Question

我们有一个每30秒运行一次的SQL作业（SQL Server 2012）

有时，没有任何理由，没有任何＆＃34;工作失败＆＃34;消息，作业停止甚至挂起（我们想要分析的行为）。

我们希望实现的是查询日志表并检测记录丢失的时间（不会每30秒发生一次）。

例如：

final case class AssociatedEvent(
  localAddress: Address,
  remoteAddress: Address,
  inbound: Boolean)
  extends AssociationEvent

final case class DisassociatedEvent(
  localAddress: Address,
  remoteAddress: Address,
  inbound: Boolean)
  extends AssociationEvent

我希望我的问题很明确。

Answer 1

这显示所有超过30秒的时段

select     A.CREATION_DATE as Start,  min(N.CREATION_DATE) as Finish
from       logTable  A
left join  logTable  N  on A.CREATION_DATE < N.CREATION_DATE
group by   A.CREATION_DATE
having     datediff(second, A.CREATION_DATE, min(N.CREATION_DATE)) > 30

Answer 2

这是我的“肮脏”解决方案，同时试图通过Luis使解决方案工作（看起来比我的好）

SELECT TOP 1000 [PROCEDURE_NAME]
      ,[CREATION_DATE]
      ,[MESSAGE]
      ,
       DATEDIFF(SECOND,
       (SELECT top 1 [CREATION_DATE] 
                    FROM [mactac].[mactac].[ERROR_LOG] f 
                    WHERE f.[CREATION_DATE]<c.[CREATION_DATE]
                    AND [PROCEDURE_NAME] like '%downl%' 
                    AND [MESSAGE]  LIKE '11 : **Fin**'
                      order by [CREATION_DATE] desc)
                    ,c.[CREATION_DATE]) as DIFF

  FROM [mactac].[mactac].[ERROR_LOG] as c
  WHERE [PROCEDURE_NAME] like '%downl%'
  AND [MESSAGE]  LIKE '11 : **Fin**'
  order by [CREATION_DATE] desc

此解决方案显示所有记录和已用时间:-D 但是......我还没想出如何添加：

AND DIFF＆gt; 35

Invalid column name 'DIFF'.

为了过滤更长的时间

Answer 3

使用CTE可以很容易。

with ordered_logs (creation_date, rnk) as
(
  select creation_date, row_number() over (order by creation_date)
  from your_log_table
)
select first.creation_date, second.creation_date, datediff(ss, first.creation_date, second.creation_date) as SecondsDifference
from ordered_logs first
inner join ordered_logs second on (first.rnk + 1) = second.rnk
where datediff(ss, first.creation_date, second.creation_date) > 30

按照创建日期对所有记录进行排序，然后将前一个记录与下一个记录连接起来（按时间顺序）。然后它只检查创建时间的差异是否超过30秒。

SQL：检测何时丢失重复日志（每30秒）

3 个答案: