OSPF star network

RouterOS Forwarding Protocols
1

Assume we have three routers.

1.) Each router has two interfaces, one inward facing and one out ward facing.

2.) All the inward facing interfaces connect to a single switch forming a basic star nework.

3.) All machines are running one instance of ospf.

4.) There are multiple different and unique subnets on each of the other outward facing interfaces connected to many different things outside of the star.

It is clear, that if any machine in the star network connected to the main central switch, wants to talk to another machine in the same star network, all it has to do is issue an arp request for the remote IP, and send its packets to the mac address that answers send it here!

However if any machine in the star network wants to talk to any machine in the outward network of machines, there needs
to be another solution, hopefully not a whole lot of manual static routes :slight_smile:

OSPF on each machine should be able to redistribute all connected subnets on all the outward facing
interfaces so that all machines know how to talk to any subnet on any machine.

What is the proper configuraton for OSPF on these these various machines?

Each interface needs a network statement, say 0.0.0.0/0 area 0.0.0.0

That would put all interfaces and subnets of all machines in area 0.0.0.0, and in the routing table a route should be listed along with an LSA
pointing any machine to any subnet on any other machine.

Is this right? Is there a better way to do it?

As a second question, how does one prevent two connected machines that have IP’s on multiple shared subnets on the
interfaces connecting them, from sending multicast LSA’s from each and every subnet to the same subnet
at the other machine.

Firewalling can stop the LSA’s from being broadcast, but that doesn’t stop ospf from trying. Waste of CPU.

Thanks,

Homer W Smith
CEO Lightlink Internet

2

OSPF should handle what you described. The exception would be the best practice of using the most specific network statements possible and to make all interfaces passive in the routing process by default and then making passive=no on the interface you want OSPF to communicate on.

The multiple IPs on shared subnets have me worried.

What do you mean by this? Are you sending VLANs from the inward interface to a switch? Is it just a lot of IP addresses assigned to the same interface?

Edit note: I must have been sleeping. In this context it should have read passive=no not passive=yes.

3

Homer,

A couple of pointers for you.

  1. Don’t redistribute into OSPF unless you absolutely have no choice. It creates external routes and breaks the area boundaries of OSPF. Use network statements instead.

  2. Areas are what contain the SPF calculations and provide boundaries for LSAs. Use them if separation is needed between core routing and edge routing. They also provide the ability to summarize if needed.

4

Thanks you for your attention.

No vlans, just simply multiple subnets, 10.16.23.0/24 and 10.16.35.0/24 on a single interface etc, plus a basic /28 perhaps connecting all the machines in the
star network together through the main switch, and a basic /30 connecting leaf sites to each closer router. Each and every interface has the potential of feeding customers off of them directly.

This network was built before we had mikrotics, we used linux boxes that had two interfaces, one in and one out, and so interfaces
tended to get crowed with different subnets meant for different clusters of customers.

Reason why we did this is something like when we expand into a new community, we often just leave them bridged to the closest router they are connected to, even though in theory they might need a router of their own one day to contain their broadcasts. Once a community gets big enough, we
stick in a new router in the community and move that community’s subnet from the router it was on to the new router, that way we don’t need to renumber
anyone. Since we have a lot of small communities connected to our network (by ubi wireless PTPs) we have lots of potential places where
routers might be one day, but for 1 to 5 customers we do not bother. But each little nest of customers has its own subnet on the router they
do connect to, in case we want to stick in a new router for that community alone.

Homer W. Smith
CEO Lightlink

5

I am sorry this leaves me confused, I am new to OSPF. Did you mean passive=no above there on the active interface?

By passive interface I take it you mean interface that listens for OSPF but does not broadcast OSPF?

Let’s start again, I lied, we do not have a star network, we have a ring network of 4 core routers, like a base ball
diamond with home plate connected to our upstream and then also connected to third and first base which are connected
to second.

Each of the other three ring routers form small star networks of 1 or more extensions going to secondary routers that
feed local communities again by ubi wireless PTMP.

Each of the 4 ring routers connect to each other using ubi wireless PTP links, and broadcast and receive OSPF between them,
and also all the remote secondary routers that connect to them.

The 4 ring routers are connected to each other by unique /30 subnets, PLUS a random number of /24 private subnets on the same interfaces
that connect the routers to each other, and intefaces
that are used to connect to customer’s home routers. Sometimes the /24 subnets have IPs on both ring routers that are connected
to each other by their individual /30, sometimes not. It’s a bit of a circus.

Originally each interface connecting the ring routers to each other had a “network x.x.x.x/30 area 0.0.0.0” declaring only its own /30 for OSPF
to work with, but I also had redistribute connected and static on. Thus every router knew where every subnet on every router was but would mark the statics and connecteds as ext routes. But it all worked just fine.

The PROBLEM I have is that there is a natural direction of packet flow from home base out to the other ring routers to get to a leaf community
that sometimes needs to be changed on a SUBNET by SUBNET basis, not on an interface by inteface basis.

The reason is that the link from home base to first base gets FULL at night and the link from home to third is empty.

So I want to reroute just one subnet, 10.16.31.0/24 so it goes from home to third to second to first to end community,
rather than directly from home to first to end community.

I did this by putting in a static route on home to direct 10.16.31.0/24 to third. This confused things because when the
packet got to second base it saw two routes to the community of equal cost, one through first base which was correct, and one
back to home which was not.

I couldn’t fix this with my limited understanding, so I changed all the network statements to 0.0.0.0/0 and now all routes
are intra area, and the problem at second base stopped happening even with the static at home pointing 10.16.30.0/24 to third base.

Basically I am too confused to know what I am doing.

So what I am trying to do is learn what the optimum strategy is for star based networks, ring based networks, linear based networks, and mixtures of the first three which we have.

Homer W. Smith
CEO Lightlink

6

I am not sure what ‘redistribute INTO OSPF’ means in this context.

Say I have two routers connected by a single link, both running an instance of OSPF with a /30 between them.

So there is a network x.x.x.x/30 area 0.0.0.0 on each router.

Now each router has one other interface to which are connected work stations. These interfaces are 10.16.0.0./24 and 10.17.0.0/24

Normally I would “redistribute connected” on each router and OSPF would find 10.16 and 10.17 as externals.

You are suggesting I trash the redistribute connected and add in a “network 10.16.0.0/24 area 0.0.0.0” on the first router and
“network 10.17.0.0./24 area 0.0.0.0” on the second router instead.

Is that correct?

How is this different than simply using network 0.0.0.0/0 area 0.0.0.0 on all routers so I don’t need a network statement
for every subnet placed on an interface?

Thanks you for your time.

Homer