[comp.dcom.sys.cisco] RIP advertising secondary IP addresses

C.Chaundy@its.unimelb.EDU.AU (09/04/90)

We are just beginning to install cisco routers to raplace a level 2 bridged
network on our campus and want to use the secondary IP address feature to
ease the transition to a fully subnetted network.  We were intending to add
a secondary address in the subnet to which we wish to migrate a group of hosts
to the ethernet interface that is connected to the current bridged backbone,
change the IP numbers of the hosts, and finally delete the secondary address
and configure a different interface with this address (and move the physical
cable).  Does this sound like the right approach?

We are also connected to the wider world by a cisco unit connected to our
backbone and most systems simply point their default route at its address.
When we added the secondary IP address to the 'local' cisco, our 'gateway'
cisco did not learn about the new subnet and I had to configure an explicit
ip route into the unit (both units are configured for RIP).  Is this to be
expected, i.e., the secondary network number is not advertised by RIP?

Regards, Chris Chaundy

Technical Manager, Networks,
Information Technology Services, The University of Melbourne
Internet: C.Chaundy@its.unimelb.EDU.AU	(DTE 505233430003)
Phone:    +61 3 344 7045  Cables Unimelb
Fax:      +61 3 347 4803  Telex AA35185
Post:     Parkville, Victoria 3052 Australia

swatt@noc.net.yale.edu (Alan S. Watt) (09/06/90)

Chris Chandry of The University of Melbourne writes:
----------------------------------------------------------------------
We are just beginning to install cisco routers to raplace a level 2 bridged
network on our campus and want to use the secondary IP address feature to
ease the transition to a fully subnetted network.  We were intending to add
a secondary address in the subnet to which we wish to migrate a group of hosts
to the ethernet interface that is connected to the current bridged backbone,
change the IP numbers of the hosts, and finally delete the secondary address
and configure a different interface with this address (and move the physical
cable).  Does this sound like the right approach?

We are also connected to the wider world by a cisco unit connected to our
backbone and most systems simply point their default route at its address.
When we added the secondary IP address to the 'local' cisco, our 'gateway'
cisco did not learn about the new subnet and I had to configure an explicit
ip route into the unit (both units are configured for RIP).  Is this to be
expected, i.e., the secondary network number is not advertised by RIP?

----------------------------------------------------------------------

In most respects, using secondary addresses is equivalent to having
multiple interfaces on the same cable.  You have stumbled upon one
of the exceptions.

Assume you have a simple configuration:


	===============+============================	(cable A)
                       |                                        
                       | (interface A: 128.128.1.1 255.255.255.0)            
		+------+--------+
                |               |
                |  cisco box    |
                |               |
		+------+--------+
                       | (interface B: 128.128.2.1 255.255.255.0)              
                       |                                        
	===============+============================	(cable B)

Now assume you have the class B network 128.128.0.0 assigned, and
subnetted on an 8-bit boundary.  Also assume interface A has address
128.128.1.1 and interface B has address 128.128.2.1.

In this configuration, the router would advertise subnet 128.128.2.0,
plus any other networks learned about through interface A onto
interface B.  It would also advertise subnet 128.128.1.0, plus any
other networks learned about through interface B onto interface A.  It
would *NOT* advertise 128.128.1 out via interface A nor 128.128.2 out
via interface B.

If you add secondary addresses (say 128.128.3.0 255.255.255.0 on
interface A), the same rules apply. 128.128.3 will be advertised out
via interface B, but not out via interface A.

This behavior is reasonable, just "not what I want" in some
circumstances.  A further caveat is the router will never generate a
datagram with a source of a secondary interface address.  You can
"ping" a secondary address and the return datagram comes back correctly
sourced; you can "telnet" to a secondary address and everything will
work.  But no SNMP traps, log messages, or routing updates will go out
sourced as a secondary address.  This can cause "routed" on some
systems (Suns) to complain about "packet from unknown router" when they
receive a routing update from an address they do not believe is on a
directly attached network.

What will work is:

  1)	Assign subnet or network numbers to cables just as you would
	if you had routers instead of bridges.

  2)	Assign secondary addresses to appropriate interface(s) of
	the router.  My convention is the gateway between the core
	network and a subnet is always node "1" in each subnet.
	I call the subnets created by these secondary addresses
	"virtual subnets".

  3)	Configure hosts on the virtual subnets to *not* run a
	routing protocol, but instead have a static default route
	to appropriate secondary interface address.

  4)	Configure any additional cisco routers on this cable to
	be in on the secret, either by further secondary addresses
	or by interface routes.

	I don't recommend trying to make UNIX-based routers smart
	about such a scheme, as they will attempt to forward
	directed broadcasts for the secondary subnets (they don't
	know it's a broadcast and haven't saved the MAC header to
	figure it out the way cisco does).  Configure them instead
	to have a static default route to a smart cisco gateway.

Ordinary host configurations are thus kept very simple; you need only
specify IP address, subnet mask, and default gateway.  Hosts on two
different virtual subnets on the same cable will route packets to
each other via the cisco box.  This means the same packet will appear
twice on the same cable, so this is not a good solution if traffic
on a particular cable is very heavy.

I use secondary addresses heavily; at one point there were 8
networks running on a sigle cable.

	- Alan S. Watt
	  High Speed Networking, Yale University
	  Computing and Information Systems
	  Box 2112 Yale Station
	  New Haven, CT  06520-2112
	  (203) 432-6600 X394
	  Watt-Alan@Yale.Edu


Disclaimer:  "Make Love, Not War -- Be Prepared For Both"
		- Edelman's Sporting Goods [and Marital Aids?]