First, reading through Tom Limoncelli’s article, I note several places where 
Tom says, “It’s probably not worth the effort to do this for a couple of 
servers,” or words to that effect. Tom’s article is promoting a mechanism 
provided by Lucent network gear while he worked at Lucent and had an 
opportunity to experiment with it. I’m probably one of the few fans of this 
idea you’ll find on this list, and I’ve never found a real justification for 
trying it.

As I recall, Juniper SRX devices are able to work in “transparent mode” (i.e. 
as a bridge) and I’d be surprised if security devices from Cisco, etc., could 
not do this. As dsr noted, you can also do it with open software. Probably a 
good idea to dedicate a couple of Linux boxes to this job in case you need 
updates and patches to maintain ebtables that you don’t want on other hosts.

While you can do many things with NATted addresses, especially since many 
protocols have been extended to allow for the nasty kludge of Firewallism. 
Depending on your reasons for maintaining your existing IP addresses on those 
servers as they are, you could end up in a world of pain.

Unless you absolutely cannot stand to ever have an Ethernet frame from a bad 
host touch be seen by the kernels of the special servers, maybe run iptables on 
the special servers to restrict traffic to the special servers? If they are 
Well Known as special, this should ease the ongoing maintenance load.

If you want a way to do this that’s REALLY crazy and unmaintainable, try using 
locked down ARP tables on your servers, with the IP to Ethernet mappings 
configured by Chef.

Bob

> On Jul 18, 2016, at 10:48 AM, Daniel Feenberg <[email protected]> wrote:
> 
> 
> 
> On Mon, 18 Jul 2016, John Stoffel wrote:
> 
>>>>>>> "Edward" == Edward Ned Harvey (bblisa4) <[email protected]> writes:
>> 
>>>> From: bblisa [mailto:[email protected]] On Behalf Of Daniel
>>>> Feenberg
>>>> 
>>>> We'd like to isolate a few machines from the rest of our LAN without
>>>> renumbering them into a subnet.
>> 
>> Edward> I don't envy the IT person or newhire who inherits this
>> Edward> environment someday. I'm sorry my comment isn't constructively
>> Edward> adding to the direction you want to go - you're probably very
>> Edward> smart and have thought this through, and considered all the
>> Edward> pros and cons, and have good management (or you are yourself,
>> Edward> management)... And I'm sorry that this email will probably
>> Edward> spark a debate about whether you should or should-not, and all
>> Edward> the reasons why, which will distract from the answer that you
>> Edward> actually want. That being said, it is almost never a good
>> Edward> management decision to do "tricks" and configure systems in
>> Edward> weird, uncommon, nonstandard ways that will be surprising or
>> Edward> confusing to new future people, or just a later version of
>> Edward> yourself, who forgot you previously did something weird. If I
>> Edward> were manager there, it would require a *very* compelling
>> Edward> reason to convince me this should be done.
>> 
>> Hear hear!  If you have machines you don't trust, why can't you
>> re-number them?
> 
> We have been asked to isolate a small subset of machines. Renumbering 
> everything else to isolate a few seemed infelicitous.
> 
>> Or even put them behind a NAT/Firewall that exposes
>> the original IPs for these hosts, but locks things down that way?
> 
> That is what we would like to do. As I understand it using an ordinary bridge 
> the original IPs to be exposed would have to be in a subnet, which they are 
> not. Nor do we have the IP space available to make a new subnet for them. 
> Hence the interest in a transparent bridge. But if we can use NAT for this 
> purpose, we are interested.
> 
> daniel feenberg
> NBER
> 
> 
>> 
>> _______________________________________________
>> bblisa mailing list
>> [email protected]
>> http://www.bblisa.org/mailman/listinfo/bblisa
>> 
> 
> _______________________________________________
> bblisa mailing list
> [email protected]
> http://www.bblisa.org/mailman/listinfo/bblisa

_______________________________________________
bblisa mailing list
[email protected]
http://www.bblisa.org/mailman/listinfo/bblisa

Reply via email to