zoukankan      html  css  js  c++  java
  • 死锁的监控分析

    死锁的监控分析

    1 背景

    1.1 报警情况

    最近整理笔记,打算全部迁移到EVERNOTE。整理到锁这一部分,里边刚好有个自己记录下来的案例,重新整理分享下给大家。

    某日中午,收到报警短信,DB死锁异常,单分钟死锁120个。

    死锁的xml文件如下:

    复制代码
     1 <deadlock-list>
     2 <deadlock victim="process810b00cf8">
     3 <process-list>
     4 <process id="process810b00cf8" taskpriority="0" logused="0" waitresource="RID: 13:1:1541136:62" waittime="7682" ownerId="3396587959" transactionname="UPDATE" lasttranstarted="2016-01-08T12:03:51.067" XDES="0xa99746d08" lockMode="U" schedulerid="41" kpid="17308" status="suspended" spid="108" sbid="0" ecid="0" priority="0" trancount="2" lastbatchstarted="2016-01-08T12:03:51.067" lastbatchcompleted="2016-01-08T12:03:51.067" lastattention="1900-01-01T00:00:00.067" clientapp="Microsoft SQL Server Management Studio - 查询" hostname="test-server" hostpid="1433" loginname="xinysu" isolationlevel="read committed (2)" xactid="3396587959" currentdb="13" lockTimeout="4294967295" clientoption1="671098976" clientoption2="390200">
     5 <executionStack>
     6 <frame procname="adhoc" line="7" stmtstart="214" stmtend="484" sqlhandle="0x020000003acf4f010561e479685209fb09a7fd15239977c60000000000000000000000000000000000000000">
     7 UPDATE FinanceReceiptNoRule SET NowSeqValue=@ReturnNum,ISRUNNING='0',LastWriteTime=GETDATE() WHERE IsRunning='1' AND SeqCode=@SeqCode </frame>
     8 </executionStack>
     9 <inputbuf>
    10 declare @SeqCode varchar(60)
    11 declare @ReturnNum bigint
    12 set @SeqCode='CGJS20160106'
    13 while(1=1)
    14 begin
    15 UPDATE FinanceReceiptNoRule SET NowSeqValue=@ReturnNum,ISRUNNING='0',LastWriteTime=GETDATE() WHERE IsRunning='1' AND SeqCode=@SeqCode
    16 end </inputbuf>
    17 </process>
    18 <process id="process18fd5d8cf8" taskpriority="0" logused="248" waitresource="KEY: 13:72057594040090624 (b3ade7c5980c)" waittime="4" ownerId="3396522828" transactionname="user_transaction" lasttranstarted="2016-01-08T12:03:05.310" XDES="0x18c1db63a8" lockMode="U" schedulerid="57" kpid="16448" status="suspended" spid="161" sbid="0" ecid="0" priority="0" trancount="2" lastbatchstarted="2016-01-08T12:03:58.737" lastbatchcompleted="2016-01-08T12:03:33.847" lastattention="2016-01-08T12:03:33.850" clientapp="Microsoft SQL Server Management Studio - 查询" hostname="test-server" hostpid="1433" loginname="xinysu" isolationlevel="read committed (2)" xactid="3396522828" currentdb="13" lockTimeout="4294967295" clientoption1="671090784" clientoption2="390200">
    19 <executionStack>
    20 <frame procname="adhoc" line="6" stmtstart="210" stmtend="400" sqlhandle="0x020000001b4f23368af7bba99098c10dec46585804f1b4ce0000000000000000000000000000000000000000">
    21 Update dbo.FinanceReceiptNoRule Set [IsRunning]='1' where SeqCode=@SeqCode and IsRunning='0' </frame>
    22 </executionStack>
    23 <inputbuf>
    24 declare @SeqCode varchar(60)
    25 declare @ReturnNum bigint
    26 set @SeqCode='CGJS20160106'
    27 while(1=1)
    28 begin
    29 Update dbo.FinanceReceiptNoRule Set [IsRunning]='1' where SeqCode=@SeqCode and IsRunning='0' 
    30 end
    31 </inputbuf>
    32 </process>
    33 </process-list>
    34 <resource-list>
    35 <ridlock fileid="1" pageid="1541136" dbid="13" objectname="fin_test.dbo.FinanceReceiptNoRule" id="lock51e8a3980" mode="X" associatedObjectId="72057594040025088">
    36 <owner-list>
    37 <owner id="process18fd5d8cf8" mode="X" />
    38 </owner-list>
    39 <waiter-list>
    40 <waiter id="process810b00cf8" mode="U" requestType="wait" />
    41 </waiter-list>
    42 </ridlock>
    43 <keylock hobtid="72057594040090624" dbid="13" objectname="fin_test.dbo.FinanceReceiptNoRule" indexname="PK_FINANCERECEIPTNORULE" id="lock7b2c6bc80" mode="U" associatedObjectId="72057594040090624">
    44 <owner-list>
    45 <owner id="process810b00cf8" mode="U" />
    46 </owner-list>
    47 <waiter-list>
    48 <waiter id="process18fd5d8cf8" mode="U" requestType="wait" />
    49 </waiter-list>
    50 </keylock>
    51 </resource-list>
    52 </deadlock>
    53 </deadlock-list>
    复制代码

    表格结构跟模拟数据如下:

    复制代码
     1 --涉及表格:
     2 CREATE TABLE [dbo].[FinanceReceiptNoRule](
     3 [SeqCode] [varchar](60) NOT NULL,
     4 [NowSeqValue] [bigint] NULL,
     5 [SeqDate] [varchar](14) NOT NULL,
     6 [IsRunning] [varchar](1) NULL,
     7 [LastWriteTime] [datetime] NULL,
     8 [Prefix] [varchar](4) NULL
     9 ) ON [PRIMARY]
    10 GO
    11 --数据模拟
    12 INSERT [dbo].[FinanceReceiptNoRule] ([SeqCode], [NowSeqValue], [SeqDate], [IsRunning], [LastWriteTime], [Prefix]) VALUES (N'TEST20150108', 1469, N'20150108', N'0', CAST(N'2015-01-08 05:05:49.163' AS DateTime), N'TEST')
    13 GO
    14 INSERT [dbo].[FinanceReceiptNoRule] ([SeqCode], [NowSeqValue], [SeqDate], [IsRunning], [LastWriteTime], [Prefix]) VALUES (N'TEST20150109', 1377, N'20150109', N'0', CAST(N'2015-01-09 04:50:26.610' AS DateTime), N'TEST')
    15 GO
    16  
    17 ALTER TABLE [dbo].[FinanceReceiptNoRule] ADD CONSTRAINT [pk_FinanceReceiptNoRule] PRIMARY KEY NONCLUSTERED 
    18 (
    19 [SeqCode] ASC
    20 )WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, SORT_IN_TEMPDB = OFF, IGNORE_DUP_KEY = OFF, ONLINE = OFF, ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON) ON [PRIMARY]
    21 GO
    复制代码

    1.2 如何监控

    捕获死锁有多种方式可以捕获,这里介绍2种:SQL SERVER Profiler工具跟Extended Events。Profiler相对比较耗资源,但是由于只监控死锁这一项,所以性能影响不是很大,其可视化界面较易上手;Extended Events耗费资源较少,实时记录到倒数第二个死锁,同时需要SQL语句来分析查询记录文件。

    如何使用 Profiler监控?
    打开 SSMS,点击<工具>,选择 <SQL Server Profiler>,如下图。

    登录到需要监控的DB实例,填写相应的跟踪属性,首先是<常规>页面,如下图。这里注意2个方面,第一,选择 <TSQL-Locks>模板,这个模板即可以用来监控死锁,也可以拿来观察 锁申请与释放情况,非常详细,有事没事可以多拿来看SELECT UPDATE DELETE等语句对锁的申请及释放情况;第二,监控结果存储,建议可以存放到某个表格中去,方便定期分析与统计。

    接着填写<事件选择>项,只需要选择 <deadlock graph> Events,其他都不需要打勾,最后点击运行就可以开始监控了。

     

    可以用一个万年常用的例子来检查是否监控正常,开3个查询窗口,按照以下顺序执行则会发生资源占用及申请互斥导致死锁,执行完第5步,等待1-3s则发生死锁。脚本提供如下:

     模拟死锁SQL

    监控到的死锁界面如下:

    如何使用Extended Events监控?

