How to detect availability

In article , Igor Mamuzic wrote: :I have the following topology:

:- R1 ---- R2 ----- R3-----Internet

:How to make that R1 knows that the link between R2 and R3 is down?

How to make R1 know that the link between R1 and R2 is down?

First you have to define "down". Do you mean only that the link has been cut (e.g., unplugged, a cable broke, a repeater burned out, a laser transceiver burned out), or do you mean that the link is not passing *usable* data (e.g., DoS attack is using the entire bandwidth), or do you mean that the link isn't passing

*any* data (e.g., equipment is wedged but carrier is there), or do you include the case where the link quality has dropped to the point where the throughput is below some quality measure (e.g., a fibre connection got dirty or a cable is frayed or there is a ground fault or a duplex mismatch) ?

Do you really care whether it is the link R2 - R3 that is down, or are you really concerned with whether R1 can get data to R3 ? It makes a difference if, for example, there is an unshown redundant connection between R1 and R3, or if R1 is just a monitoring device and all the important data is joining the topology at R2. If the concern is getting data between R1 and R3, then ping R3...

For more advanced methods with various tradeoffs, see Vincent C. Jones' book about High Availability Networking Using Cisco Routers.

One way would be to use routing protocol and create a loopback on R1 and announce it in mentioned routing protocol. If route to that "network" is not available at R1, that should mean that R1 has no connectivity to R3.

If you have a route-map directing traffic from certain source addresses on the incoming interface of R1 and the route-map routes this traffic to R2 then you will be reliant on R2's routing table to re route this traffic. As R2-R3 is down then (guessing your setup) that the alternate route would be back via R1 (dynamic routing) then the traffic would route back via the invisibal link you talk about.

I am guessing you want a method of R1 knowing about the link between R2-R3 down in advance to control a route-map.

I personally don't know the method but would be interested in other peoples ideas here.

Regards

Toby

Hi Ivan

I have been trying to achieve this in my lab since this was posted and so far have failed.

Can you expand on your idea? I have used the below topology.

R1----R2----R3---R4 | | R5-----------

Traffic can come from 2 source ip addresses on R1 (and i'm not interested in return traffic at this stage) and in my lab terminate on R5 loopback address

R2 has a route map on the interface from R1 so certain source addresses route to R5's loopback route through R3-R4-R5 other source addresses use R2-R5 by default. (I am using OSPF over the network and have tweaked the bandwidth on links to ensure traffic for R5 route viaR4 from R3).

What the original poster is querying and me too now is that as the routemap on R2 uses a next hop for certain source addresses from R1. How can we tweak the config so that traffic entering R2 from R1 does not use the route-map when the link from R3-R4 is down. i.e. to stop this traffic wasting bandwidth between R2-R3 just to be routed back to R2 and onward to R5.

My initial thought was to set the next hop in the route-map to an address of the R4 Interface but as this is not strictly the next hop all I get is hits on the route-map but the routing is ignored via the set statement.

Is there a way to detect distant link failures to stop this bandwidth wastage.

Regards

Toby

Ivan,

Did you mean to create loopback interface on R1 as a placeholder for this network and then redistribute it into RIP? If so, I'm already have it at R1 since this lo is my placeholder for sending my network into ISP with RIP. ISP router is R2 on my topology "drawing". When R2 looses RIP updates from me (R1) then my ISP knows that it must route traffic destined for my network via alternative path which works good. The problem is that R1 is default gateway for several internal networks and using route-map it routes one internal network via RIP to the ISP (R2) and uses default route from ISP's OSPF router (not show on this topology) to route second internal network. I'm seeking for a solution which will enable R1 to reroute traffic originally routed with route-map to default gateway in routing table (OSPF) in the case that path to the Internet via R2 is down. R2 is connected with my R1 with UTP cabling, but R2 and R3 are connected with wireless link, so that I'm not worried about loosing connectivity between R2 and R1, but between R2 and R3. If the communication between R2 and R3 is lost my route-map (on R1) will still forward packers to R2 (which is black hole), since route-map next-hop (R2) is still in the R1's routing table.

Best Regards, Igor

Hi Igor,

Now I have a little better insight in your problem and I see now that my first solution should be altered to get what you need. The problem with it is that you use two mechanisms for basic routing (routing via default route and via route-map).

The thing is that this is a complex situation but solvable since I've solved such things in two ways at some of my customers.

This could be easily solved with PBR Support for multiple tracking options which is available in IOS 12.3T versions. It could be also solved by some other techniques but I think that they would be much more complex than this.

Remember, when you have situation like this it has to be clear that you actually have two components of this problem:

1) to find out if link is "alive" 2) to route traffic in response to (1)

both can be achieved with technique that I've mentioned and can be found at:

I hope this will help.

Ivan, thanks for the link...This is exactly what I have been looking for:)... I think that I saw an article on that subject in some of "Packet Magazine" issues, but I couldn't find it.

Best Regards, Igor