadmin    = 1       THEN 'yes'     ELSE '' END AS IsBulkadmin       ,
  6.        CASE WHEN diskadmin    = 1       THEN 'yes'     ELSE '' END AS IsDiskadmin       ,
  7.        CASE WHEN processadmin = 1       THEN 'yes'     ELSE '' END AS IsProcessadmin    ,
  8.        CASE WHEN serveradmin  = 1       THEN 'yes'     ELSE '' END AS IsServeradmin     ,
  9.        CASE WHEN setupadmin   = 1       THEN 'yes'     ELSE '' END AS IsSetupadmin    
  10. FROM syslogins
  11. --WHERE NAME='loginname'
 
查看最大工作线程数
 
Code Snippet
  1. SELECT  max_workers_count
  2.   FROM  sys.dm_os_sys_info

 

查看当前用户进程的会话ID

 

SELECT @@SPID

 

查询当前会话使用哪种协议

Code Snippet
  1. SELECT net_transport
  2. FROM   sys.dm_exec_connections
  3. WHERE session_id = @@SPID;
 
查看当前连接的会话信息

--进程号1--50是SQL Server系统内部用的
SELECT * FROM sys.dm_exec_sessions WHERE session_id >=51
 
--查看某台机器的连接会话信息
SELECT * FROM sys.dm_exec_sessions WHERE session_id >=51 AND host_name='PO130018801'
 
--查看某个登录名的连接会话信息
SELECT * FROM sys.dm_exec_sessions WHERE session_id >=51 AND login_name='username'
 
--查看活动的连接会话信息
SELECT * FROM sys.dm_exec_sessions WITH(NOWAIT) WHERE session_id >=51 AND status ='running'
 
--查找连接到服务器的用户并返回每个用户的会话数
SELECT  login_name ,
        COUNT(session_id) AS session_count
FROM    sys.dm_exec_sessions
GROUP BY login_name ;
 
 
查看正在执行的SQL语句
 

方法1: 选择数据库实例,单击右键,选择”活动监视器“,监控/查看正在执行的SQL


方法2: profile去跟踪,比较耗费资源。

 

方法3:

 
Code Snippet
  1. SELECT[Spid] = session_Id ,
  2.       ecid ,
  3.       [Database] = DB_NAME(sp.dbid) ,
  4.       [User] = nt_username ,
  5.       [Status] = er.status ,
  6.       [Wait] = wait_type ,
  7.       [Individual Query] = SUBSTRING(qt.text, er.statement_start_offset / 2,
  8.                                      ( CASE WHEN er.statement_end_offset = -1
  9.                                             THEN LEN(CONVERT(NVARCHAR(MAX), qt.text))
  10.                                                  * 2
  11.                                             ELSE er.statement_end_offset
  12.                                        END - er.statement_start_offset ) / 2) ,
  13.       [Parent Query] = qt.text ,
  14.       Program = program_name ,
  15.       Hostname ,
  16.       nt_domain ,
  17.       start_time
  18.   FROMsys.dm_exec_requests er
  19.       INNER JOIN sys.sysprocesses sp ON er.session_id = sp.spid
  20.       CROSS APPLY sys.dm_exec_sql_text(er.sql_handle) AS qt
  21.  WHEREsession_Id >= 51

方法4:

Code Snippet
  1. SELECT   m.session_id ,
  2.          m.start_time ,
  3.          m.command    ,
  4.          m.wait_type  ,
  5.          m.cpu_time   ,
  6.         CAST(s.text AS VARCHAR(1000)) AS sqlText
  7.  FROMmaster.sys.dm_exec_requests m WITH ( NOLOCK )
  8.      CROSS APPLY fn_get_sql(m.sql_handle) s
  9.  
  10. SELECT  r.session_id,
  11.         r.start_time      ,
  12.         r.command         ,
  13.         r.wait_type       ,
  14.         r.cpu_time        ,
  15.         s.text  
  16.  FROMsys.dm_exec_requests r
  17.      CROSS APPLY sys.dm_exec_sql_text(r.sql_handle) s

