Does the issue go away when you make ping part of the high priority queue?
On Mon, Jul 21, 2008 at 7:55 AM, Amaury De Ganseman
<[EMAIL PROTECTED]> wrote:
> Hi all,
>
> I run OpenBSD 4.3 on my gateway. But when a machine behind the
> NAT/gateway uses bittoreent (or gtk-gnutella) I loss packets.
> For example when I try to do a ping www.google.com I can see "ping:
> sendto: No buffer space available" (on my gateway)
> It's the same if I use gtk-gnutella. I think it's related to the huge
> number of states (about 1500 for bittorent)
>
> Any idea to the problem ?
>
> Thanks for help.
>
>
> Amaury De Ganseman
> infos:
> ---------
>
> dmesg in attach
>
> I use user-space pppoe
>
> I use traffic shaping:
>
> #ALTQ CBQ
>
> altq on $ext_if cbq bandwidth 350Kb queue {std}
>
> queue std bandwidth 100% { hprio, lprio}
> queue lprio bandwidth 80% priority 0 cbq(default, borrow,red, ecn )
> queue hprio bandwidth 20% priority 7 cbq(borrow, red, ecn)
>
>
>
> gateway# netstat -m
> 37 mbufs in use:
> 32 mbufs allocated to data
> 1 mbuf allocated to packet headers
> 4 mbufs allocated to socket names and addresses
> 32/306/6144 mbuf clusters in use (current/peak/max)
> 708 Kbytes allocated to network (10% in use)
> 0 requests for memory denied
> 0 requests for memory delayed
> 0 calls to protocol drain routines
>
> I use user-space pppoe
