Reported distance, computed distance, feasible distance, feasibility condition, successor, feasible successor

Most material defines feasible distance as the metric of the best path, and that is wrong. RFC 7868 defines it as the least-known total metric to a destination since the route last went from active to passive: a memory, not a measurement, which can fall but not rise while the route is passive. A neighbour is a feasible successor only if its reported distance is less than FD.

Free to read14 min read · 6 min video · lab: EIGRP

By , founder of Daveth · Updated

Opening frame of the video walkthrough of Reported distance, computed distance, feasible distance, feasibility condition, successor, feasible successor

Watch this lesson

6 min · Video walkthrough

In Daveth Pro (Network)

The full 6 min walkthrough is in Daveth Pro (Network). See what the plan opens

Upcoming practical lab

Lab C5: EIGRP

What the lab covers

What this page answers

What is EIGRP feasible distance?

The least-known total metric to a destination from this router since the last transition from active to passive. It is a history, not the current best metric: while the route is passive it can fall or stay where it is, but cannot rise. Read it in “A memory, not a measurement”

What is the EIGRP feasibility condition?

A neighbour's reported distance to the destination must be less than this router's feasible distance (RD below FD). A neighbour reporting exactly the FD fails it. Read it in “One comparison, and why it proves anything”

What is the difference between reported distance and computed distance?

Reported distance is what the neighbour told you, its own distance to the destination. Computed distance is your total through that neighbour: its reported distance plus the cost of your link to it. Read it in “Two numbers in brackets, and one on the line above”

Define feasible distance, right now, before you read on

In your head. One sentence.

If your answer was the metric of the best path to the destination: that is the definition almost everyone carries, it is what most material teaches, and it is wrong. Not approximately. Wrong in a way that makes the rest of EIGRP impossible to reason about, because with that definition the feasibility condition would reject nearly everything and EIGRP would have almost no backup paths at all.

This page exists to replace that one definition, and it does it by making a router remember something.

First, an acronym that ruins conversations. RD is reported distance. You will also see it called advertised distance, abbreviated AD, which is also the abbreviation for administrative distance, the entirely different thing from 1.2.a. Two quantities, one acronym, routinely in the same sentence. This course says reported distance and never abbreviates it.

Two numbers in brackets, and one on the line above

R1 has three neighbours and all three offer 10.1.1.0/24: R2, R3 and R4, which its topology table names by their addresses, 10.0.0.2, 10.0.0.3 and 10.0.0.4. What has R1 written down about each of them?

R1# show ip eigrp topology all-links
EIGRP-IPv4 Topology Table for AS(100)/ID(10.255.0.1)
Codes: P - Passive, A - Active

P 10.1.1.0/24, 1 successors, FD is 2560
        via 10.0.0.2  (2560/256), Serial1/0
! 10.0.0.2 is 256 from the destination; going through it costs you 2560. The difference is your link.
        via 10.0.0.3  (3328/1280), Serial1/1
        via 10.0.0.4  (4096/3072), Serial1/2

Reported distance is what the neighbour told you: its own distance to the destination, which arrived in an EIGRP message and which you wrote down. It is not yours.

Computed distance is what that means for you: the neighbour's reported distance plus the cost of your link to it. Your total, if you send traffic that way. The pair prints (CD/RD): yours first, theirs second.

Your neighbour's distance is not your distance through your neighbour. Say a colleague tells you they are ten minutes from the office. Useful, and it is not your answer, because you still have to get to where they are standing. Their ten minutes plus whatever it takes you to reach them: that is your number.

The comparison stops at the arithmetic. Nobody is walking anywhere, and these are composite metric values rather than minutes or kilometres. But the shape holds: they measured their part of it. They did not measure yours.

Which gives you one relationship to carry into every via line you will ever read: CD ≥ RD, always. Your total through a neighbour cannot be less than that neighbour's own distance, because your link is in it.

