cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
2301
Views
10
Helpful
16
Replies

EIGRP FD and FS Calculation and Semantics

kfarrington
Level 3
Level 3

Guys,

Just a couple of question in regards to EIGRP entries in the topology table.

One :-

-------

Would it be correct to say that there is only ONE FD per prefix in the topology table. Ie, If I see the following output from the "show ip eigrp topo all" entries, the "ONE FD for the prefix" FD is 3072 but on the individual lines, each numeric value before the "/" is not the FD, this is just the total metric to the destination network and the entry after the "/" is the RD.

RTR1>sh ip eigrp top all

IP-EIGRP Topology Table for AS(121)/ID(155.195.58.44)

Codes: P - Passive, A - Active, U - Update, Q - Query, R - Reply,

r - reply Status, s - sia Status

P 155.195.44.64/26, 3 successors, FD is 3072, serno 478092

via 155.195.45.93 (3072/2816), Vlan455

via 155.195.45.94 (3072/2816), Vlan455

via 155.195.45.95 (3072/2816), Vlan455

via 155.195.45.91 (28416/28160), Vlan455

via 155.195.45.86 (28416/28160), Vlan455

via 155.195.47.13 (3328/3072), Vlan302

via 155.195.47.5 (3328/3072), Vlan202

via 155.195.45.81 (28416/28160), Vlan455

via 155.195.45.79 (28416/28160), Vlan455

via 155.195.45.75 (28416/28160), Vlan455

via 155.195.45.71 (28416/28160), Vlan455

via 155.195.45.76 (28416/28160), Vlan455

via 155.195.45.92 (28416/28160), Vlan455

via 155.195.45.82 (28416/28160), Vlan455

Ie, in the output above, we have ONE FD value of 3072, and the other values of 28416, 3328 you cannot call these feasible distances and would just call them metrics to the destination.

OR is is better to use the term there is only one ACTIVE FD per prefix and all the other entries above, ie 28416 and 3328 are FDs for that particular neighbor?

Two:

-------

Please see the diagrams attached for the age old question about RD == FD and this is not an FS.

Firstly on the diagram "EIGRP Topology- No Feasible Successor.jpg" as FD == RD, there is no FS and thus when the successor link is brought down, EIGRP queries for a new path and this is then indicated by the fact a NEW FD is calculated for the prefix as is seen in the topology table.

Secondly, on the diagram "EIGRP Topology- With a Feasible Successor.jpg" you see that when the FS takes over (no EIGRP query) in the topology table the FD does NOT change but the routing table metric does.

So the clarification here is that is it safe to assume when an FS takes over from a successor, you can tell this in the output of the eigrp topology table as the prefix metric will be HIGHER than the feasible distance and the reported distance will be LOWER than the FD. This I think is an important statement to clarify.

Three:

-------

In regards to question two, If a prefix successors route goes down and there is a feasible successor, as mentioned, the FD for the prefix will NOT change. But then say that the successor does not come back into operation, say a decommissioned link, so the FS (which is now the S) is now the preferred path and will remain that way for ever. Is there a timer value for DUAL to recalculate the route - realigning a new FD for the prefix in the topology table and finding any new FSs than now align with the new FD.

I hope this does make sense and please let me know if anyone needs further clarification.

Kind regards,

Ken

2 Accepted Solutions

Accepted Solutions

pkhatri
Level 11
Level 11

Hi Ken,

I've tried to address all your points below.

One

---

Yes, there is only one FD per prefix. As you correctly stated "each numeric value before the "/" is not the FD, this is just the total metric to the destination network and the entry after the "/" is the RD"

You cannot indeed call the other non-3072 values the FDs - calling them metrics is the better term. Your first statement is the correct one, not the second one. The other metrics cannot be called the FDs for those particular neighbors.

Two

---

Agree with your first statement- regarding "EIGRP Topology- No Feasible Successor.jpg"

As for "EIGRP Topology- With a Feasible Successor.jpg", what you have to keep in mind is that the FD is the minimum distance from the current router toward the destination since the last time the router transitioned from active to passive state for that destination. That is, this is the lowest metric since the last diffusing computation was performed for that route. In the case with a FS, only a local computation was performed, so the FD did not change.

