Fw: question about realms file format

classic Classic list List threaded Threaded
2 messages Options
| Threaded
Open this post in threaded view
|

Fw: question about realms file format

Thanh Lim
Oops.  Sorry, sent to the wrong list.

-thanh

----- Original Message -----
From: "Thanh Lim" <[hidden email]>
To: <[hidden email]>
Sent: Tuesday, August 30, 2005 2:46 PM
Subject: question about realms file format


> How much longer will we need to support the old realms file format?  I'm
> looking into updating all code dealing with the REALM structure, and
> improving performance for lookups, etc. and am unsure if I should be
> updating that code as well.
>
> Also, I notice within the REALM structure, every single time, a linked
> list iteration occurs, many times, the active and account active flag is
> checked, and set, even if it's not the ip address and port that we are
> looking for. Why was it done this way, and shouldn't we just update the
> active flag only when we find the right one(s)?
>
> Thanks,
> -thanh
>

-
List info/subscribe/unsubscribe? See http://www.freeradius.org/list/devel.html
| Threaded
Open this post in threaded view
|

Re: Fw: question about realms file format

Alan DeKok
"Thanh Lim" <[hidden email]> wrote:
> How much longer will we need to support the old realms file format?

  I'd like to fix it in the CVS head, which will hopefully become 2.0.

>  I'm looking into updating all code dealing with the REALM
> structure, and improving performance for lookups, etc. and am unsure
> if I should be updating that code as well.

  Sure.

> Also, I notice within the REALM structure, every single time, a linked
> list iteration occurs, many times, the active and account active flag is
> checked, and set, even if it's not the ip address and port that we are
> looking for. Why was it done this way, and shouldn't we just update the
> active flag only when we find the right one(s)?

  It's for things like load-balancing & fail-over.  If you're
load-balancing requests to "N" home servers, then you need to wake the
dead ones up, even though you're not sending packets to them.  If you
don't wake them up, you'll *never* send packets to them.

  A better solution would be to have something run through the lists
every second or so, and wake up the dead servers.  That would have the
same effect, with less performance impact.

  Alan DeKok.

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