And the successor is just the lowest computed distance. 10.0.0.2 at 2560: that is what goes in the routing table. One successor, here.

Now the destination line. FD is 2560, and in this snapshot it happens to equal the best computed distance. That is exactly why the wrong definition survives. It is often true. It is not what FD means.

A memory, not a measurement

Here is the definition, in the specification's own words:

Defined as the least-known total metric to a destination from the current router since the last transition from ACTIVE to PASSIVE state.

Least-known. Since the last transition. That is a history, not a measurement: the smallest computed distance this router has ever seen for this destination since the route last became passive.

And the consequence is the thing to hold: while the route is passive, FD can fall or stay where it is. It cannot rise.

State 1 The link to R2 gets worse

Predict the destination line before you look.

R1# show ip eigrp topology all-links
EIGRP-IPv4 Topology Table for AS(100)/ID(10.255.0.1)
Codes: P - Passive, A - Active

P 10.1.1.0/24, 1 successors, FD is 2560
! The computed distance rose. This number did not.
        via 10.0.0.2  (3072/256), Serial1/0
        via 10.0.0.3  (3328/1280), Serial1/1
        via 10.0.0.4  (4096/3072), Serial1/2

The computed distance through 10.0.0.2 went up, and the metric in the routing table went up with it. The feasible distance stayed at 2560, because 2560 is still the smallest distance this router has ever known for this destination, and that fact did not stop being true because things got worse.

Which numbers in these blocks come from where. Every value in them is computed from the specification's definitions, not captured from a device. The definition says the feasible distance holds while the route stays passive; what IOS XE itself prints as the feasible distance when the successor's own distance rises and it stays the successor was not verified for this page.

State 2 Then the link comes back better than it has ever been

A new record. Watch what it does to a neighbour nobody touched.

R1# show ip eigrp topology all-links
EIGRP-IPv4 Topology Table for AS(100)/ID(10.255.0.1)
Codes: P - Passive, A - Active

P 10.1.1.0/24, 1 successors, FD is 1280
! A new low, so FD moves down: the only direction it moves in.
        via 10.0.0.2  (1280/256), Serial1/0
        via 10.0.0.3  (3328/1280), Serial1/1
        ! Reports 1280. The bar is now 1280. Not below it.
        via 10.0.0.4  (4096/3072), Serial1/2

10.0.0.3 has stopped being a usable backup. Nothing happened on its side of the network. Its reported distance is what it always was, and the link to it costs what it always cost. The bar moved.

That is the first thing the wrong definition cannot explain. If FD were the metric of the best path, then a better best path would be unambiguously good news. It is not: a lower FD is a stronger claim about how close you have been, so fewer neighbours can prove they are closer still.

State 3 And back to exactly where it started

Count what changed.

R1# show ip eigrp topology all-links
EIGRP-IPv4 Topology Table for AS(100)/ID(10.255.0.1)
Codes: P - Passive, A - Active

P 10.1.1.0/24, 1 successors, FD is 1280
! Not available from anywhere in this table.
        via 10.0.0.2  (2560/256), Serial1/0
        via 10.0.0.3  (3328/1280), Serial1/1
        via 10.0.0.4  (4096/3072), Serial1/2

Every link cost is its original value. Every neighbour reports what it reported at the start. Every computed distance in that table is identical to the one on the first step. And the feasible distance is 1280: a distance not currently available from anywhere.

The router remembers being closer than it is. That is not stale data and it is not a bug. It is the mechanism, and the next section is why it is worth having.

Each of those readings carries one feasible distance: the one in force at that moment. Laid side by side, the four look like this:

after                         best CD   FD        backups
as built                      2560      2560      10.0.0.3
the link to R2 worsens        3072      2560      10.0.0.3
and then beats its record     1280      1280      none
and returns to its old cost   2560      1280      none

Read the last column. Same table, different answer. R1 had 1 usable backup on the first step and 0 on the fourth, with every number in the topology table identical. The difference is entirely in what happened in between.

One comparison, and why it proves anything

The Feasibility Condition is met when a neighbor's advertised cost, (RD) to a destination is less than the FD for that destination.

RD < FD. One comparison, against a number the router already has. Work the three neighbours yourself before reading the table, and take the middle one slowly:

neighbour   RD      RD < 1280?  role
10.0.0.2    256     yes         successor
10.0.0.3    1280    no          neither
10.0.0.4    3072    no          neither

Did the middle row catch you? That neighbour reports a distance of exactly the feasible distance, and the answer is still no.

Look hard at the sign, because it is doing real work: less than. Not less than or equal to. A height limit on a car park says below 2 metres. A van measuring exactly 2 metres is not nearly fine, and it is not a judgement call. It does not go in. Equal is close. The less-than sign is not interested in close.

And the comparison stops there: a height bar is about whether you physically fit. This one is about whether a path can be proved not to loop, which is what the rest of this section works out.

A Feasible Successor is a neighbor that meets the Feasibility Condition. A Feasible Successor is regarded as a downstream neighbor towards the destination.

Now: why does that comparison prove anything? It is one of the genuinely elegant ideas in routing, and it is worth working out rather than accepting.

FD is the closest this router has ever been to that destination. If a neighbour reports a distance lower than that, then that neighbour is closer to the destination than this router has ever managed. And a neighbour that is closer than you have ever been cannot be reaching it through you: a path through you would have to be at least as long as your own best, plus the link back.

So the neighbour is genuinely downstream, and it cannot be sending traffic back through you to get there. That is the whole proof, and it fits in one line of arithmetic. Which is why:

Every path meeting the Feasibility Condition is guaranteed to be loop free.

Guaranteed. Not likely: guaranteed, from one comparison against a number the router already has. That is why a feasible successor can be promoted instantly: no query, no computation, no risk.

And that also gives you the two states, which are defined by the condition rather than separately: passive means I have a provably loop-free best path; active means I do not, and I have to go and ask.

A feasible distance belongs to a destination

Not to a router, not to an interface, and not to a link. Which has a consequence worth seeing, because R1 is about to learn a second destination across the same three links.

2001:db8:1::/64 comes up. Same neighbours, same links, same costs. A different destination, and one that has just become passive, so under the definition it has nothing earlier to remember. Its feasible distance is set from where it is now.

neighbour   link    ipv4 (CD/RD)  RD < FD?   ipv6 (CD/RD)  RD < FD?
10.0.0.2    2304    (2560/256)    yes        (2560/256)    yes
10.0.0.3    2048    (3328/1280)   no         (3328/1280)   yes
10.0.0.4    1024    (4096/3072)   no         (3072/2048)   yes

destination       feasible distance   state
10.1.1.0/24       1280                passive
2001:db8:1::/64   2560                passive

Read the two (CD/RD) columns for 10.0.0.3. They are the same pair. Same link, same reported distance, same computed distance, and it is a feasible successor for one destination and not for the other, because the two destinations remember different things.

This is derived, not quoted. The specification defines a feasible distance to a destination; an IPv4 prefix and an IPv6 prefix are two destinations; and the two address families are configured and managed separately. The conclusion follows from those, and this page marks it as followed rather than read.

10.0.0.4 is the sharper case. It reports 3072 for one destination and 2048 for the other, which is not a slip: the network behind a neighbour is not the same shape for both families, so there is no reason for one number to predict the other.

What the guarantee is worth, and what it costs

The link to 10.0.0.2 fails. Both destinations lose their successor at the same instant, across the same two remaining links.

neighbour   link    ipv4 (CD/RD)  RD < FD?   ipv6 (CD/RD)  RD < FD?
10.0.0.3    2048    (3328/1280)   no         (3328/1280)   yes
10.0.0.4    1024    (4096/3072)   no         (3072/2048)   yes

destination       feasible distance   state
10.1.1.0/24       1280                ACTIVE
2001:db8:1::/64   2560                passive

One of them never notices. 10.0.0.4 passes the IPv6 route's feasibility condition, so it is promoted immediately: no query, no computation, no risk, because a neighbour reporting a distance below the best this router has ever managed cannot be routing back through it. That is the guarantee being cashed.