Agree with your last statement here too.

Three

-----

If the current successor (which has failed) does not come back, it is possible that the new S (which was the old FS) will remain that way forever. EIGRP does not perform periodic DUAL computations. The only time this will happen is if this new successor goes down and there are no more FSs for the route - a full DUAL computation will be performed and the FD will once again equal the best metric.

For a real good understanding of some of this, it may be advisable to read some of the original work on DUAL by J. Garcia-Lunes-Aceves.

Hope that answers some of your queries.

Paresh

View solution in original post

Hi Ken,

I believe your scenario would indeed work as you have stated and the new path would not be considered a FS in this case. That is the price you (i..e EIGRP) pay for taking short cuts. Local computations are an optimization used to speed up convergence and therefore, can result in such situations. Until a full diffusing computation is done, this situation will prevail. Remember that the principal aim of the exercise is to ensure that whatever path you pick is loop-free. That is indeed satisfied by the local computation process.

The one thing to note is that even in the case above, routing will still be optimal. It will, however, have an impact if you have configured a variance > 1 and that second path is not considered feasible and therefore not installed in the routing table.

In your case above, if you lose the current successor, DUAL will run and eventually sort things out.

Hope that helps - pls rate the post if it does.

Regards,

Paresh

View solution in original post

16 Replies 16

pkhatri
Level 11
Level 11

Hi Ken,

I've tried to address all your points below.

One

---

Yes, there is only one FD per prefix. As you correctly stated "each numeric value before the "/" is not the FD, this is just the total metric to the destination network and the entry after the "/" is the RD"

You cannot indeed call the other non-3072 values the FDs - calling them metrics is the better term. Your first statement is the correct one, not the second one. The other metrics cannot be called the FDs for those particular neighbors.

Two

---

Agree with your first statement- regarding "EIGRP Topology- No Feasible Successor.jpg"

As for "EIGRP Topology- With a Feasible Successor.jpg", what you have to keep in mind is that the FD is the minimum distance from the current router toward the destination since the last time the router transitioned from active to passive state for that destination. That is, this is the lowest metric since the last diffusing computation was performed for that route. In the case with a FS, only a local computation was performed, so the FD did not change.

Agree with your last statement here too.

Three

-----

If the current successor (which has failed) does not come back, it is possible that the new S (which was the old FS) will remain that way forever. EIGRP does not perform periodic DUAL computations. The only time this will happen is if this new successor goes down and there are no more FSs for the route - a full DUAL computation will be performed and the FD will once again equal the best metric.

For a real good understanding of some of this, it may be advisable to read some of the original work on DUAL by J. Garcia-Lunes-Aceves.

Hope that answers some of your queries.

Paresh

kfarrington
Level 3
Level 3

Paresh, Mahy thx indeed for that. You are most helpful.

Just an update on the current topology table entry and FD since yesterday (in regards to Question 3)

The FD is still 156160 and the routing metric is 158464/155904 so there has been no aging out of this entry as yet to indicate that this link may have been decommisioned as you have stated. (output below)

One last thing, if I may,

Also, looking at the "show ip eigrp top detail-links" there seems to be stats associated with the topology entry for which I have no idea what they mean and if this has any relation to anything.

Paresh, Once again, many thx for the input :)

Paresh, if you or anyone could help with this other point, I would be very greatful indeed.

Kindest regards,

Ken.

