How is FreeRADIUS supposed to know when a user disconnects and frees up the IP address from the pool if the NAS doesn’t tell it? Anything else is not exactly reliable. If you have a user with a long duration session that lasts longer than your timeout the IP could be put back into the pool when it is still in use.

The best solution would be to fix the NAS to send the packets or fix the network to make sure they get delivered.

Michael

----------------------------------------------------------------------

Michael J. Hartwick, VE3SLQ                      hartwick@hartwick.com

Hartwick Communications Consulting                      (519) 396-7719

Kincardine, ON, CA                             http://www.hartwick.com

----------------------------------------------------------------------

From: freeradius-users-bounces+hartwick=hartwick.com@lists.freeradius.org [mailto:freeradius-users-bounces+hartwick=hartwick.com@lists.freeradius.org] On Behalf Of Tabacchiera Stefano
Sent: Tuesday, May 04, 2010 15:39
To: freeradius-users@lists.freeradius.org
Subject: Re: R: Re: R: Re: R: rlm_ippool: No available ip addresses in pool

>Tabacchiera Stefano wrote:
>> Alan,
>> here's the content of gdbm db:
>
>  Ah... it's the DBM pools.

I already stated that in the subject of my mail (did you notice the module name?).

>
>  Well.. use "rlm_ippool_tool" to manage the pool.
>

Great idea! Too bad that tool allow only to cleal *all* the entries in the DBM pool.

Or am I just missing something?

>  Or, get your NAS to send accounting stop packets.  It's *supposed* to
>send stop packets when a user session is cleared.

As I already said, I know the NAS sometimes doesn't send acct-stop pkts, but it's out of my control.

My questions (still unanswered, let me say) are:

1) is maximum_timeout useless?

2) Is there a way to keep my dbm pool safe and updated (I mean no expired addresses), even in the case some acct-stop pkt are loss?

3) Should I switch to sql pool, 'cause dbm it's actually unreliable?