Problem with rml_sqlcounter with GigaByte datavolume

[Hanno Schupp]

Hi Yves,
thanks for your response. I understand the difference between the storing
gigawords in the database and the sqlcounter response. I do not speak C, but
have no doubt that what you are saying is correct. I just find it hard to
believe that FreeRadius has such a basic limitation, which it appears can be
easily overcome (if you can do so with a PERL script using rlm_perl, it must
be possible to do it with C in rlm_sqlcounter, right), but so far no one has
bothered to address. While this might have been OK in former times, when
data traffic was expensive and limited, in a broadband scenario multi
GigaByte data allowances over 4GB per month are very common, and I am
gobsmacked that FreeRadius cannot handle what seems a very basic
requirement.

So accepting the point that this is a limitation of FreeRadius in its
current version, I reword my question to Alan:
Are there any plans to address this shortcoming of FreeRadius in the near
future?

Failing that, are there any rlm_perl scripts out there on the wiki or in the
wider user community, that can handle gigawords on the radcheck values that
actual usage is checked against? My language is PHP and I know from
experience that while PHP code can be embedded into FreeRadius, it is
probably the least performing option. So I would love to avoid to roll my
own in PHP, if at all possible.



YvesDM wrote:

You confuse gigawords storage in the database coming from acct updates/stop
packets of the nas with the reply from sqlcounter.
FR is capable of saving gigawords in the database when a nas is sending
them, that's not the problem.
But, the sqlcounter's code was never changed to reply gigawords to the nas.
Check the C code and you will see.

Kind regards
Y.


On Mon, Jun 6, 2011 at 1:24 PM, Hanno Schupp <hanno.schupp at gmail.com> wrote:

> Thank you for this reply.
>
> I thought the limitation might come from the wrapping around 4.3 GB due to
> the limitations of a 32bit system with 2147483648 being the highest signed
> and 4294967296 being the highest unsigned number. 1705032704 is then exactly
> the difference to 6GB, after the system wrapped at 4.29GB. I requite the
> log:
>
>
>
> Sat Jun  4 23:10:21 2011 : Debug: rlm_sqlcounter: Rejected user lapzel14,
> check_item=1705032704, counter=2147513300
>
>
>
> Exactly the 1705032704 one would expect based on highest 32bit unsigned
> integer.
>
>
>
> Now here is my problem: Why does it wrap at 32Bit, if the system is a x64
> server? Does not make a lot of sense to me.
>
>
>
> Also, the FAQ is containing instructions how to deal with gigawords in
> terms of the sql statements that handle the calculation of the counter
> value. And as this is implemented, the counter value is not the problem here
> – it is the check_item value that as I understand is based on my
> configuration, taken straight out of the radcheck table.
>
>
>
> I am sorry, but this sounds like a limitation/bug of the standard system,
> that could be overcome. After all, if it can be resolved with custom perl
> code as I understand you suggest, why should the standard system not be able
> to handle data limits larger than 4.29GB out of the box?
>
> Or am I missing something?
>
>
>
> Alan, can you enlighten us on this issue?
>
>
>
> Regards
>
>
>
> Hanno
>
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freeradius.org/pipermail/freeradius-users/attachments/20110608/a83ab10a/attachment.html>