按几列分组,包括 MAX() 以限制行,但显示不在分组依据中的列
Group by a few columns including MAX() to limit rows but display columns not in group by
我每 5 分钟将 sp_whoisactive 输出到一个日志 table 中,并且想要聚合诸如已用时间或 cpu 之类的列。我获得唯一记录的最佳机会是 session_id、login_time、login_name,我希望每条唯一记录的 MAX(collection_time)。
我的问题是如何在一个查询中执行此操作?
我尝试了各种变体,但始终无法获得所需的输出。我最终执行了 2 个步骤 1) 按 ##Temp table 分组和 2) 然后将 ##temp 加入日志 table 以获得报告中我想要的所有列。
第 1 步查询:
enter code hereIF(OBJECT_ID('tempdb..##TEMP') IS NOT NULL) BEGIN DROP TABLE ##TEMP END
SELECT *
INTO ##TEMP
FROM (
select distinct Session_id, login_name, login_time, max(collection_time) as MaxCollection
from [MyDB].[DBA].[WhoIsActive]
group by login_name, session_id, login_time
) AS X
第 2 步查询:
DECLARE @yesterday DATETIME, @today DATETIME
SET @yesterday = DATEADD(hh,12,DATEADD(dd,DATEDIFF(dd,0,GETDATE()-1),0))
SET @today = DATEADD(hh,11,DATEADD(dd,DATEDIFF(dd,0,GETDATE()-0),0))
SELECT
tt.[session_id]
,tt.[tasks]
,tt.[login_time]
,tt.[login_name]
,tt.[start_time]
,tt.[collection_time]
,tt.[dd hh:mm:ss.mss] as ElapsedTime
,(Substring(tt.[dd hh:mm:ss.mss],1,2) * 86400) + (Substring(tt.[dd hh:mm:ss.mss],4,2) * 3600) + (Substring(tt.[dd hh:mm:ss.mss],7,2) * 60) + (Substring(tt.[dd hh:mm:ss.mss],10,2)) as totalseconds
,tt.[host_name]
,tt.[status]
,tt.[CPU]
,tt.[program_name]
FROM [MyDB].[DBA].[WhoIsActive] tt
right JOIN ##TEMP kk
on tt.session_id = kk.session_id and tt.collection_time = kk.MaxCollection
--where kk.start_time between @yesterday and @today --and kk.login_name like '%ETL%'
order by tt.login_name asc, tt.session_id asc, login_time desc
My best chance for unique record is session_id
, login_time
, login_name
and I want the MAX(collection_time)
for each unique record.
如果你只想要每个元组的最大收集时间,那么使用聚合:
select session_id, login_time, login_name, max(collection_time) max_collection_time
from whoisactive
group by session_id, login_time, login_name
另一方面,如果您想要 整个 对应的行,那么一个选项是 row_number()
:
select *
from (
select w.*,
row_number() over(partition by session_id, login_time, login_name order by collection_time desc) rn
from whoisactive w
) t
where rn = 1
我每 5 分钟将 sp_whoisactive 输出到一个日志 table 中,并且想要聚合诸如已用时间或 cpu 之类的列。我获得唯一记录的最佳机会是 session_id、login_time、login_name,我希望每条唯一记录的 MAX(collection_time)。
我的问题是如何在一个查询中执行此操作?
我尝试了各种变体,但始终无法获得所需的输出。我最终执行了 2 个步骤 1) 按 ##Temp table 分组和 2) 然后将 ##temp 加入日志 table 以获得报告中我想要的所有列。
第 1 步查询:
enter code hereIF(OBJECT_ID('tempdb..##TEMP') IS NOT NULL) BEGIN DROP TABLE ##TEMP END
SELECT *
INTO ##TEMP
FROM (
select distinct Session_id, login_name, login_time, max(collection_time) as MaxCollection
from [MyDB].[DBA].[WhoIsActive]
group by login_name, session_id, login_time
) AS X
第 2 步查询:
DECLARE @yesterday DATETIME, @today DATETIME
SET @yesterday = DATEADD(hh,12,DATEADD(dd,DATEDIFF(dd,0,GETDATE()-1),0))
SET @today = DATEADD(hh,11,DATEADD(dd,DATEDIFF(dd,0,GETDATE()-0),0))
SELECT
tt.[session_id]
,tt.[tasks]
,tt.[login_time]
,tt.[login_name]
,tt.[start_time]
,tt.[collection_time]
,tt.[dd hh:mm:ss.mss] as ElapsedTime
,(Substring(tt.[dd hh:mm:ss.mss],1,2) * 86400) + (Substring(tt.[dd hh:mm:ss.mss],4,2) * 3600) + (Substring(tt.[dd hh:mm:ss.mss],7,2) * 60) + (Substring(tt.[dd hh:mm:ss.mss],10,2)) as totalseconds
,tt.[host_name]
,tt.[status]
,tt.[CPU]
,tt.[program_name]
FROM [MyDB].[DBA].[WhoIsActive] tt
right JOIN ##TEMP kk
on tt.session_id = kk.session_id and tt.collection_time = kk.MaxCollection
--where kk.start_time between @yesterday and @today --and kk.login_name like '%ETL%'
order by tt.login_name asc, tt.session_id asc, login_time desc
My best chance for unique record is
session_id
,login_time
,login_name
and I want theMAX(collection_time)
for each unique record.
如果你只想要每个元组的最大收集时间,那么使用聚合:
select session_id, login_time, login_name, max(collection_time) max_collection_time
from whoisactive
group by session_id, login_time, login_name
另一方面,如果您想要 整个 对应的行,那么一个选项是 row_number()
:
select *
from (
select w.*,
row_number() over(partition by session_id, login_time, login_name order by collection_time desc) rn
from whoisactive w
) t
where rn = 1