The other goes active. Both remaining paths are still sitting in the IPv4 table. Neither clears the lower bar. So what does the command most people type first show for it?

R1# show ip eigrp topology
EIGRP-IPv4 Topology Table for AS(100)/ID(10.255.0.1)
Codes: P - Passive, A - Active

A 10.1.1.0/24, 0 successors
! No feasible distance on this line: this course has not verified what a device prints there while a route is active, and it is not derivable from the definitions this page is built on.

That is the filtered view, and it is empty, which is precisely the view a person reaches for when asking why was there no backup, and precisely the one that hides the neighbours they are asking about. The other command shows them:

R1# show ip eigrp topology all-links
EIGRP-IPv4 Topology Table for AS(100)/ID(10.255.0.1)
Codes: P - Passive, A - Active

A 10.1.1.0/24, 0 successors
! Active. Two paths reported, neither of them provably safe.
        via 10.0.0.3  (3328/1280), Serial1/1
        via 10.0.0.4  (4096/3072), Serial1/2

Are those paths looped? Almost certainly not. The condition did not say they were. It said this router cannot prove they are not. The feasibility condition is sufficient but not necessary: everything that passes is definitely safe, and things that fail may be safe too. EIGRP chose a conservative test it can evaluate instantly over an exact test it would have to go and compute.

Which produces the failure to recognise: your successor dies, you can see an obvious alternative sitting right there in the topology table, and the route goes active anyway. Nothing is broken. There was nothing to promote, so EIGRP has to start asking.

What it asks, and what that costs, is 1.3.c.i and 1.3.c.iv. This page stops here, and so does its model, which refuses to step past an active route rather than showing you a mechanism it has not planned to teach.

Proves: show ip eigrp topology and show ip eigrp topology all-links give you FD on the destination line and (CD/RD) on each via line. With those two numbers you can classify every neighbour as successor, feasible successor, or neither, which is the central skill in this topic.
Does not prove: that a neighbour missing from the list is unreachable. The plain command shows successors and feasible successors; all-links shows every path reported, including the ones that failed. If the question is "why was there no backup", the first command hides exactly the rows you are asking about.

Take it somewhere else

Two routers offer you a destination. Neighbour A reports 1000, neighbour B reports 1500. Your FD is 1200 and A is your successor, so the link to A costs 200, because 1000 + 200 = 1200. B is not a feasible successor: 1500 is not below 1200.

Now something improves on A's side of the network: a better path opens behind A and A's reported distance drops from 1000 to 700. Your computed distance through A is now 900, a new record, so FD falls to 900.

Does B become feasible? Think before reading on.

No: it gets worse. FD dropped to 900, so the bar B has to clear is lower than before. B at 1500 was 300 above the old bar and is now 600 above the new one.

And notice why the improvement had to happen on A's side. Your computed distance is A's reported distance plus the link, so it can never fall below A's own reported distance: tightening your own link to A can only take you down to A's RD, never past it. CD ≥ RD, on every via line you will ever read.

Which is a real design consequence. Improving your own best path makes it harder for other neighbours to qualify as backups. A network tuned hard for optimal primary paths can end up with fewer feasible successors and more active routes than a slightly less optimal one. Recovering from a failure without asking anyone and having the best primary path are not the same goal, and occasionally they pull against each other.

Watch the number refuse to move

Everything above is six steps and one router. The full experience steps through them, and at any point you can read the topology table, the feasibility check worked out neighbour by neighbour, and the feasible distance across all six steps, where a topology table shows only the current one.

Hold on to this

FD is a memory, not a measurement.

The interactive lesson

The whole lesson, with its drawings and its questions. Free with a Daveth account.

Open the lesson