自加入 SQL 具有聚合的服务器

Question

我有一个查询，returns 所有会话都在当天开始，然后给我当前活动会话的计数和自上次登录以来的时间。（这让我知道是否有内核出现任何导致用户无法登录的问题）

SELECT [server]
     , appserver
     , count(*) as open_sessions
     , max(start_time) Latest_New_Login
     , datediff(minute, max(start_time) , GETDATE() ) elapsed_mins_since_last_login
FROM [session_logs]
where DATEDIFF(day, START_TIME, GETDATE()) <= 0  
and end_time is null
group by server, appserver
order by 1,2

数据返回

+---------+-----------+---------------+-------------------------+-------------------------------+
| server  | appserver | open_sessions |    Latest_New_Login     | Elapsed_mins_since_last_login |
+---------+-----------+---------------+-------------------------+-------------------------------+
| SERVER1 | PK1       |            13 | 2018-01-26 16:24:47.087 |                             1 |
| SERVER1 | PK2       |            12 | 2018-01-26 16:19:00.157 |                             6 |
| SERVER2 | PK1       |             8 | 2018-01-26 16:25:41.297 |                             0 |
| SERVER2 | PK2       |             6 | 2018-01-26 16:07:31.027 |                            18 |
| SERVER3 | PK1       |             7 | 2018-01-26 16:18:56.597 |                             7 |
| SERVER3 | PK2       |             5 | 2018-01-26 16:19:06.550 |                             6 |
+---------+-----------+---------------+-------------------------+-------------------------------+

一切都很好，但是我需要一种解决方法来确保显示所有服务器和应用程序服务器。今天我们遇到了一个问题，服务器 3 PK1 上没有显示任何活动登录，这从我的结果中删除了它的条目 table。

有没有办法让我始终显示以下数据集并让我的聚合函数与它们相符？一个自连接查询，但我无法让它工作。

select distinct [server], appserver
FROM [session_logs]

returns 值

Server 1 PK1
Server 1 PK2
Server 2 PK1
Server 2 PK2
Server 3 PK1
Server 3 PK2

Answer 1

做一个LEFT JOIN:

SELECT
    serv.[server],
    serv.appserver,
    t.open_sessions,
    t.Latest_New_Login,
    t.elapsed_mins_since_last_login
FROM (
    SELECT DISTINCT [server], appserver FROM [session_logs]
) serv
LEFT JOIN (
    SELECT [server]
         , appserver
         , COUNT(*) AS open_sessions
         , MAX(start_time) AS Latest_New_Login
         , DATEDIFF(MINUTE, MAX(start_time), GETDATE()) AS elapsed_mins_since_last_login
    FROM [session_logs]
    WHERE
        DATEDIFF(day, START_TIME, GETDATE()) <= 0  
        AND end_time is null
    GROUP BY [server], appserver
) t
    ON t.[server] = serv.[server]
    AND t.appserver = serv.appserver
ORDER BY
    serv.[server], serv.appserver

如果您有一个 table 来保存您的服务器列表，请使用它来代替：

SELECT DISTINCT [server], appserver FROM [session_logs]

Answer 2

不使用的另一种选择JOIN

select
    [server]
    , appserver
    , count(start_time) as open_sessions
    , max(start_time) Latest_New_Login
    , datediff(minute, max(start_time) , GETDATE() ) elapsed_mins_since_last_login
from (
    SELECT [server]
         , appserver
        , start_time
    FROM 
        [session_logs]
    where 
        DATEDIFF(day, START_TIME, GETDATE()) <= 0  
        and end_time is null
    union all select distinct [server], appserver, null FROM [session_logs]
) t
group by server, appserver

Answer 3

构建一个包含所有唯一服务器的通用 Table 表达式，并在其上进行 LEFT JOIN：

;with cte([server], appserver)
as (
    select distinct [server], appserver
    from [session_logs]
    -- let's limit with records from last 30 days in case [session_logs] is a massive table
    -- I assume all servers you care about had some activity in that time
    where DATEDIFF(day, START_TIME, GETDATE()) <= 30
)
SELECT cte.[server]
     , cte.appserver
     , count(start_time) as open_sessions
     , max(start_time) Latest_New_Login
     , datediff(minute, max(start_time), GETDATE()) Elapsed_mins_since_last_login
FROM cte
left join [session_logs] as sl on cte.[server] = sl.[server] and cte.appserver = sl.appserver
where DATEDIFF(day, START_TIME, GETDATE()) <= 0  
and end_time is null
group by cte.[server], cte.appserver
order by cte.[server], cte.appserver

自加入 SQL 具有聚合的服务器

Self Join SQL Server with aggregates

sql-server

self-join