    建立扩展事件监控的脚本如下:(扩展事件很赞,2012版支持可视化操作,感兴趣的可以上 MSDN了解:https://msdn.microsoft.com/zh-cn/library/bb630282.aspx,本文就不分析语法等知识点了)

    复制代码
    1 CREATE EVENT SESSION [DeadLock] ON SERVER 
    2 ADD EVENT sqlserver.xml_deadlock_report 
    3 ADD TARGET package0.event_file(SET filename=N'F:eventsdeadlockdeadlock.xel',max_file_size=(20)),
    4 ADD TARGET package0.ring_buffer(SET max_events_limit=(100),max_memory=(10240),occurrence_number=(50))
    5 WITH (MAX_MEMORY=4096 KB,EVENT_RETENTION_MODE=ALLOW_SINGLE_EVENT_LOSS,MAX_DISPATCH_LATENCY=30 SECONDS,MAX_EVENT_SIZE=0 KB,MEMORY_PARTITION_MODE=NONE,TRACK_CAUSALITY=OFF,STARTUP_STATE=ON)
    6 GO
    复制代码

    查询SQL如下,这里需要注意:查询是基于buffer还是基于filer分析,一般buffer存储的个数都是有限的,比如上文我们只分配了4M存储,file分析则是完整的,但是要看保留的文件个数。这里我们给出buffer的查询SQL如下,file的查询大家感兴趣的可以动手写下。

    复制代码
    DECLARE @deadlock_xml XML
    SELECT @deadlock_xml=(
                           SELECT
                                  (
                                    SELECT
                                          CONVERT(XML, target_data)
                                    FROM sys.dm_xe_session_targets st
                                    JOIN sys.dm_xe_sessions s ON s.address = st.event_session_address
                                    WHERE s.name = 'deadlock' AND st.target_name = 'ring_buffer'
                                  ) AS [x]
                           FOR XML PATH('') , TYPE
                          )

    SELECT dateadd(hour,+6,tb.col.value('@timestamp[1]','varchar(max)')) TimePoint, tb.col.value('(data/value/deadlock/process-list/process/executionStack/frame)[1]','VARCHAR(MAX)') statement_parameter_k, tb.col.value('(data/value/deadlock/process-list/process/executionStack/frame)[2]','VARCHAR(MAX)') statement_k, tb.col.value('(data/value/deadlock/process-list/process/executionStack/frame)[3]','VARCHAR(MAX)') statement_parameter, tb.col.value('(data/value/deadlock/process-list/process/executionStack/frame)[4]','VARCHAR(MAX)') [statement], tb.col.value('(data/value/deadlock/process-list/process/@waitresource)[1]','VARCHAR(MAX)') waitresource_k, tb.col.value('(data/value/deadlock/process-list/process/@waitresource)[2]','VARCHAR(MAX)') waitresource, tb.col.value('(data/value/deadlock/process-list/process/@isolationlevel)[1]','VARCHAR(MAX)') isolationlevel_k, tb.col.value('(data/value/deadlock/process-list/process/@isolationlevel)[2]','VARCHAR(MAX)') isolationlevel, tb.col.value('(data/value/deadlock/process-list/process/@waittime)[1]','VARCHAR(MAX)') waittime_k, tb.col.value('(data/value/deadlock/process-list/process/@waittime)[2]','VARCHAR(MAX)') waittime, tb.col.value('(data/value/deadlock/process-list/process/@clientapp)[1]','VARCHAR(MAX)') clientapp_k, tb.col.value('(data/value/deadlock/process-list/process/@clientapp)[2]','VARCHAR(MAX)') clientapp, tb.col.value('(data/value/deadlock/process-list/process/@hostname)[1]','VARCHAR(MAX)') hostname_k, tb.col.value('(data/value/deadlock/process-list/process/@hostname)[2]','VARCHAR(MAX)') hostname FROM @deadlock_xml.nodes('//event') as tb(col)
    复制代码

    这个SQL可以查询的出非常详细的资源争夺情况,如果想要有效的使用扩展事件,建议大家详细查看下官网的xml语法(SQL SERVER对xml的支持也是棒棒哒,期待2016版中的json支持)

    是不是很清晰,一目了然,有了这个就可以去分析拉!