S1Rtr1#sh ip eigrp top det

IP-EIGRP Topology Table for AS(10)/ID(192.168.12.129)

Codes: P - Passive, A - Active, U - Update, Q - Query, R - Reply,

r - reply Status, s - sia Status

P 10.192.221.0/24, 1 successors, FD is 156160, serno 314, Stats m(20)M(452)A(128)c(6)

via 10.192.30.2 (158464/155904), FastEthernet0/0

P 10.192.30.0/24, 1 successors, FD is 28160, serno 166

via Connected, FastEthernet0/0

P 10.192.20.0/24, 1 successors, FD is 28416, serno 318

via 10.192.30.2 (30464/27904), FastEthernet0/0

S1Rtr1#

S1Rtr1#show ip route 10.192.221.0

Routing entry for 10.192.221.0/24

Known via "eigrp 121", distance 90, metric 158464, type internal

Redistributing via eigrp 121

Last update from 10.192.30.2 on FastEthernet0/0, 19:03:59 ago

Routing Descriptor Blocks:

* 10.192.30.2, from 10.192.30.2, 19:03:59 ago, via FastEthernet0/0

Route metric is 158464, traffic share count is 1

Total delay is 5190 microseconds, minimum bandwidth is 100000 Kbit

Reliability 255/255, minimum MTU 1500 bytes

Loading 1/255, Hops 2

S1Rtr1#sh ip eigrp top 10.192.221.0 255.255.255.0

% IP-EIGRP (AS 10): Route not in topology table

IP-EIGRP (AS 121): Topology entry for 10.192.221.0/24

State is Passive, Query origin flag is 1, 1 Successor(s), FD is 156160

Routing Descriptor Blocks:

10.192.30.2 (FastEthernet0/0), from 10.192.30.2, Send flag is 0x0

Composite metric is (158464/155904), Route is Internal

Vector metric:

Minimum bandwidth is 100000 Kbit

Total delay is 5190 microseconds

Reliability is 255/255

Load is 1/255

Minimum MTU is 1500

Hop count is 2

S1Rtr1#

Also,

If we know the following formula and this is infact correct?

Current metic (in topo or RT) to destination prefix < current prefix FD = router is using an FS to destination network (dual has not run)

Can we not have an add on to the "show ip eigrp topology" command to indicate any prefix in this state? Either just with a flag with the prefix in the output of the original command or use a new argument to the existing commands? Providing the statement will always meet the criteria?

Just a thought?

Regards,

Ken

I believe you meant:

Current metic (in topo or RT) to destination prefix > current prefix FD = router is using an FS to destination network (dual has not run) (you had a 'less than' sign)...

Your statement is correct that if the above formula holds true, the current successor is the result of a local computation. However, I would not say that the router is using an 'FS' to the destination network. The FS used to be an FS; after the local computation, it is the successor (apart from the different FD, the router has no recollection of who the previous successor was).

As to the second part of your post, I'm not sure what such a flag would achieve ? Does it matter that a route is using a path determined using local computation ? I don't think so. It really has no bearing on anything and besides, it is implied by the difference in the FD and the current metric.

Hope that helps .. pls rate posts that help.

regards,

Paresh.

Yes, sorry I did mean Current Metric > FD :)

To the second point, Cool, I was just scratching at anything that may be worth while, I suppose it is not :)

Top-stuff fella

Apprieciate it :)

Ken

I'm a bit confused, I think. You're saying you see a route or routes where the FD is not equal to the current metric? Can you provide an output that shows this? In reality, you would have to catch the router in the state where an update has been received, and the route has been installed in the topo table, but DUAL hasn't yet been called to resort the topo table. That's normally a period of a few milliseconds, so you'd be lucky to catch this once every few years--I've been working on EIGRP networks for 10 years+, and I've never seen this state in a real, live, network, nor even in a lab network.

Can you explain further?

Thanks!

:-)

Russ.W

Hi Russ,

What we are saying is that we have a prefix that has a successor and an FS. We down the sucessor and the FS takes over (no query run) - normal behaviour.

Its just that now the FS becomes the S, but the FD for the prefix in the topology table remians as the original FD (becuase dual has not run) and the metric of the route is higher than the FD.

I have attached a before and after we fail the sucessor route.

Thx very much - let me know your thoughts if you dont mind :)

Regards,

Ken

Hi Russ,

I'm afraid I will have to disagree with you on this. I do not believe that this is a transient condition at all. When a router performs a local computation, it selects a new metric (based on its FS) but the new FD is:

FD = min(previous FD, metric of new path)

Since the FS will always have a worse metric, this implies that the old FD will remain resulting in it being non-equal to the current metric.

