Redistribution in EIGRP routing protocol

[fabio.marino]

Hi to all.

Studying route redistribution   between different routing protocols i missed some concept.

Contest is redistribution of OSPF in EIGRP protocol.

Following the lab topology:

My target is to redistribuite the OSPF routes into EIGRP.

Based on the official cisco book for ccnp route certification i got that during redistribution the seed metric should be set to a value larger than the largest metric within the receiving autonomous system to help prevent suboptimal routing and routing loops.

So my first task was to calculate the highest metric value within tha EIGRP AS.

After that, i implemented the redistribution under the EIGRP process on R2 border router:

router eigrp 100

redistribute ospf 100 metric "BW" "DE" "REL" "LOAD" "MTU"

My question is: which values of bandwidth, delay, reliability, loading, mtu have i to use? Is there connection with the highest metric used in the AS EIGRP 100?

In different example suggestion is to adopt default value such us:

bw in kbps = 10000

Delay in tens of microseconds = 100

Reliability = 255

Load = 1

MTU = 1500 bytes

But in this way you don't have any connection with the metric values used in tha AS accepting route redistribution.

I'll try searching and reading a lot of documentation but i would like to have opinion about this concept.

Thanks a lot, Fabio.

[cadet alain]

Hi,

you could put all 1's it would work. You only got one exit point here so suboptimal routing is no concern as well as routing loops.

Don't forget you must do mutual redistribution or otherwise use a static route for OSPF routers to know about EIGRP routes.

Regards.

Alain.

[milan.kulik]

Hi,

in a more complex topology where loops could occur, it's recommended to use the Delay value for manipulating the EIGRP metric.

I'm using

bw  =1000000

Delay  = 100000

Reliability = 255

Load = 1

MTU = 1500 bytes

in one of my router configurations which gives a composite metric of 25602560 - far worse than any metric coming from my EIGRP cloud.

You can check the existing EIGRP prefix metric using "sh ip eigrp topo .." command.

HTH,

Milan

[fabio.marino]

Thanks Alain, but i don't understand why you suggest to use value of 1.

You have to consider that under the  "redistribute ospf 100 metric " command values that you can use have minimum values of:

bandwidth = 1

delay = 0

reliability = 0

loading = 1

mtu = 1

and they are not the "K" values used to caluclate the EIGRP metric.

[cadet alain]

Hi Fabio,

What I meant is in your topology you have no suboptimal path or routing loop issues so you can use any seed metric you like but i use all 1s because it's simpler to type ( i know these are not the K values ).

Regards.

Alain.

[fabio.marino]

Thanks Alain for your time,

may be my question is not clear..i would like to discuss about redistribution in EIGRP AS and i'm using the above topology to have a reference. I'm interesting to understand, considering to above topology, which criteria to use to obtain routes redistribution in EIGRP AS especially in the contest of choose metric values such as MTU, reliability, delay, load, bandwidth.

My target is not to solve tha bove topology, but just to get some baseline to have redistribution in EIGRP, that's all.

Now i agree with you about your suggestion but i would like to know how to menage this type of problem using simple guideline, that's all.

[Peter Paluch]

Fabio and Alain,

Please allow me to join the discussion.

Regarding the "seed metric" configured when redistributing from one routing protocol to another, I tell my students that if the redistribution is performed on a single router in the network then the seed metric is largely arbitrary. It can be set to almost anything, provided that there still is sufficient space for this starting metric to grow as the redistributed routes propagate further inside the target routing protocol.

Often, I suggest using metrics that are realistic, i.e. that reflect the metrics of the interface that connects to the source routing protocol's domain. Alternatively, using "typical" metrics derived from a normative 10Mbps or 100Mbps Ethernet interface are completely fine as well.

If there are multiple routers redistributing routes between the same routing protocols then the individual seed metrics depend on the requirements on the connectivity between the routing domains. It may be that one of the redistributing routers is preferred - logically, that router should use the lowest seed metric to win over other routers. The exact value of the seed metric is not that important, rather, it is their relation and their difference that matters.

I would say that it is very much dependent on the expectations how the routing should work in the ideal case.

Best regards,

Peter

[fabio.marino]

Great Peter, you satisfied my need.

I understood that all depend on the contest where are you working for, and may be is very hard to find guidelines or prefixed best practices able to satisfy all request or all topologies.

Now i', rising new question on loop during redistribution and i'm going to open a new discusion.

So far, even though you explained very well basic concepts how i expected, i don't want to close the discussion at all, may be we can gather more useful information or point-of.view based on experience.

Thanks again Peter.