<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=ISO-8859-1" http-equiv="Content-Type">
</head>
<body bgcolor="#ffffff" text="#000000">
Rune Kock wrote:
<blockquote
cite="mid:fa8654f10808221454w4bfc4181v4683a3c04698b5b6@mail.gmail.com"
type="cite">
<pre wrap="">On Fri, Aug 22, 2008 at 09:35, Paul Chambers <a class="moz-txt-link-rfc2396E" href="mailto:bod@bod.org"><bod@bod.org></a> wrote:
</pre>
<blockquote type="cite">
<pre wrap="">Sadly the Dell Powerconnect 2716 does not support SNMP at all. The so-called
'RMON' support is limited to aggregated statistics for the entire switch,
and is only visible within the web interface, not through SNMP (I think it's
very misleading on Dell's part to call that RMON, to be honest).
So no joy with this particular model, sorry. It's possible that there are
other mid-level switches which do support this info via SNMP, but I have no
information to offer.
</pre>
</blockquote>
<pre wrap=""><!---->Thanks for taking the time to test it. I'll start looking at
different switches and see if one turns up at a reasonable price.
Anyway, just knowing that this is an option is a great help. It never
occurred to me to look for advanced switches to solve this.
</pre>
</blockquote>
Yes, the trick is going to be finding something affordable with the
right feature set. I did a quick search on PriceGrabber for 'managed
switch' and was pleasantly surprised at some of the pricing - it's
dropped a lot in the year or so since I bought the 2716. Particularly
if you don't need gigabit ethernet on every port.<br>
<br>
The '802.1v VLAN classification by port and protocol' standard that
some managed switches offer caught my eye. Though a quick search didn't
turn up a clear explanation of what that meant. Maybe someone on the
list knows. If 'protocol' could equal 'DHCP', that could be a way to
make sure all DHCP traffic is switched to a central point.<br>
<blockquote
cite="mid:fa8654f10808221454w4bfc4181v4683a3c04698b5b6@mail.gmail.com"
type="cite">
<blockquote type="cite">
<pre wrap="">Another random idea: how about attempting to 'starve' a rogue DHCP server of
addresses to hand out? i.e. monitor for another DHCP server, and if one
appears, repeatedly request fresh DHCP addresses until it has no more to
hand out? would the requests need to come from unique MAC addresses? does it
help to pretend to be a bootp relay? This isn't an area I know a whole lot
about, to be honest.
</pre>
</blockquote>
<pre wrap=""><!---->
A very interesting idea. I don't know how a DHCP-server reacts when
it runs out of IPs. Whether it just becomes silent, or whether it
sends an error back. The first case would be great, the latter might
cause the client to give up, and thus not solve anything.
</pre>
</blockquote>
Yep, no way of knowing, and perhaps inconsistent amongst products. A
device may even recycle the 'oldest' lease. Not 'correct' behavior, but
I've seen plenty of incorrect behavior from networking gear.
And there's always a possibility that you'd tickle a bug in the device
in question.<br>
<blockquote
cite="mid:fa8654f10808221454w4bfc4181v4683a3c04698b5b6@mail.gmail.com"
type="cite">
<pre wrap="">If this idea works, it might be possible to hack an existing program
such as dhcp_probe to do this.
</pre>
</blockquote>
I expect so, it should be possible on a Linux or BSD box.<br>
<blockquote
cite="mid:fa8654f10808221454w4bfc4181v4683a3c04698b5b6@mail.gmail.com"
type="cite">
<blockquote type="cite">
<pre wrap="">It's a shame DHCP doesn't offer a mechanism to handle such situations more
gracefully. I guess we could always extend dnsmasq to add one, in an attempt
to establish a de facto standard :)
</pre>
</blockquote>
<pre wrap=""><!---->I wonder if IPv6 handles this any better.
</pre>
</blockquote>
My limited experiences with IPv6 have been that it makes things more
complex, not simpler... :)<br>
<blockquote
cite="mid:fa8654f10808221454w4bfc4181v4683a3c04698b5b6@mail.gmail.com"
type="cite">
<pre wrap="">If we were to extend the DHCP-standard, I would suggest a priority
field. Routers would have a low priority until they are explicitly
configured with a higher one. And the top priorities might require a
certificate signed by some CA.
</pre>
</blockquote>
Yep, that'd be one piece of the puzzle. I'm a little surprised that the
DHC working group hasn't considered the problem, to be honest. I've
seen the same problem occur on networks in my workplace more than once.
Sure-fire way to irritate the admins ;)<br>
<br>
Paul<br>
</body>
</html>