解释死锁并修复它

我需要了解这是如何可能的，即我想了解这个机制。解释死锁并修复它

enter image description here

受害者过程是一个很大的联合型查询涉及约6表（实体框架）

工艺上的右侧是插入/更新批次由多个语句。

我无法给SQL，因为它们在死锁事件中被截断。

我见过涉及2/3表和写操作的死锁，但我无法解释这个。发生了什么？我看到索引涉及，这是有效的索引 - 我需要它。

每周也许3次我得到这个死锁一种系统，其中每分钟约100个这样的选择并每秒

<deadlock-list> 
<deadlock victim="process2fdc4e088"> 
    <process-list> 
    <process id="process2fdc4e088" taskpriority="0" logused="0" waitresource="KEY: 28:72057594054049792 (7303a0672d4e)" waittime="713" ownerId="928827354" transactionname="user_transaction" lasttranstarted="2012-06-27T06:32:35.030" XDES="0x9982e3b0" lockMode="S" schedulerid="3" kpid="15300" status="suspended" spid="84" sbid="2" ecid="0" priority="0" trancount="1" lastbatchstarted="2012-06-27T06:32:35.137" lastbatchcompleted="2012-06-27T06:32:35.030" clientapp="sss" hostname="aaa" hostpid="4080" loginname="aaa" isolationlevel="read committed (2)" xactid="928827354" currentdb="28" lockTimeout="4294967295" clientoption1="671088672" clientoption2="128056"> 
    <executionStack> 
    <frame procname="adhoc" line="1" sqlhandle="0x020000009b2fc809e2d580f750aacd7697bcc9fab8b85647"> 
SELECT 
[UnionAll2].[C3] AS [C1], 
[UnionAll2].[C4] AS [C2], 
[UnionAll2].[C5] AS [C3], 
[UnionAll2].[C6] AS [C4], 
[UnionAll2].[C7] AS [C5], 
[UnionAll2].[C8] AS [C6], 
[UnionAll2].[C9] AS [C7], 
[UnionAll2].[C10] AS [C8], 
[UnionAll2].[C11] AS [C9], 
[UnionAll2].[C12] AS [C10], 
[UnionAll2].[C13] AS [C11], 
[UnionAll2].[C14] AS [C12], 
[UnionAll2].[C15] AS [C13], 
[UnionAll2].[C16] AS [C14], 
[UnionAll2].[C17] AS [C15], 
[UnionAll2].[C18] AS [C16], 
[UnionAll2].[C19] AS [C17], 
[UnionAll2].[C20] AS [C18], 
[UnionAll2].[C21] AS [C19], 
[UnionAll2].[C22] AS [C20], 
[UnionAll2].[C23] AS [C21], 
[UnionAll2].[C24] AS [C22], 
[UnionAll2].[C25] AS [C23], 
[UnionAll2].[C1] AS [C24], 
[UnionAll2].[C26] AS [C25], 
[UnionAll2].[C27] AS [C26], 
[UnionAll2].[C28] AS [C27], 
[UnionAll2].[C29] AS [C28], 
[UnionAll2].[C30] AS [C29], 
[UnionAll2].[C31] AS [C30], 
[UnionAll2].[C32] AS [C31], 
[UnionAll2].[C33] AS [C32], 
[UnionAll2].[C34] AS [C33], 
[UnionAll2].[C35] AS [C34], 
[UnionAll2].  </frame> 
    </executionStack> 
    <inputbuf> 
SELECT 
[UnionAll2].[C3] AS [C1], 
[UnionAll2].[C4] AS [C2], 
[UnionAll2].[C5] AS [C3], 
[UnionAll2].[C6] AS [C4], 
[UnionAll2].[C7] AS [C5], 
[UnionAll2].[C8] AS [C6], 
[UnionAll2].[C9] AS [C7], 
[UnionAll2].[C10] AS [C8], 
[UnionAll2].[C11] AS [C9], 
[UnionAll2].[C12] AS [C10], 
[UnionAll2].[C13] AS [C11], 
[UnionAll2].[C14] AS [C12], 
[UnionAll2].[C15] AS [C13], 
[UnionAll2].[C16] AS [C14], 
[UnionAll2].[C17] AS [C15], 
[UnionAll2].[C18] AS [C16], 
[UnionAll2].[C19] AS [C17], 
[UnionAll2].[C20] AS [C18], 
[UnionAll2].[C21] AS [C19], 
[UnionAll2].[C22] AS [C20], 
[UnionAll2].[C23] AS [C21], 
[UnionAll2].[C24] AS [C22], 
[UnionAll2].[C25] AS [C23], 
[UnionAll2].[C1] AS [C24], 
[UnionAll2].[C26] AS [C25], 
[UnionAll2].[C27] AS [C26], 
[UnionAll2].[C28] AS [C27], 
[UnionAll2].[C29] AS [C28], 
[UnionAll2].[C30] AS [C29], 
[UnionAll2].[C31] AS [C30], 
[UnionAll2].[C32] AS [C31], 
[UnionAll2].[C33] AS [C32], 
[UnionAll2].[C34] AS [C33], 
[UnionAll2].[C35] AS [C34], 
[UnionAll2] </inputbuf> 
    </process> 
    <process id="processd5471948" taskpriority="0" logused="1204" waitresource="PAGE: 28:1:102676" waittime="864" ownerId="928827514" transactionname="user_transaction" lasttranstarted="2012-06-27T06:32:35.363" XDES="0x1e0f0d3f0" lockMode="IX" schedulerid="4" kpid="19116" status="suspended" spid="77" sbid="2" ecid="0" priority="0" trancount="2" lastbatchstarted="2012-06-27T06:32:35.377" lastbatchcompleted="2012-06-27T06:32:35.377" clientapp="DitatTMS" hostname="sss" hostpid="4080" loginname="sss" isolationlevel="read committed (2)" xactid="928827514" currentdb="28" lockTimeout="4294967295" clientoption1="671088672" clientoption2="128056"> 
    <executionStack> 
    <frame procname="adhoc" line="1" stmtstart="50" stmtend="250" sqlhandle="0x020000003859c72d14e4f731cc12f95e6e3ed8b75668b3b9"> 
update [dbo].[MBLTripStopAttribute] 
set [AttributeValue] = @0 
where ([TripStopAttributeKey] = @1)  </frame> 
    <frame procname="unknown" line="1" sqlhandle="0x000000000000000000000000000000000000000000000000"> 
unknown  </frame> 
    </executionStack> 
    <inputbuf> 
(@0 varchar(max) ,@1 int)update [dbo].[MBLTripStopAttribute] 
set [AttributeValue] = @0 
where ([TripStopAttributeKey] = @1) 
select [RowVersion] 
from [dbo].[MBLTripStopAttribute] 
where @@ROWCOUNT &gt; 0 and [TripStopAttributeKey] = @1 </inputbuf> 
    </process> 
    </process-list> 
    <resource-list> 
    <keylock hobtid="72057594054049792" dbid="28" objectname="DitatApp.dbo.MBLTripDriver" indexname="IX_MBLTripDriver_UpdatedOn_DriverKey" id="lock873ee900" mode="X" associatedObjectId="72057594054049792"> 
    <owner-list> 
    <owner id="processd5471948" mode="X"/> 
    </owner-list> 
    <waiter-list> 
    <waiter id="process2fdc4e088" mode="S" requestType="wait"/> 
    </waiter-list> 
    </keylock> 
    <pagelock fileid="1" pageid="102676" dbid="28" objectname="DitatApp.dbo.MBLTripStopAttribute" id="lock92d2c200" mode="SIU" associatedObjectId="72057594055163904"> 
    <owner-list> 
    <owner id="process2fdc4e088" mode="S"/> 
    </owner-list> 
    <waiter-list> 
    <waiter id="processd5471948" mode="IX" requestType="convert"/> 
    </waiter-list> 
    </pagelock> 
    </resource-list> 
</deadlock> 
</deadlock-list>

