SQL Server阻塞的检查
1. 阻塞
除了内存、cpu、i/o这些系统资源以外,阻塞和死锁是影响数据库应用性能的另一大因素。
所谓的「阻塞」,是指当一个数据库会话中的事务,正在锁定其他会话事务想要读取或修改的资源,造成这些会话发出的请求进入等待的状态。sql server 默认会让被阻塞的请求无限期地一直等待,直到原来的事务释放相关的锁,或直到它超时、服务器关闭、进程被杀死。一般的系统中,偶尔有短时间的阻塞是正常且合理的;但若设计不良的程序,就可能导致长时间的阻塞,这样就不必要地锁定了资源,而且阻塞了其他会话欲读取或更新的需求。遇到这种情况,可能就需要手工排除阻塞的状态。
2.阻塞和死锁可能带来的问题
(1)并发用户少的时候,一切还都正常。但是随着并发用户的增加,性能越来越慢。
(2)应用程序运行很慢,但是sql server 这个cpu和磁盘利用率很低。
(3)客户端经常受到以下错误。
error 1222--lock request time out period exceeded.
error 1205--your transaction(process id #xx) was deadlocked on resources with another process and has been chosen as the deadlock victim. return your transaction.
超时错误--timeout expired. the timeout period elapsed prior to completion of the operation or the server is not responding.
(4)有些查询能够进行,但是有些特定的查询或修改总是不能返回。
(5)重启sql server就能解决。但是有可能跑一段时间以后又会出现问题。
3.阻塞的检查
3.1 主要的系统表或函数
sys.sysprocesses 系统表是一个很重要的系统视图,主要用来定位与解决sql server的阻塞和死锁。主要字段1.spid:sql servr 会话id 2.blocked:正在阻塞求情的会话 id。如果此列为 null,则标识请求未被阻塞 3. program_name:应用程序的名称,就是 连接字符串中配的 application name 4. hostname:建立链接的客户端工作站的名称。
sys.dm_exec_requests、sys.dm_exec_sql_text返回指定spider的 sql 查询文本。
dbcc inputbuffer 显示从客户端发送到 microsoft sql server 实例的最后一个语句。
sp_lock 系统存储过程,报告有关锁的信息。
3.2 check逻辑
对应的存储为dblockcheck(job为db_lockcheck),主要check逻辑如下:
3.3 保存的数据
所收集的数据保存dblock_information中,主要包含信息如截图,定期的统计分析可获得经常被阻塞和引起阻塞sql语句和table,这些信息是进行数据库优化的一个角度。
select top 100* from dblock_information order by transdatetime desc
4.代码实现
4.1 table的创建脚本
create table [dbo].[dblock_information]( [id] [int] identity(1,1) not null, [message] [nvarchar](300) null, [locktype] [char](1) null, [spid1] [int] null, [spid2] [int] null, [eventtype] [nvarchar](100) null, [parameters] [nvarchar](10) null, [eventinfo] [nvarchar](3000) null, [individualquery] [nvarchar](1000) null, [transdatetime] [datetime] null constraint [df_dblock_information_transdatetime] default (getdate()), [appname] [varchar](50) null, [hostname] [varchar](50) null, constraint [pk_dblock_information] primary key clustered ( [id] asc )with (pad_index = off, statistics_norecompute = off, ignore_dup_key = off, allow_row_locks = on, allow_page_locks = on) on [primary] ) on [primary]
4.2 存储的创建脚本
create procedure [dbo].[dblockcheck] @alarmthreshold as tinyint=10 as set nocount on -------------------------------------------------------------------------------------------------------- --*program*: <dblock check for job> --*programer*:<> --*date*:<> --*description*:<query sql locking process> --*unify*:<ua> --########## parameter description begin ########## --########## parameter description end # ########## --##########update log begin ################### --##########update log end # ################### -------------------------------------------------------------------------------------------------------- declare @sql as varchar(200) declare @subject as varchar(200) declare @body as nvarchar(max) declare @spname as nvarchar(max) declare @message as nvarchar(200) declare @dbname varchar(15) declare @ip varchar(20) declare @cnt as int declare @cnt2 int declare @individualquery nvarchar(1000) declare @hostname varchar(50) declare @appname varchar(50) set @dbname=db_name() select @ip='xxx.xxx.xxx.xxx' ----不手动定义ip也可通过以下函数来实现 declare @serverip nvarchar(30)='', @servername nvarchar(60)='' select top 1 @servername = @@servername ,@serverip=local_net_address from sys.dm_exec_connections where local_net_address is not null -------- begin declare @spid int,@bl int, @inttransactioncountonentry int, @introwcount int, @intcountproperties int, @intcounter int create table #tmp_lock_who ( id int identity(1,1), spid smallint, bl smallint) create table #tmp_lock_information ( id int identity(1,1), message nvarchar(200), locktype char(1), spid1 int, spid2 int, eventtype nvarchar(100), parameters nvarchar(10), eventinfo nvarchar(max), individualquery nvarchar(1000), appname varchar(50), hostname varchar(50) ) if @@error<>0 return @@error insert into #tmp_lock_who(spid,bl) select 0 ,blocked from (select * from master..sysprocesses where blocked>0 ) a where not exists(select * from (select * from master..sysprocesses where blocked>0 ) b where a.blocked=spid) union select spid,blocked from master..sysprocesses where blocked>0 if @@error<>0 return @@error -- 找到临时表的记录数 select @intcountproperties = count(*),@intcounter = 1 from #tmp_lock_who if @@error<>0 return @@error if @intcountproperties=0 select n'现在没有阻塞信息!' as message -- 循环开始 while @intcounter <= @intcountproperties begin -- 取第一条记录 select @spid = spid,@bl = bl from #tmp_lock_who where id = @intcounter begin select @individualquery= substring (qr.text,qs.statement_start_offset/2, (case when qs.statement_end_offset = -1 then len(convert(nvarchar(max), qr.text)) * 2 else qs.statement_end_offset end - qs.statement_start_offset)/2) from sys.dm_exec_requests qs outer apply sys.dm_exec_sql_text(qs.sql_handle) as qr where qr.text is not null and qs.session_id=@bl select @hostname=left(hostname,50),@appname=left(program_name,50) from master..sysprocesses with(nolock) where spid=@bl set @sql='dbcc inputbuffer ('+cast(@bl as char(20))+')' if @spid =0 begin select @message=n'引起数据库阻塞的是: '+ cast(@bl as nvarchar(100)) + n'进程号,其执行的sql语法如下' --set @sql='dbcc inputbuffer ('+cast(@bl as char(20))+')' insert into #tmp_lock_information(eventtype,parameters,eventinfo) exec(@sql) update #tmp_lock_information set locktype='1',spid1=@bl,spid2=@bl,individualquery=@individualquery,[message]=@message,appname=@appname,hostname=@hostname where [message] is null end else begin select @message=n'进程号spid:'+ cast(@spid as nvarchar(100))+ n'被' + n'进程号spid:'+ cast(@bl as nvarchar(10)) +n'阻塞,其当前进程执行的sql语法如下' insert into #tmp_lock_information(eventtype,parameters,eventinfo) exec(@sql) update #tmp_lock_information set locktype='2', spid1=@spid,spid2=@bl,individualquery=@individualquery,[message]=@message,appname=@appname,hostname=@hostname where [message] is null end end -- 循环指针下移 set @intcounter = @intcounter + 1 end drop table #tmp_lock_who if exists(select 0 from #tmp_lock_information) begin insert into dblock_information(message,locktype,spid1,spid2,eventtype,parameters,eventinfo,individualquery,appname,hostname) select [message],locktype,spid1,spid2,eventtype,parameters,substring(eventinfo,1,500),individualquery,appname,hostname from #tmp_lock_information end drop table #tmp_lock_information return 0 end
推荐阅读