can someone help me out with the following scenario

I am new to OSPF, so bare with me.

My eth1 faces the internet with a public IP (several) my customers are natted to their cpe wlan, and then again natted across to various public IP's in the gateway according to their service type.

Eth2 has an address in the range 12.1.0.0/24 and Eth3 has an address in the range 13.1.0.0/24

Routers are seperated by IP ranges at present, both clockwise and anticlockwise as far as point Z... Eg 12.1.1.x - 12.1.10.x clockwise. And antiiclockwise 13.1.1.x - 13.1.10.x

everything is routed back to the gateway, however at point Z I have to choose between one direction or the other

Having closed the remaining segmant in my loop.. I want to try more advanceed things with OSPF.

Starting with redundancy and maybe splitting the load across the 2 halves.

I am about to change the entire structure to one address range in blocks of /30.... All same range, say 10.1.x.x

Forgetting the stubs and the more mundain stuff,

OSPF must be set in R1 to redistribute default, correct!!

And in both R2 and R3, they must be set to Redistribute connected only (as well as all others)

And the entire loop is same network 10.1.0.0/24 to encompass all /30 blocks.

The loop is also called the same name "backbone" . Each RB will have a "loopback" with a /32 address added into the ospf routes.

My question is right back at R1, I am confused I cannot have 2 Ip's in the same range across seperate interfaces, I also do not want to use a bridge. I can of course segment both Eth2 and Eth3 by a /30 block.

So when install OSPF on the gateway, do I install OSPF to run on ALL interfaces (eth2 and 3) or do I have 2 seperate instances of OSPF running.. Open one tab and run it on Eth2 and open a second tab and run another service on eth3

Maybe I am just thinking too hard.

Anyone !

I recommend using a setup to separate interfaces, make yo able to prioritize over your bakup links. an easy setup would be default. but have a 100 prio to the backup p2p ospf connection.

hi samsung, I had almost given up on the forum.. I was about to experiment with a standard set up.

1 instance running on both interfaces, adding in 1 route to cover all the /30 addresses, also 1 route for all the loopbacks and routes for each 192.169.xx Ap range (clients)

redistribute default (only on gateway)
redistribute connected type 1 on all others.

What did you mean by prioritise...

As this is my first experiment with OSPF, do you think it would be wise for me to keep to something simple to start with, or how complicated is your plan ?... And can you explain more the benefits.

Simon

I don't think you need to redistribute connected unless you are not advertising on the connected network. So if your "Connected" network is listed in the "Networks" tab of OSPF, then AFAIK you shouldn't need to redistribute those connected routes.

By default OSPF will calculate all paths via the Dijkstra algorithm, which is similar to using router hop counts. So if you want it to take another course, you will need to add a Cost under "Interface" tab for that interface. So if you have a link between two router that you want to force as a failover link, then simply add the Interface with a higher cost, such as 50 (if it is high enough). Make sure you do the same for both sides of the link (in two routers) or you will see transmit and receive traffic going different directions. OSPF is a great way to handle redundancy and networks can be engineered to go through certain links via costs. But by itself, there is not an easy way to do load balancing with two attached links from the same subnet.