来源

2012-06-27 katit

图像是否被截断？我看到箭从侧面进来。 – Andomar

使用死锁源XML。 [图像往往是ambigous]（http://rusanu.com/2010/05/12/the-puzzle-of-u-locks-in-deadlock-graphs/） –

@Andomar我打破了图像在2 - 否则它太小了。视觉上“合并”顶部和底部部分作为左侧和右侧 - 中间部分共享 – katit

约1，例如插入件看起来像处理84上保持的共享意图更新pagelock甲它正在等待一个共享密钥锁B.

过程77 B.持有排按键锁定它在等待一个意向排页锁A.

下一步是获取SQL语句，这些进程正在执行。

来源

2012-06-27 14:21:20 Andomar

我可以发布近似的SQL，对于左侧的查询将会非常有用，但右侧查询将仅以插图的数量“说明” /该事务中的更新有所不同。它有助于更好地诊断？就像我说的XML没有任何帮助 - 查询大和截断.. – katit

@katit：我们需要这两个查询，而不仅仅是当前的查询，但所有查询是该事务的一部分。如果它每周发生三次，您可以[启用跟踪标志3604，然后是1204.]（http://www.sqlservercentral.com/articles/Performance+Tuning/tracingdeadlocks/1324/） – Andomar

解释死锁并修复它

回答

相关问题