I have seen this on numerous occasions. In addition, I believe this to be the correct behaviour.

This is also pretty clearly stated in a number of publications including:

'Loop-Free Routing Using Diffusing Computations'

by J. J. Garcia-Lunes-Aceves

which appeared in IEEI/ACM TRANSACTIONS ON NETWORKING, VOL. 1, NO, 1, FEBRUARY 1993

I'm happy to be proven wrong as this caused me a bit of grief in understanding when I was first leaning EIGRP !

Paresh.

Aha! You're talking strictly about the local topology table, not the routing table, etc. Okay... You are correct, if we switch from a successor to an FS, we will not run DUAL, so the routing entry in the topology table (and the topology table only) doesn't change. This means the FD we are comparing against to find FS' will not change.

However, the cost in the routing table will change, and we will send an update when we switch to an FS, to all of our neighbors. I've never, however, seen the metric in the routing table not match the route metric in the topology table, which is what I thought someone was saying... :-)

The correct term for the FS in the routing table is the metric, or cost. For the composite metric in the network entry in the topology table, it's the feasible distance. For the composite metric for the best route in the topology table, it's the successor metric (which can be different from the feasible distance, as noted here). The remainder are just composite route metrics (we always call non-feasible routes "feable successors").... :-)

The bandwidth, delay, etc., are called vector metrics.

HTH.

:-)

Russ.W

Hi Guys, This is very interesting :)

This is why I was asking the point that in theory, the topology table FD could be out of date what could happen and if there was any aging timer:-

1. Normal operation with two Paths, 1 S and 1 FS

2. If there was a link that was decommisioned, and an FS took over - Then the new metric of the route (which is not the FD) is set.

3. Then potentially a newly learned path was installed, that DID NOT meet the FC as the old FD remaining in the topology table, but if the FD was set correctly, it would.

Example below, but I have not tested in the LAB so I assume this could be rubbish stated below :)

Example :-

step 1 - two paths available in topo table

rtr 1 has in topo table

prefix 10.192.221.0/24 metrio 290/190 via path A

prefix 10.192.221.0/24 metrio 200/100 via path B

FD = 200

Route metric 200

Sucessor is via Path B

Feasible Sucessor is via Path A

Step 2 - Sucessor goes down,

rtr 1 has in topo table

prefix 10.192.221.0/24 metrio 290/190 via path A

FD = 200 (NOT 290)

Route Metric is 290

Sucessor is via Path A

no Feasible Sucessor

Step 3 - New Path is introduced via Path X

(Path would be a FS if FD was reset and met the FC (ie RD is less than FD), but is NOT FS as old FD still in effect and does not meet the FC)

rtr 1 has in topo table

prefix 10.192.221.0/24 metrio 290/190 via path A

prefix 10.192.221.0/24 metrio 310/210 via path X

FD = 200 (NOT 290)

Route Metric is 290

Sucessor is via Path A

no Feasible Sucessor ** as OLD FD is in effect

Thanks again guys,

Ken

Hi Ken,

I believe your scenario would indeed work as you have stated and the new path would not be considered a FS in this case. That is the price you (i..e EIGRP) pay for taking short cuts. Local computations are an optimization used to speed up convergence and therefore, can result in such situations. Until a full diffusing computation is done, this situation will prevail. Remember that the principal aim of the exercise is to ensure that whatever path you pick is loop-free. That is indeed satisfied by the local computation process.

The one thing to note is that even in the case above, routing will still be optimal. It will, however, have an impact if you have configured a variance > 1 and that second path is not considered feasible and therefore not installed in the routing table.

In your case above, if you lose the current successor, DUAL will run and eventually sort things out.

Hope that helps - pls rate the post if it does.

Regards,

Paresh

Paresh, Thats great.

I know its only a small point, and I think it is important to understand mathematically how this can affect convergence when using EIGRP when designing various network topologies.

Many thx indeed,

Ken

t

Paresh, Here are the lab results - mess 1 of two

Getting Started

Find answers to your questions by entering keywords or phrases in the Search bar above. New here? Use these resources to familiarize yourself with the community: