Charles Sprickman <sp...@bway.net> wrote
  in <alpine.osx.2.00.1105170300090.1...@hotlap.nat.fasttrackmonkey.com>:

sp> First, the easy one.  For IPv6 aliases, what is the proper subnet?

 Normally it is a /64.  See also Section 2.5.4 in RFC 4291.

sp> And the second one, which is also probably easy.  We're going to move
sp> at some point from a bunch of subnets on the same wire to having our
sp> own router that gets our blocks routed to it.  At that point I'd like
sp> to move to routing individual IPs (or small subnets) to each host
sp> behind the router.
sp>
sp> For example, say we have the following routed to our router:
sp>
sp> 10.1.0.0/27
sp> 10.2.0.0/27
sp> 10.3.0.0/27
sp>
sp> All the hosts behind our router are in 10.1.0.0/27.  I want to send
sp> some IPs from 10.2.0.0/27 and 10.3.0.0/27 to a host at 10.1.0.2, so I
sp> do the equivalent of "ip route 10.2.0.0 255.255.255.248 10.1.0.2"
sp> (cisco speak) on the router box.  How should the aliases on 10.1.0.2
sp> be defined?  Should they all have /32 masks?  Should the first get a
sp> /29 and the rest a /32?
sp>
sp> Is this even a valid config?  In reality, we have way more subnets,
sp> totally non-contiguous, varying masks.  With VRRP on the provider's
sp> side, we immediately lose 2 IPs from each subnet in our current setup,
sp> plus the network and broadcast IPs.  I'm hoping that in a routed setup
sp> I can regain not only the VRRP IPs but the top and bottom of each
sp> subnet... Considering the scarcity of IPs these days, that would be a
sp> big help.

 Well, I could not understand what you are trying... Is 10.1.0.2
 located on 10.1.0.0/27 and acting as another nexthop router?  If you
 want to split three subnets on a single wire into three subnets on
 three wires, simply configuring three /27 addresses to each interface
 on the router works.  If you want to route a part of the traffic from
 specific addresses to a specific host, you can add a specific route
 for the address range.

-- Hiroki

Attachment: pgpLFnQHNuek4.pgp
Description: PGP signature

Reply via email to