用于死锁检测的SQL扩展事件会话

Question

是否有一种方法来增加死锁扩展事件会话捕获的死锁<inputbuf>元素的大小？

我们希望看到完整的查询，以帮助定位应用程序代码中的问题。

它似乎仅限于1024个字符+/-。可以增加吗？

有关示例XML，请参阅下面。您可以看到<inputbuf>元素中的查询文本在select列表的中间被切断：

<deadlock>
 <victim-list>
  <victimProcess id="processc9c0829848" />
 </victim-list>
 <process-list>
  <process id="processc9c0829848" taskpriority="0" logused="0" waitresource="PAGE: 5:1:40600276 " waittime="696" ownerId="255115931225" transactionname="SELECT" lasttranstarted="2019-04-24T09:29:25.950" XDES="0xc8dfa8da40" lockMode="S" schedulerid="13" kpid="8480" status="suspended" spid="245" sbid="2" ecid="0" priority="0" trancount="0" lastbatchstarted="2019-04-24T09:29:25.950" lastbatchcompleted="2019-04-24T09:29:25.950" lastattention="1900-01-01T00:00:00.950" clientapp="EntityFramework" hostname="MSR-PRD-BDB02" hostpid="43440" loginname="IUSR_BuildDB" isolationlevel="read committed (2)" xactid="255115931225" currentdb="5" lockTimeout="4294967295" clientoption1="671088672" clientoption2="128056">
   <executionStack>
    <frame procname="adhoc" line="1" stmtstart="66" stmtend="7378" sqlhandle="0x02000000638e8b1acc45f82c476cd42914e32866e87c4fd60000000000000000000000000000000000000000">
unknown    </frame>
    <frame procname="unknown" line="1" sqlhandle="0x0000000000000000000000000000000000000000000000000000000000000000000000000000000000000000">
unknown    </frame>
   </executionStack>
   <inputbuf>
(@p__linq__0 int,@p__linq__1 int)SELECT 
[Project1].[CachedBuildStateId] AS [CachedBuildStateId], 
[Project1].[BuildVersionId1] AS [BuildVersionID], 
[Project1].[ProjectID] AS [ProjectID], 
[Project1].[VersionName] AS [VersionName], 
[Project1].[PlatformID] AS [PlatformID], 
[Project1].[VersionPath] AS [VersionPath], 
[Project1].[RegionID] AS [RegionID], 
[Project1].[TestStatusID] AS [TestStatusID], 
[Project1].[TestResult] AS [TestResult], 
[Project1].[TestReportPath] AS [TestReportPath], 
[Project1].[TypeID] AS [TypeID], 
[Project1].[Size] AS [Size], 
[Project1].[VersionDate] AS [VersionDate], 
[Project1].[Approved] AS [Approved], 
[Project1].[Archived] AS [Archived], 
[Project1].[CLData] AS [CLData], 
[Project1].[CLCode] AS [CLCode], 
[Project1].[CLSound] AS [CLSound], 
[Project1].[XDKSDKVersion] AS [XDKSDKVersion], 
[Project1].[Comments] AS [Comments], 
[Project1].[DateAdded] AS [DateAdded], 
[Project1].[DateModified] AS [DateModified], 
[Project1].[CacheLocally] AS [CacheLocally],   </inputbuf>
  </process>
  <process id="processc9bf5cf468" taskpriority="0" logused="436" waitresource="PAGE: 5:1:3752363 " waittime="724" ownerId="255115931408" transactionname="user_transaction" lasttranstarted="2019-04-24T09:29:25.973" XDES="0xcf6b186408" lockMode="IX" schedulerid="25" kpid="11228" status="suspended" spid="209" sbid="2" ecid="0" priority="0" trancount="2" lastbatchstarted="2019-04-24T09:29:25.973" lastbatchcompleted="2019-04-24T09:29:25.970" lastattention="1900-01-01T00:00:00.970" clientapp="EntityFramework" hostname="MSR-PRD-BDB04" hostpid="6248" loginname="IUSR_BuildDB" isolationlevel="read committed (2)" xactid="255115931408" currentdb="5" lockTimeout="4294967295" clientoption1="671088672" clientoption2="128056">
   <executionStack>
    <frame procname="adhoc" line="1" stmtstart="78" stmtend="408" sqlhandle="0x020000009438fe362ae846385f4e36448d4eac52a5f052e10000000000000000000000000000000000000000">
unknown    </frame>
    <frame procname="unknown" line="1" sqlhandle="0x0000000000000000000000000000000000000000000000000000000000000000000000000000000000000000">
unknown    </frame>
   </executionStack>
   <inputbuf>
(@0 int,@1 datetime,@2 datetime,@3 int)update [dbo].[CachedBuildState]
set [CachingStatusId] = @0, [StartTimeUtc] = @1, [LastUpdateTimeUtc] = @2, [MessageDetails] = null
where ([CachedBuildStateId] = @3)
   </inputbuf>
  </process>
 </process-list>
 <resource-list>
  <pagelock fileid="1" pageid="40600276" dbid="5" subresource="FULL" objectname="BuildDB.dbo.CachedBuildState" id="lockcda4d62280" mode="IX" associatedObjectId="72057594075283456">
   <owner-list>
    <owner id="processc9bf5cf468" mode="IX" />
   </owner-list>
   <waiter-list>
    <waiter id="processc9c0829848" mode="S" requestType="wait" />
   </waiter-list>
  </pagelock>
  <pagelock fileid="1" pageid="3752363" dbid="5" subresource="FULL" objectname="BuildDB.dbo.CachedBuildState" id="lockd70db08700" mode="S" associatedObjectId="72057594075283456">
   <owner-list>
    <owner id="processc9c0829848" mode="S" />
   </owner-list>
   <waiter-list>
    <waiter id="processc9bf5cf468" mode="IX" requestType="wait" />
   </waiter-list>
  </pagelock>
 </resource-list>
</deadlock>

