Re: Radius Bug

Dale E. Reed Jr. (daler@comtch.iea.com)
Fri, 17 May 1996 10:29:20 -0700 (PDT)

On Fri, 17 May 1996, Steve Van Natta wrote:

> Wed Jan 3 07:53:12 1996
> Acct-Status-Type = Start
>
> Wed Jan 3 15:01:01 1996
> Acct-Status-Type = Stop
> Acct-Session-Time = 25673
>
> You list the total seconds as 25673. However when you subtract the
> entry in the stop log, from the entry in the start log, you have a
> total of 25428 seconds or 7:03:48. (HH:MM:SS)

But you didn't show Acct-Delay-Time in the list. Remember the DATE
stamp is from the Radius Server. The TIMES are from the NAS. This
is the difference. If the accounting packet (either the start or
stop) sits in the NAS for any length of time (which is sounds like
the start packet did) then you must modify the date at which you
display to the user by the Acct-Delay-Time entry.

Standard Radius STAMPS the packet in the detail file, and leaves
this as an exercise to the user. We use a simple SQL function in
our ODBC version to obtain correct times from the database and
haven't seen this happen.

> This is a pretty substantial difference (slightly over 4 minutes) and in the
> course of logins over the period of a month, this can really add up. The
> biggest embarrasment though would be to release statistics of start and stop
> times in an itemized listing, like the one shown above, that don't match.
>
> I believe this is an error, since it doesn't always happen. Sometimes the
> stuff adds up.

Again, check Acct-Delay-Time on the ones that do and don't add up.

> If for some reason this isn't an error, which is correct? Your start and
> stop time stamps? Or your total seconds under Acct-Session-Time? We
> endeavor to be accurate.

I would always use the Acct-Session-Time field.

Dale