SQL, Get average job runtime from a log table - sql

Question

I have a table LOGS with this attributes:
ID(int)
Date(datetime)
Core(int)
Source(string)
Message(string)
That table contains log entries from multiple jobs.
A job has multiple log entries, but the start/end entries always have the same message.
Example:
->1, 07.12.2016 10:49:00, 2, Some DLL, Calling Execute on CleanUp // Start
2, 07.12.2016 10:49:01, 3, Other DLL, BLABLABLA
3, 07.12.2016 10:49:10, 1, Other DLL, BLABLABLA
->4, 07.12.2016 10:50:15, 2, Other DLL, BLABLABLA // Job does sth.
->5, 07.12.2016 10:50:50, 2, Other DLL, Execution completed // End
The rows marked with an arrow belonging to the same job.
As you can see, a job starts with 'Calling Execute...' and ends with 'Execution completed'.
What I want to achieve:
My task is to get the average job running times. The initial aproach was to filter with
WHERE Message LIKE '%JOBNAME%' OR Message LIKE 'Execution completed'
and comparing the dateTimes. This worked for some jobs, but some jobs run rarely so I only get "Execution completed" and the precision is not that great doing this manually.
At the end I want a list with following attributes:
ID(start log),
Start-Date,
End-Date,
Core,
Source-Start,
Source-End,
Message-Start,
Message-End
So later it's easy to calculate the difference and do the avg on it.
My idea
-> Get jobs by searching for a message.
-> Get a list with the message "Executing completed" having:
a higher ID (end log is always after start log)
a later datetime
the same core
For example:
Having a job with the attributes
1, 07.12.2016 11:33:00, 2, Source 1, Calling Execute on job Cleanup
Then searching for all logs with
ID>1,
dateTime>07.12.2016 11:33:00,
Core=2,
Message="Execution completed"
Picking the first item of that list should be the end log of the job
How can I do this with a sql query?
PS: I cannot change anything in the database, I can only read data.

Answer 1

You can identify the jobs using a correlated subquery to get the next end record. The following shows how to get these fields:
select l.*, lend.*
from (select l.*,
(select min(l2.date)
from logs l2
where l2.core = l.core and
l2.message like '% End'
l2.date > l.date
) as end_date
from logs l
where l.message like '% Start'
) l join
logs lend
on lend.core = l.core and lend.date = l.end_date;
This assumes that the date/time values are unique for a given "core".