Answer 1

每个帧都有一个sqlhandle，因此您可以使用sys.dm_exec_sql_text函数获得完整的查询：

SELECT text
FROM sys.dm_exec_sql_text(sql_handle)

sys.dm_exec_sql_文本

例如，从sys.dm_exec_query_stats获取一些语句的文本(stmtstart、stm趋向、sqlhandle，而不是XML报告中的statement_start_offset、statement_end_offset、sql_handle )。

SELECT TOP (100)
    CASE   
        WHEN s.[statement_start_offset] > 0 THEN  
            --The start of the active command is not at the beginning of the full command text 
            CASE s.[statement_end_offset]  
               WHEN -1 THEN  
                  --The end of the full command is also the end of the active statement 
                  SUBSTRING(t.TEXT, (s.[statement_start_offset]/2) + 1, 2147483647) 
               ELSE   
                  --The end of the active statement is not at the end of the full command 
                  SUBSTRING(t.TEXT, (s.[statement_start_offset]/2) + 1, (s.[statement_end_offset] - s.[statement_start_offset])/2+1)   
            END  
        ELSE  
            --1st part of full command is running 
            CASE s.[statement_end_offset]  
               WHEN -1 THEN  
                  --The end of the full command is also the end of the active statement 
                  RTRIM(LTRIM(t.[text]))  
               ELSE  
                  --The end of the active statement is not at the end of the full command 
                  LEFT(t.TEXT, (s.[statement_end_offset]/2) +1)  
            END  
        END AS [executing statement]
FROM sys.dm_exec_query_stats s 
    CROSS APPLY sys.dm_exec_sql_text(s.[sql_handle]) t

UPD

感谢@JohnEisbrener的合理评论，我应该注意到，在清理查询缓存之前，来自DMV的信息将是实际的。因此，如果您希望在缓存清理后(例如服务器重新启动后)使用该信息，则必须在事件发生后立即将该数据保存在某个永久表中。