    2 分析

    根据xml文件内容或者扩展事件的监控内容,都可以整理为以下信息(开头的那个死锁分析):

    查看事务1及事务2的执行计划如下:

     结合表格及执行计划,可以大致推测死锁过程:

    会话1:

    • 根据主键SeqCode查找到键值所在的 索引页 Index_Page,找到该页上面的 keyhashvalue 键值行 Index_key,对Index_Page持有IU锁,对Index_key持有U锁;
    • 由于该表是堆表,bookmark lookup是通过 RID查找 ,即通过行标识符查找,找到RID所对应的行数据所在的 数据页  Data_Page,然后在该页面上找到RID指向槽号上的行数据,对该行数据持有U锁;
    • 这个时候,已经查找到了需要更新的行数据,可以把数据页 Data_Page上的IU锁 升级为IX锁,RID指向的行数据 从U锁升级为X锁,升级结束后,释放索引页跟键值行上面的 IU锁及U锁。
    • 则此时,会话1 持有 Data_Page 上的IX锁、RID行上的 X锁.

    这个过程中,刚好会话2进行这样的锁申请:

    • 找出事务2中持有锁资源是哪个索引,可以根据sys.partitions 可以查看到72057594038910976是主键pk_FinanceReceiptNoRule,主键列是:SeqCode。
    • 根据主键SeqCode查找到键值所在的 索引页 Index_Page,找到该页上面的 键值行 Index_key,对Index_Page持有IU锁,对Index_key持有U锁;
    • 由于该表是堆表,bookmark lookup是通过 RID查找 ,即通过行标识符查找,找到RID所对应的行数据所在的 数据页  Data_Page,然后在该页面上找到RID指向槽号上的行数据,准备该行数据持有U锁,但是发现RID行上被会话1持有了X锁,导致其申请 U锁 Timeout。
    • 则此时 会话2 持有 Index_Page上的IU锁、Index_key上的U锁、Data_Page上的IU锁,请求 RID行的 U锁。

    假设这个时候,会话1 中又执行了一次update操作(同一个事务中):

    • 根据主键SeqCode查找到键值所在的 索引页 Index_Page,找到该页上面的 键值行 Index_key,对Index_Page持有IU锁,准备对Index_key持有U锁,但是发现 Index_key被会话2持有了U锁。

    那么这个时候死锁就产生了(详见下图):

    • 会话1 持有 Data_Page 上的IX锁、RID行上的 X锁,申请 Index_key 的U锁(等待会话2释放)
    • 会话2 持有 Index_Page上的IU锁、Index_key上的U锁、Data_Page上的IU锁,请求 RID行的 U锁(等待会话1释放)

    3 解决

    想法子除去RID查找,直接index就找到数据,就不会发生这个死锁,也就是,在主键上面重新建立聚集索引,丢弃原先的非聚集索引主键。因为这样排除了RID的U锁申请与持有,直接是保持X锁 直至事务结束,同时可以直接根据主键来修改键值所在的数据页,减少的RID查询行的时间。

    修改后的执行计划如下:

    其锁申请释放的流程如下(详见截图):

    • 根据主键SeqCode查找到键值所在的 索引页 Index_Page,找到该页上面的 keyhashvalue 键值行 Index_key,对Index_Page持有IU锁,对Index_key持有U锁;
    • 由于该表已经是聚集索引表,主键所在的页上包含 行数据,则可以直接 对Index_Page持有IU锁升级为IX锁,对Index_key持有U锁升级为X锁,避免了RID逐个找行数据的锁申请

     
    分类: SQL SERVER
  • 相关阅读:
    leetcode 309. Best Time to Buy and Sell Stock with Cooldown
    leetcode 714. Best Time to Buy and Sell Stock with Transaction Fee
    leetcode 32. Longest Valid Parentheses
    leetcode 224. Basic Calculator
    leetcode 540. Single Element in a Sorted Array
    leetcode 109. Convert Sorted List to Binary Search Tree
    leetcode 3. Longest Substring Without Repeating Characters
    leetcode 84. Largest Rectangle in Histogram
    leetcode 338. Counting Bits
    git教程之回到过去,版本对比
  • 原文地址:https://www.cnblogs.com/Leo_wl/p/6539091.html
Copyright © 2011-2022 走看看