Date: Fri, 24 Jan 1997 15:34:11 +0200 (EET) From: Andrew Stesin <stesin@gu.net> To: michael butler <imb@scgt.oz.au> Cc: jdp@polstra.com, hackers@freebsd.org Subject: Re: Fault-tolerant network with 2 ethernets Message-ID: <Pine.BSI.3.95.970124151806.5513N-100000@creator.gu.kiev.ua> In-Reply-To: <199701241245.XAA08098@asstdc.scgt.oz.au>
next in thread | previous in thread | raw e-mail | index | archive | help
On Fri, 24 Jan 1997, michael butler wrote: > > > Yes. But make sure that a set of IP addresses > > used for loopback-aliased router-IDs is _not_ covered > > by any of the subnets (address ranges). > > You need to exercise some degree of caution with this. The temptation is > to use a subnet with a netmask so small as to only accomodate that of the > loopback interface (255.255.255.252 is commonly used on Ciscos). 255.255.255.255 will do (at least does for me both on cisco's and on Gated's loopbacks). This means that you get your loopback as a "stub host" in OSPF terms, not a "stub net" -- why loose addresses due to unneccesary "wide" netmask? I just take a /27 subnet (32 addresses) and assign them one-by-one to routers as their IDs, with 0xffffffff netmask, and with no regard to wherever the router topologically or physically belongs to (keeping only administrative issues in mind). And this just works. > This is all well and good when everything else listening to your IGP > understands what it means. If you have "legacy" hardware lying about (some > terminal servers and Cisco 1003s come to mind), you need to be very careful > about what you propagate into other, classful IGPs (such as RIPv1 or IGRP) > so that they too can understand how to forward packets properly. I think that that wouldn't be a cool idea -- to use legacy crap in such a setup; especially with 100Mbit ether :) (Show me that old dumb device, which a) doesn't understand OSPF and b) has 100Mbit ether interface! :) And I think that a great majority of today's terminal/access servers and other "hardware" routers understand and implement OSPFv2, just any device smarter than an oak table should do. ;) > I suggest trying to arrange things such that appropriate route summarisation > can be achieved by judicious rules in gated.conf on suitable (and > comprehensible) subnet boundaries. In my case, it was easy - such hardware > only gets a default route and nothing else :-) Yes. Me thinks - this is the very best approach if you should deal with dumb, RIP-1-only devices: let "smart" devices (cisco 25xx and better, FreeBSD+gated) play all the routing game, and let them just decide where others should go; "smarties" redistribute only an appropriate default into RIP, redirecting all others to wherever is appropriate at the moment. But I'd better avoid that "way too dumb" devices at all. > > michael > -- Best, Andrew Stesin nic-hdl: ST73-RIPE
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSI.3.95.970124151806.5513N-100000>