[Rack] Rack Power Outage?

[Dr. Jesus]

On Mon, Sep 13, 2010 at 4:18 PM, Jonathan Lassoff <jof at thejof.com> wrote:
> On Mon, Sep 13, 2010 at 4:13 PM, Dr. Jesus <j at hug.gs> wrote:
>> On Mon, Sep 13, 2010 at 3:58 PM, Jonathan Lassoff <jof at thejof.com> wrote:
>>> Awesome. Seems good now.
>>>
>>> Do you think the altq parameters on r00ter are still the right rates?
>>> The network performance seems to have dropped precipitously in the
>>> last month or two.
>>
>> Performance as in speed, or performance as in BER?
>
> Performance as in ICMP pings taking 4+ seconds at time, SSH going
> really high latency, and TCP performance is in the crapper (50 kbps or
> less). It seems intermittent, and probably at high-use times.
>
> I haven't dug into it while it's been happening though.
>
> Come to think of it, I don't think ICMP is prioritized. I wonder if we
> should also match traffic with ToS/DSCP bits set.

I think what you really want is to prioritize small packets.  I
believe SSH and ICMP are already prioritized:

pass  out on $ext_if inet proto icmp from ($ext_if) to any keep state
queue ssh_im_out
pass  out on $ext_if inet proto tcp from ($ext_if) to any port $ssh_ports \
        flags S/SA keep state queue(std_out, ssh_im_out)

Hmm, it looks like pf doesn't provide a mechanism for matching packet
size without involving a divert socket and looking at the packet in
userspace.  Maybe I'm wrong, but I don't see anything in the pf syntax
definition about packet size.