Question

情况是存在某些存储过程和/或特殊SQL导致我们的CPU从30％上升到80％并且将所有索引从内存中删除，我想知道是否已经建立了良好的将性能峰值（CPU，磁盘读取等）与运行SP的特定实例相关联的方法。

Answer 1

我可以想到两种间接方式

注意CPU，读取，通过存储过程调用的探查器写入我们最近使用它来发现昂贵的IO / CPU与持续时间相比
使用DMV在我们的系统上找到最贵的，这也显示存储过程

示例：

SELECT TOP 10 
    QS.total_logical_reads,
    QS.total_logical_writes, 
    QS.execution_count,
    QS.total_logical_reads + QS.total_logical_writes AS [IO_total],
    QS.total_logical_reads / QS.execution_count AS Avg_Reads_per_Execution,
    ST.[text] AS query_text,
    db_name(ST.dbid) AS database_name,
    ST.objectid AS OBJECT_ID
FROM 
    sys.dm_exec_query_stats  QS
    CROSS APPLY 
    sys.dm_exec_sql_text(sql_handle) ST
WHERE 
    QS.total_logical_reads + QS.total_logical_writes > 0 
    --AND ST.dbid IS NULL 
ORDER BY 
    [IO_total] DESC;

Answer 2

性能监视器。如果是Sql Server 2008 - ＆gt; 启动，程序文件，Microsoft Sql Server 2008，性能工具，Sql Server Profiler 。 生成跟踪。跟踪包含各种信息。例如，通过一些工作，可以找到存储过程的开始和结束跟踪事件，请注意它运行的时间覆盖了峰值，然后通过查看跟踪收集的资源计数器进一步验证。在sql server性能上尝试此链接：http://www.sql-server-performance.com/articles/per/correlate_trace_performance_p1.aspx
在SQL Server代理中使用警报来触发自定义T-SQL，该T-SQL捕获正在运行@特定时间的内容。 Sql Server代理，警报，常规，新警报，Type = SQL Server性能状况警报（根据需要设置），响应，执行作业。该作业有一个sql脚本写入表 - 告诉你发生了什么。各种各样的世界可能是你的牡蛎。例如，你可以问自己 - 嘿，现在记忆中的clr是什么？

现在记忆力是什么？

SELECT
  a.*
 ,cla.*
 ,clr_ad.*
 ,state_desc = CASE clr_ad.[state]
     WHEN N'E_APPDOMAIN_CREATING' THEN 'The AppDomain is being created.'
     WHEN N'E_APPDOMAIN_SHARED' THEN 'The runtime AppDomain is ready for use by multiple users.'
     WHEN N'E_APPDOMAIN_SINGLEUSER' THEN 'The AppDomain is ready for use in DDL operations. These differ from E_APPDOMAIN_SHARED in that shared AppDomains are used for CLR integration executions as opposed to DDL operations. Such AppDomains are isolated from other concurrent operations.'
     WHEN N'E_APPDOMAIN_DOOMED' THEN 'The AppDomain is scheduled to be unloaded, but there are currently threads executing in it.'
     WHEN N'E_APPDOMAIN_UNLOADING' THEN 'SQL Server has requested that the CLR unload the AppDomain, usually because the assembly that contains the managed database objects has been altered or dropped.'
     WHEN N'E_APPDOMAIN_UNLOADED' THEN 'The CLR has unloaded the AppDomain. This is usually the result of an escalation procedure due to ThreadAbort, OutOfMemory, or an unhandled exception in user code.'
     WHEN N'E_APPDOMAIN_ENQUEUE_DESTROY' THEN 'The AppDomain has been unloaded in CLR and set to be destroyed by SQL Server.'
     WHEN N'E_APPDOMAIN_DESTROY' THEN 'The AppDomain is in the process of being destroyed by SQL Server.'
     WHEN N'E_APPDOMAIN_ZOMBIE' THEN 'The AppDomain has been destroyed by SQL Server; however, not all of the references to the AppDomain have been cleaned up.'
     ELSE '?'
     END
FROM
  sys.dm_clr_loaded_assemblies cla
  JOIN sys.assemblies a
    ON cla.assembly_id = a.assembly_id
  JOIN sys.dm_clr_appdomains clr_ad
    ON cla.appdomain_address = clr_ad.appdomain_address

如何将SQL Server性能指标与导致问题的存储过程相关联

2 个答案: