i group my events in transactions by user and day
...| transaction user day and then calculate duration, eventcount, time of transaction started and finished
in logs there are fields actiontype and actiontime
i want transaction to start with actiontype=login and actiontime for this action_type for this user and day is minimal,
i.e. i want start transaction since first time user login this day
something like this... | transaction user day startswith=min(actiontime for actiontype=login)
What commands should i write?
Thank a lot for answers!
... | transaction user day startswith(action_type="login")
This will start a transaction on first action_type="login" and not close it until the next day.
When you use startswith, you can have it be freeform text, an eval, or a valid search string. They have different syntax which is somewhat confusing in the documentation. The () indicates search syntax instead of simple text matching.
Thanks, but this way i will get number of transactions equal to number of user
s logins for this day, but i need 1 transaction with all user actions since first login for day
Because we don't specify an 'endswith' or 'maxpause' it won't close the transaction at all. Each transaction will start with the first user login and end on the last event for that user. It will not start subsequent transactions for the same user in the same day.
You could look at the
map command. Not much experience with that myself, but I believe that it could work here.
sourcetype=blah earliest=@d action_type=login | stats min(_time) as first_login by user | map search="search sourcetype=blah starttimeu=$first_login$ user=$user$ | table _time action_type user " | eval time = strftime(_time, "%Y-%m-%d %H:%M:%S") | eval time_action = time . " - " . action_type | stats list(time_action) as Time-Action by user
This should produce a list like;
user Time-Action ----------------------- bob 2013-04-25 11:22:33 - logout 2013-04-25 11:22:20 - change_value 2013-04-25 11:20:10 - login 2013-04-25 09:30:52 - logout 2013-04-25 09:26:14 - add_user 2013-04-25 09:21:36 - login caesar 2013-04-25 11:45:17 - logout 2013-04-25 11:35:10 - list_products 2013-04-25 11:17:39 - del_user 2013-04-25 11:09:29 - login
| reverse | just before the final
stats will list the
Time-Action in chronological order (per user).
This is perhaps not exactly what you want, and I'd be happy to see improvements to this query. Actually, it's the first time I've played with the
EDIT: Typo in the search. Sorry.