Question regarding Routing Loops

bermovick

I'm reviewing Lammle's CCENT book and on page 393 it discusses routing loops. I understand the concept, but not the example he uses.

For those without the book, here's the diagram in ASCII (forgive the periods used for formatting).
A---B---C---E---
..\./
...D
B-C network is network 3, C-E is network 4, E-?? is network 5.

In the example, network 5 drops. E tells router C there's no route to 5 anymore. Routers A, B and D still think there's a route. So far so good (packets get dropped upon reaching C). Eventually C lets B know there's no route to network 5 - but then when A sends out it's update, B and D think there's a route to network 5 through A. I understand this, -BUT- what I don't understand is why router C doesn't think there's a route through router B? (assuming B sent an update before C)? Or am I over-analyzing this?

chmorin

bermovick wrote: »

I'm reviewing Lammle's CCENT book and on page 393 it discusses routing loops. I understand the concept, but not the example he uses.

For those without the book, here's the diagram in ASCII (forgive the periods used for formatting).
A---B---C---E---
..\./
...D
B-C network is network 3, C-E is network 4, E-?? is network 5.

In the example, network 5 drops. E tells router C there's no route to 5 anymore. Routers A, B and D still think there's a route. So far so good (packets get dropped upon reaching C). Eventually C lets B know there's no route to network 5 - but then when A sends out it's update, B and D think there's a route to network 5 through A. I understand this, -BUT- what I don't understand is why router C doesn't think there's a route through router B? (assuming B sent an update before C)? Or am I over-analyzing this?

Which routing protocol is this?

alan2308

Are we talking RIP (I assume so)?

Router B has network 5 marked dead. And since the hold down timer hasn't expired yet, its going to ignore the route to network 5 coming from Router A because its an inferior route compared to the one that i knew from Router C.

bermovick

chmorin wrote: »

Which routing protocol is this?

The book hadn't yet gotten into the specific protocol (RIP/RIP2) covered. This is discussing distance-vector protocols in general.

bermovick

alan2308 wrote: »

Are we talking RIP (I assume so)?

Router B has network 5 marked dead. And since the hold down timer hasn't expired yet, its going to ignore the route to network 5 coming from Router A.

The example specifically states that router B will accept router A's update, causing the routing loop (A has a route to B, B has a route back to A).

Local Native

bermovick wrote: »

I'm reviewing Lammle's CCENT book and on page 393 it discusses routing loops. I understand the concept, but not the example he uses.

For those without the book, here's the diagram in ASCII (forgive the periods used for formatting).
A---B---C---E---
..\./
...D
B-C network is network 3, C-E is network 4, E-?? is network 5.

In the example, network 5 drops. E tells router C there's no route to 5 anymore. Routers A, B and D still think there's a route. So far so good (packets get dropped upon reaching C). Eventually C lets B know there's no route to network 5 - but then when A sends out it's update, B and D think there's a route to network 5 through A. I understand this, -BUT- what I don't understand is why router C doesn't think there's a route through router B? (assuming B sent an update before C)? Or am I over-analyzing this?

I just read this section an hour ago, so I'll try my hand at explaining my analysis. I believe you are simply analyzing it one-step further than Lammle's example (b/c he proved the routing loop in less steps).

Network 5 goes down. E tells C no route. A,B, and D do not know yet. They will keep sending out their updates. C sends out its update telling B that Network 5 is down. A sends out its information (that it thinks it can get to Network 5 through , and B goes "You know Network 5, I've been looking for it!" and adds it to its routing table as going through Router A (and you have the routing loop). Now your question is why doesn't Router C think theres a route through Router B? I believe after Router C's route flush timer goes off, it erases Network 5 from its routing table, and will eventually get stuck into the routing loop through B's routing table claiming it can connect to Network 5 through A. Or, Router C will send outs its update that Network 5 doesn't exist, but by the time B updates its message, A already sends outs its message stating Network 5 is on the table (thus creating a routing loop between A and B, with C always trying to fight it saying "No! Network 5 is connected to me and is down")

I'll have to test out the impact of the route flush timer on a routers memory, whether it deletes everything even if the network is still plugged into an interface and disabled, or if the routing loop is just causing problems via C can't get its message to A (an extra hop) by the time A already tricks B due to the interval of update timers.

I imagine its the latter scenario, with the constant fight of C telling B there's no route, and A sending its update, and B becoming an idiot and believing A before it forwards C's message to everyone.

bermovick

Yeah; I understood that; what I didn't get is why he took it all the way to the A<-->B update? Why wouldn't B's update cause the same issue with C?

Note he hadn't discussed any of the things that prevent this from happening (poisoning, holddowns, etc).

Another real-life question. If RIPs update timer is 30 seconds, how can I create a topology in GNS3, set up router RIP on one of the furthest routers from my bridged NIC/loopback, and immediately log in to the furthest router (network-ally) from that and already see it's route? There's 2 other routers between the 2 (1 virtual, 1 physical)

wbosher

If you can, try to get your hands on CBT nuggets. Jeremy explains routing loops with RIP really well, in an easy to understand way.

alan2308

bermovick wrote: »

Yeah; I understood that; what I didn't get is why he took it all the way to the A<-->B update? Why wouldn't B's update cause the same issue with C?

Note he hadn't discussed any of the things that prevent this from happening (poisoning, holddowns, etc).

Another real-life question. If RIPs update timer is 30 seconds, how can I create a topology in GNS3, set up router RIP on one of the furthest routers from my bridged NIC/loopback, and immediately log in to the furthest router (network-ally) from that and already see it's route? There's 2 other routers between the 2 (1 virtual, 1 physical)

I think the key point is that Router C still knows the route to Network 5 through router E at this point so it doesn't care about the route that its getting from router B yet because its an inferior route (higher hop count). It took the route out of its routing table, but it still knows the route.

For your second question, RIP supports triggered updates, which means that as soon as a router sees a change to the network, it sends an update to its neighbors immediately, rather than waiting for its next scheduled update.

bermovick

I think I'm understanding, and I'm just over-analyzing it. I think it's one of my gripes with Lammle's writing style is you have to know more than what he's told you for his examples to make sense.

Regarding the 2nd answer, that makes sense and was what I was suspecting it was doing. Thanks for verifying

alan2308

bermovick wrote: »

I think I'm understanding, and I'm just over-analyzing it. I think it's one of my gripes with Lammle's writing style is you have to know more than what he's told you for his examples to make sense.

What he's doing here is introducing the basic concepts of a DV protocol. It's all hypothetical at this point, he's working with a completely made up and brain dead routing protocol. The basic concepts are whats important here. What does every router know at the point network 5 disappeared? 30 seconds later, every router sends an update, so what does every router know now? 30 second later every router sends another update.... and so on. Im not sure we know enough specifics about this pretend protocol to really answer your question anyway.

But, you see what routing loops are and how they can happen. You see how slowly a DV protocol converges after a change happens. Now you'll move on to real protocols where the specifics are known.