--查看某个会话ID正在执行的SQL

Code Snippet
  1. SELECT  m.session_id ,
  2.         m.start_time ,
  3.         m.command    ,
  4.         m.wait_type  ,
  5.         m.cpu_time   ,
  6.       CAST(s.text AS VARCHAR(1000)) AS sqlText
  7.   FROMmaster.sys.dm_exec_requests m WITH ( NOLOCK )
  8.       CROSS APPLY fn_get_sql(m.sql_handle) s
  9.  WHEREm.session_id = 342
  10.  
  11.     SELECT   r.session_id      ,
  12.              r.start_time      ,
  13.              r.command         ,
  14.              r.wait_type       ,
  15.              r.cpu_time        ,
  16.              s.text  
  17.   FROM sys.dm_exec_requests r
  18.       CROSS APPLY sys.dm_exec_sql_text(r.sql_handle) s
  19. WHERE r.seesion_id =342

查看SQL SERVER进程执行的语句

Code Snippet
  1. USE master
  2.  
  3. DECLARE @spid INT ;
  4.  
  5. DECLARE @sql_handle BINARY(20) ;
  6.  
  7. SET @spid = 56
  8.  
  9.   SELECT@sql_handle = sql_handle
  10.     FROMsysprocesses AS A WITH ( NOLOCK )
  11.    WHEREspid = @spid ;
  12.  
  13.   SELECTtext
  14.     FROM::fn_get_sql(@sql_handle) ;
 
查找TOP N语句

按平均 CPU 时间返回排名前十个的查询的相关信息。此示例将根据查询的查询哈希对查询进行聚合,以便按照查询的累积资源消耗来分组在逻辑上等效的查询。
--注意:SQL 2005 某些版本,没有sys.dm_exec_query_stats系统动态视图没有query_hash视图。

Code Snippet
  1. USE DBNAME;
  2. GO
  3. SELECT TOP 10 query_stats.query_hash             AS "Query Hash",
  4.     SUM(query_stats.total_worker_time) /
  5.     SUM(query_stats.execution_count)             AS "Avg CPU Time",
  6.     MIN(query_stats.statement_text)              AS "Statement Text"
  7. FROM
  8.     (SELECT QS.*,
  9.     SUBSTRING(ST.text,(QS.statement_start_offset/2) + 1,
  10.     ((CASE statement_end_offset
  11.         WHEN -1 THEN DATALENGTH(st.text)
  12.         ELSE QS.statement_end_offset END
  13.             - QS.statement_start_offset)/2) + 1) AS statement_text
  14.      FROM sys.dm_exec_query_stats AS QS
  15.      CROSS APPLY sys.dm_exec_sql_text(QS.sql_handle) as ST) as query_stats
  16. GROUP BY query_stats.query_hash
  17. ORDER BY 2 DESC;
  18. GO
查看会话阻塞/死锁信息
 

