I am trying to use transactions to better summarize what is going on in sessions.
sourcetype="blah" response="200" session_id!="-" | transaction session_id maxspan=30m
Thus, a session is defined to have no more than 30 minutes of idle time between individual events.
We would like to record some summary stats to the summary index for 1 hour spans. The problem, however, is that the the session may actally start before that hour is up and thus be excluded.
And it will be excluded in the following hour as well as its events did not all occur in the 1 hour interval.
What is the best practice when trying to do something like this?
You may want to look at the concurrency search command, if you're trying to just count the number of transactions running in a span: http://www.splunk.com/base/Documentation/latest/SearchReference/Concurrency
If you're doing more than that, I'm not sure, it would depend exactly what you were trying and what the data looked like.