Poisoned reverse

Have a question or want to start a discussion? Post it! No Registration Necessary.  Now with pictures!

Threaded View

I don't quite see the point of poisoned reverse.

Scenario (with RIP):
- 2 routers with a hub (and me tracing) between them.
- One router is turned off.
- After 3 mins, the remaining router realizes its partner is dead, and then
sends the poisoned reverse (to the dead router), informing it that is is no
longer alive (hopcount=16).

What I don't understand is, if someone is dead, what is the point of
informing them of this fact (rip does this twice (60 secs apart), and then
clears the entry from its routing table; it also sends a poisoned route to
its other mates, and that makes perfect sense).


Re: Poisoned reverse

Nospa wrote:
Quoted text here. Click to load it

The is no single mechanism is adequate for routing loop avoidance. And
route poisoning alone doesn't cut it, and I thinink you already know
this. Along with triggered updates, route poisoning accomplishes good
convergence time.
Regarding poison reverse,
The poison reverse is an acknoledgment of reception of route
poisoning. And ALL routers on the segment connected to the router
issuing route poison MUST send back a poison reverse. If the routers
dont send poison reverse, this means that the router has lost the
right way to reach the dead network. This is considered neccessary in
distance vector protocols because they don't see the full scope of the

I hope this helps.


Re: Poisoned reverse
Mohammed Alani wrote:
Quoted text here. Click to load it

Very interesting question, I think its quite confusing myself, but I
think the answer is that the poison reverse is a required rule and the
neighbour router has no concept of a router being turned off. Its a very
simple algorithm. So its just sends it anyway.

The point of poison reverse is to inform the dead router that there is
no route via the neighbour to the dead routers network.

Re: Poisoned reverse
Quoted text here. Click to load it


Poison Reverse prevents loose loops.

Think of something like this
|- rtrA  <---> rtrB <---> rtrC -|

if rtrC link to net fails it sends a flash update with a poisoned
route  to rtrB, which in turns sends a poisoned route to either rtrC and
rtrA which also responds to rtrB with a poisoned route. The poisoned route
is then advertised with the scheduled updates for xx times.

Poison Reverse can also work ( cisco routers don't do this if I remember
correctly ) with Split Horizon advertising incoming routes out the SAME
interface BUT with an infinite metric.


Site Timeline