方法1:查看那个引起阻塞,查看blk不为0的记录,如果存在阻塞进程,则是该阻塞进程的会话 ID。否则该列为零。 

    EXEC sp_who active

方法2:查看那个引起阻塞,查看字段BlkBy,这个能够得到比sp_who更多的信息。

    EXEC sp_who2 active

方法3:sp_lock 系统存储过程,报告有关锁的信息,但是不方便定位问题


方法4:sp_who_lock存储过程

Code Snippet
  1. USE master;
  2. GO
  3.  
  4. SET ANSI_NULLS ON;
  5. GO
  6.  
  7. SET QUOTED_IDENTIFIER ON;
  8. GO
  9.  
  10. IF EXISTS(SELECT 1 FROM sysobjects WHERE id=OBJECT_ID(N'sp_who_lock')
  11.                                        AND OBJECTPROPERTY(id, 'IsProcedure') =1)
  12.     DROP PROCEDURE sp_who_lock;
  13. GO
  14.  
  15. --==================================================================================================
  16. --            ProcedureName             :            sp_who_lock
  17. --            Author                    :            作者不详,出自网络
  18. --            CreateDate                :            2013-05-13
  19. --            Description               :            查看阻塞和死锁信息
  20. /**************************************************************************************************
  21.         Parameters                    :              参数说明
  22. ***************************************************************************************************
  23.                                                      无参存储过程
  24. ***************************************************************************************************
  25.         Modified Date            Modified User     Version                 Modified Reason
  26. ***************************************************************************************************
  27.        2013-06-03                    Kerry        V01.00.01 调整存储过程格式,代码部分修改以及增加注释信息
  28. ***************************************************************************************************/
  29. --=================================================================================================
  30.  
  31.   CREATE PROCEDURE sp_who_lock
  32.     AS
  33. BEGIN
  34.  
  35. DECLARE @spid                            INT;
  36. DECLARE @block                           INT;
  37. DECLARE @RowCount                        INT;
  38. DECLARE @RowIndex                        INT;
  39.  
  40. --创建临时表,保持被阻塞或正阻塞其他SQL的SQL语句信息
  41. CREATE TABLE #tmp_lock_who
  42. (
  43.       id    INT IDENTITY(1, 1) ,
  44.       spid    SMALLINT ,
  45.       block SMALLINT
  46. )
  47.     
  48.     
  49. IF @@ERROR<>0 RETURN @@ERROR;
  50.  
  51. INSERT INTO #tmp_lock_who
  52.         (
  53.           spid ,
  54.           block
  55.         )
  56.         SELECT  0 ,
  57.                 blocked
  58.         FROM    ( SELECT    *
  59.                   FROM      sysprocesses
  60.                   WHERE     blocked > 0
  61.                 ) a
  62.         WHERE   NOT EXISTS( SELECT *
  63.                              FROM   ( SELECT    *
  64.                                       FROM      sysprocesses
  65.                                       WHERE     blocked > 0
  66.                                     ) b
  67.                              WHERE  a.blocked = spid )
  68.         UNION
  69.         SELECT  spid ,
  70.                 blocked
  71.         FROM    sysprocesses
  72.         WHERE   blocked > 0;
  73.         
  74.         
  75. IF @@ERROR<>0 RETURN @@ERROR;
  76.  
  77.  
  78. -- 找到临时表的记录数
  79.   SELECT@RowCount = COUNT(1) ,
  80.         @RowIndex = 1
  81.     FROM#tmp_lock_who
  82.  
  83.  
  84. IF @@ERROR<>0 RETURN @@ERROR;
  85.  
  86.   IF@RowCount=0
  87.   SELECT  N'现在没有阻塞和死锁信息' AS MESSAGE;
  88.  
  89.  
  90. -- -- 循环开始
  91. WHILE @RowIndex <= @RowCount
  92. BEGIN
  93.     -- 取第一条记录
  94.     SELECT     @spid    = spid,
  95.                @block   = block
  96.     FROM #tmp_lock_who
  97.     WHERE Id = @RowIndex
  98.  
  99.     IF @spid = 0
  100.         SELECT N'引起数据库死锁的是: ' + CAST(@block AS VARCHAR(10))
  101.             +  N'进程号,其执行的SQL语法如下';
  102.     ELSE
  103.         SELECT N'进程号SPID:' + CAST(@spid AS VARCHAR(10)) + N'被进程号SPID:'
  104.             +  CAST(@block AS VARCHAR(10)) +'阻塞,其当前进程执行的SQL语法如下';
  105.         
  106.     DBCC INPUTBUFFER(@block )
  107.  
  108.  
  109.     SET @RowIndex = @RowIndex + 1;
  110.  
  111. END;
  112.  
  113.     DROP TABLE #tmp_lock_who;
  114.  
  115.     RETURN 0;
  116. END

方法5:右键服务器-选择“活动和监视器”,查看进程选项。注意“任务状态”字段。

方法6:右键服务名称-选择报表-标准报表-活动-所有正在阻塞的事务。


小结:总结之后,才发现居然有这么多方法,MGD,系统的整理、梳理知识点是非常有必要的,你能更全面、深入的了解。

 
查看内存状态

dbcc memorystatus


具体如何分析,请查看官方文档http://support.microsoft.com/kb/907877/zh-cn

 
查看脚本执行时间


方法1: 查看SSMS管理器,查询窗口右下角
方法2:

Code Snippet
  1. DECLARE @exectime DATETIME
  2.   SELECT@exectime = GETDATE()
  3. --SQL 语句
  4. PRINT N'SQL执行耗时:' + CONVERT(VARCHAR(10), DATEDIFF(ms, @exectime, GETDATE()))

方法3:

SET STATISTICS TIME ON

--SQL 语句

 

查看进程正在执行的SQL语句 

  dbcc inputbuffer ()

查看那些表缺少索引 

 

