Question

我们有一个简单的存储过程，该存储过程声明了几个变量，从一些查询计数中设置了变量，然后返回值。存储过程将在.5s内针对sql express（2017）在本地执行。本地数据库是完整的生产备份，已使用与生产相同的数据在本地还原。

当我们在Azure中对同一数据库执行存储过程时，查询将超时并无限期执行。

我尝试执行每个select语句（以下带有select count（distinct ..）的语句，每个语句将在Azure SQL上运行并在大约1s内正确返回计数，但是当我尝试使用声明或我尝试在存储过程中执行选择，如下所示，它们将无法完成。以下代码将无法在Azure上完全执行

SET ANSI_NULLS ON
GO
SET QUOTED_IDENTIFIER ON
GO
ALTER proc [MySchema].[UserCounts] (
     @HUserCount int out 
    ,@NUserCount int out
) as

set @HUserCount = (
    select count(distinct userTable.Id) as UserCount
    from [MySchema1].[Active] active with (nolock)
        inner join [MySchema2].[UserActive] userTable with (nolock) on active.UserId = userTable.Id
        inner join [MySchema2].[AppSession] appSession with (nolock) on appSession.UserId = userTable.Id
    where appSession.ComponentId = '1'
)

set @NUserCount = (
    select count(distinct userTable.Id) as SUserCount
    from [MySchema1].[Active] active with (nolock)
        inner join [MySchema2].[UserActive] userTable with (nolock) on active.UserId = userTable.Id
        inner join [MySchema2].[AppSession] appSession with (nolock) on appSession.UserId = userTable.Id
    where appSession.ComponentId = '2'
)

当我执行该过程时，以上代码将挂起/超时。另外，如果我只是从存储过程中拉出代码，它也将无法完成/超时。这是超时的类似代码。

declare @HUserCount int 
declare @NUserCount int

select @HUserCount = (
    select count(distinct userTable.Id) as UserCount
    from [MySchema1].[Active] active with (nolock)
        inner join [MySchema2].[UserActive] userTable with (nolock) on active.UserId = userTable.Id
        inner join [MySchema2].[AppSession] appSession with (nolock) on appSession.UserId = userTable.Id
    where appSession.ComponentId = '1'
)

select @NUserCount = (
    select count(distinct userTable.Id) as SUserCount
    from [MySchema1].[Active] active with (nolock)
        inner join [MySchema2].[UserActive] userTable with (nolock) on active.UserId = userTable.Id
        inner join [MySchema2].[AppSession] appSession with (nolock) on appSession.UserId = userTable.Id
    where appSession.ComponentId = '2'
)

以下代码将在Azure上按预期执行

select count(distinct userTable.Id) as UserCount
    from [MySchema1].[Active] active with (nolock)
        inner join [MySchema2].[UserActive] userTable with (nolock) on active.UserId = userTable.Id
        inner join [MySchema2].[AppSession] appSession with (nolock) on appSession.UserId = userTable.Id
    where appSession.ComponentId = '1'

以及

select count(distinct userTable.Id) as SUserCount
    from [MySchema1].[Active] active with (nolock)
        inner join [MySchema2].[UserActive] userTable with (nolock) on active.UserId = userTable.Id
        inner join [MySchema2].[AppSession] appSession with (nolock) on appSession.UserId = userTable.Id
    where appSession.ComponentId = '2'

我缺少明显的东西吗？这些表没有很多行，可能不超过50,000行，对于SQL Server而言几乎没有。

立即编辑/更新和解决方法

作为一个临时解决方案，我修改了存储的proc，以便它不使用set语句，然后再使用select count（distinct ... ...似乎select count（distinct作为独立的语句运行良好，但不能正常运行在set语句之后，该语句将变量设置为select count（distinct ..

这是结果代码/工作代码和经过修改的过程

SET ANSI_NULLS ON
GO
SET QUOTED_IDENTIFIER ON
GO
ALTER proc [MySchema].[UserCounts] (
     @HUserCount int out 
    ,@NUserCount int out
) as

select distinct userTable.Id into #UserCount
    from [MySchema1].[Active] active with (nolock)
        inner join [MySchema2].[UserActive] userTable with (nolock) on active.UserId = userTable.Id
        inner join [MySchema2].[AppSession] appSession with (nolock) on appSession.UserId = userTable.Id
    where appSession.ComponentId = '1'
set @HUserCount = (select count(*) from #UserCount)

select distinct userTable.Id into #SUserCount
    from [MySchema1].[Active] active with (nolock)
        inner join [MySchema2].[UserActive] userTable with (nolock) on active.UserId = userTable.Id
        inner join [MySchema2].[AppSession] appSession with (nolock) on appSession.UserId = userTable.Id
    where appSession.ComponentId = '2'
set @NUserCount = (select count(*) from #SUserCount)

DROP TABLE #UserCount
DROP TABLE #SUserCount

Azure挂起执行在本地立即运行的存储过程

0 个答案: