On Mon, 19 Jul 2004, jesk wrote:

> > > What's the time difference between sending the accounting-request and
> > receiving
> > > the accounting-response.
> >
> > time radclient -f acctfile xx.xxx.xxx.xxx acct secret
> > Received response ID 34, code 5, length = 20
> >
> > real    0m0.080s
> > user    0m0.057s
> > sys     0m0.009s
> >
> > > The process is serialized yes, but more slots will allow you to have
> more
> > > outstanding requests. This can help only if radrelay is using up it's
> > slots (in
> > > other words when the accepting radius server is slow).
> >
> > i increased the slot variable to 255 and you are right radrelay
> performance
> > better now. it sends more serialized requests out before it will go
> through
> > the while(1)loop, that saves time. radrelay is uptodate with its sending
> > accounting
> > data, and the file is rotated very fast.
> >
> > thank you for your help.
> >
> > best regards,
> > christian
> >
> >
> >
> > -
> > List info/subscribe/unsubscribe? See
> http://www.freeradius.org/list/users.html
> >
> >
>
> is there the possibility to increase NR_SLOTS above 256. the file is growing

No 256 is the maximum for obvious reasons (we can't have more than 256 radius
packets pending) unless we use more fd's.

> again under very high load
> but after 20 minutes it gets rotated.

Again i believe that the accepting radius server does not handle accounting
requests as quickly as they are written in the detail file (and thus processed
by radrelay). If you have 256 outstanding requests then your radius is slow, not
the application sending those requests. Even worse if this application
serializes requests.

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

--
Kostas Kalevras         Network Operations Center
[EMAIL PROTECTED]       National Technical University of Athens, Greece
Work Phone:             +30 210 7721861
'Go back to the shadow' Gandalf

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

Reply via email to