> > If I have the number of SQL connections say set to 10, then when I
> > launch a large number of RADIUS requests the log files say that
> > there isn't enough SQL connections and it seems to then reject the
> > connection.  So I get a few accepts and hundreds of rejects.
> 
>   Yes... because the SQL queries are taking too long.  It's simple
> math.
> 
>   The server gets 100 requests in a second, and if the SQL queries
> take 1/10 of a second each, it will take 10 seconds to respond to all
> of the requests.  At which point, some will have timed out, and the
> NAS will have sent even more.
> 
>   Fix your SQL database to be faster, or run it on a faster machine.

I have tried running it from a RAM disk with no noticable improvement.  The machine it 
is running on is a 2GHz system, however the same system is the one generating the 
requests also.  I will try moving the radius off onto a non local system and see how 
it behaves.

> > How is this achieved? I thought that the RADIUS protocol only
> > permitted 256 sessions, (well per connection)?
> 
>   There is no "connection" between NAS and RADIUS server.
> 
>   Read the RFC's for descriptions of how more than 256 requests from
> one NAS are allowed.

I have been reading through the RFC's I shall dig deeper.

> > if I have less than 256 it rejects a lot of the requests.
> 
>   Because your SQL database is slow.  Adding more SQL sockets means
> that even more requests can *appear* to be handled.  It doesn't solve
> the underlying problem.

Would switching to non-threaded increase performance for fewer connections?

> > Any idea of timescale for this?
> 
>   Before 1.0.

:)

> >  hanging back on responces to slow the sender down?
> 
>   That won't make any difference.

Thats what I thought

-- 
-----
Graeme Hinchliffe (BSc)
Core Internet Systems Designer
Zen Internet (http://www.zen.co.uk)

ICQ 3842605 (link)

Sales : 0870 6000 971
Fax   : 0870 6000 972


- 
List info/subscribe/unsubscribe? See http://www.freeradius.org/list/users.html

Reply via email to