Since the code will only trigger this query on accounting on and off packets, perhaps we should set the AcctStatusType = '%{Acct-Status-Type}' which should result in "Accounting-On" or "Accounting-Off" I just don't like the idea of leaving it set to "Start", and I do agree setting it to "Stop" would be misleading... Other suggestions? -----Original Message----- From: freeradius-devel-bounces+duanec=mail.illicom.net@lists.freeradius.org [mailto:freeradius-devel-bounces+duanec=mail.illicom.net@lists.freeradius.or g] On Behalf Of Guy Fraser Sent: Monday, December 17, 2007 5:28 PM To: FreeRadius developers mailing list Subject: Re: reasoning/question about accounting_onoff_query On 2007-Dec-17, at 15:33, Peter Nixon wrote:
On Mon 17 Dec 2007, Duane Cox wrote:
Correct me if I'm wrong...
But I believe the "accounting_onoff_query" is to be executed by a rebooting nas, or other similiar situation, so that all accounts that are currently "online" or specifiically all accounting records that are open (START), are to be closed, only accounts pertaining to that specific nas.. Well that's how I read the conf file.
If so, souldn't the query also update the AcctStatusType to "Stop" ??
I've looked over all the db examples, and they are pretty much the same, update the StopTime, SessionTime, TerminateCause, etc, but they don't update the StatusType... why?
Possibly, although if there is a acctstoptime then the StatusType is not really necessary. Maybe the StatusType should be set to "off" or something similar to be explicit about how the record was closed... (These type of behaviours are not defined by RFCs so we try to come up with the most logical/usefull setup)
Cheers
I agree with Peter on this one. I would set the Acct-Terminate-Cause to something like NAS-Reset. Setting the Acct-Status-Type to Stop would in fact be misrepresenting the facts. In practice we do no "Bill" customers for use if we do not get a valid stop record, but we also configure a maximum session time, because our NAS provider does not support interim updates. - List info/subscribe/unsubscribe? See http://www.freeradius.org/list/devel.html