下面语句功能强大,执行结果受统计信息的影响

Code Snippet
  1. SELECT sys.objects.name table_name,
  2.        mid.statement full_name,
  3.     (migs.avg_total_user_cost * migs.avg_user_impact) *(migs.user_seeks + migs.user_scans) AS Impact,
  4.     migs.avg_user_impact *(migs.user_seeks + migs.user_scans) Avg_Estimated_Impact,
  5.     'CREATE NONCLUSTERED INDEX IDX_' + sys.objects.name + '_N ON '
  6.         + sys.objects.name COLLATE DATABASE_DEFAULT
  7.         + ' ( ' + IsNull(mid.equality_columns, '')
  8.         + CASE WHEN mid.inequality_columns IS NULL
  9.                 THEN ''
  10.             ELSE
  11.                 CASE WHEN mid.equality_columns IS NULL
  12.                     THEN ''
  13.                 ELSE ','
  14.                 END + mid.inequality_columns
  15.             END + ' ) '
  16.         + CASE WHEN mid.included_columns IS NULL
  17.                 THEN ''
  18.             ELSE 'INCLUDE (' + mid.included_columns + ')' END
  19.         + ';' AS CreateIndexStatement,
  20.     mid.equality_columns,
  21.     mid.inequality_columns,
  22.     mid.included_columns
  23. FROM sys.dm_db_missing_index_group_stats AS migs
  24.     INNER JOIN sys.dm_db_missing_index_groups AS mig ON migs.group_handle = mig.index_group_handle
  25.     INNER JOIN sys.dm_db_missing_index_details AS mid ON mig.index_handle = mid.index_handle
  26.         AND mid.database_id = DB_ID()
  27.     INNER JOIN sys.objects WITH (nolock) ON mid.OBJECT_ID = sys.objects.OBJECT_ID
  28. WHERE (migs.group_handle IN
  29.         (
  30.             SELECT TOP (500) group_handle
  31.             FROM sys.dm_db_missing_index_group_stats WITH (nolock)
  32.             ORDER BY (avg_total_user_cost * avg_user_impact) *(user_seeks + user_scans) DESC))
  33.     AND OBJECTPROPERTY(sys.objects.OBJECT_ID, 'isusertable')=1
  34. --ORDER BY [Impact] DESC, [full_name] DESC
  35. ORDER BY [table_name], [Impact] desc

查看应该被移除的索引

 

查看那些多余的、应该被移除的索引

SQL 1:

Code Snippet
  1. SELECT OBJECTNAME = OBJECT_NAME(I.OBJECT_ID),
  2.     INDEXNAME = I.NAME,
  3.     I.INDEX_ID
  4. FROM SYS.INDEXES I
  5.     JOIN SYS.OBJECTS O ON I.OBJECT_ID = O.OBJECT_ID
  6. WHERE OBJECTPROPERTY(O.OBJECT_ID,'IsUserTable') = 1
  7.     AND I.INDEX_ID NOT IN(
  8.                             SELECT S.INDEX_ID
  9.                             FROM SYS.DM_DB_INDEX_USAGE_STATS S
  10.                             WHERE S.OBJECT_ID = I.OBJECT_ID
  11.                                 AND I.INDEX_ID = S.INDEX_ID
  12.                                 AND DATABASE_ID = DB_ID())
  13. ORDER BY OBJECTNAME, I.INDEX_ID, INDEXNAME ASC

SQL 2: 分析下面语句,移除那些没有必要的索引

Code Snippet
  1. SELECT DB_NAME(database_id)         AS  N'DataBaseName'  ,
  2.        OBJECT_NAME(U.object_id)     AS  N'Table_Name'    ,
  3.        I.name                       AS  N'Index_Name'    ,
  4.        user_seeks                   AS  N'用户索引查找次数',
  5.        user_scans                   AS  N'用户索引扫描次数',
  6.        last_user_seek               AS  N'最后查找时间'   ,
  7.        last_user_scan               AS  N'最后扫描时间'   ,
  8.        rows                         AS  N'表中的行数'
  9. FROM sys.dm_db_index_usage_stats AS U
  10.    INNER JOIN sys.indexes I ON U.index_id= I.index_idAND U.object_id= I.object_id
  11.   INNER JOIN sysindexesT ON I.object_id = T.id
  12. WHERE database_id= DB_ID('DbName')
  13. AND OBJECT_NAME(U.object_id)='TableName'
  14. ORDER BY user_seeks, user_scans, object_name(U.object_id);