how to EIGRP propagate loopback interfaces?

[jmaxwellUSAF]

Hi.

I've studied the idea of loopback interface. Now it's time i actually implement loopback interfaces network wide.

Please let me know if this makes sense (I'm just being logical, with no exp here.):

We have approximately 100 network devices. Only for nomenclature reasons, I will choose a "loopback subnet" for all my

loopback addresses-- 192.168.250.x.

I will actually make all

addresses 192.168.250.x/32

because these IP addresses will exist throughout my network, and thus I am creating a

discontinuous 192.168.250.x address

space.

-On each device I will add an EIGRP network statement, as an example--

network 192.168.250.77 255.255.255.255

-My expected result is that I will be able to

ping/SSH

into the above example device on

IP address 192.168.250.77

, if any of the L3 ports are up/up.

May you please comment on my proposed details? What is the correct implementation on my intent?

Thank you!

[David Ruess]

Hello,

This is not quite how this would work. A couple things would cause issues:

1. When you do this and you try to log into that address you will log into the closest router (metric wise)  that is advertising that IP address. This wont be an "access any" type situation.

2. If this loopback happens to be the highest IP address and the RID of EIGRP has not been configured then EIGRP will use that (upon next reload). That being the case then as each router reloads they will use the same RID and not exchange routes.

Your best bet is to get a

/24

(or even

/26

since you only have 100 devices) and use those for your loopback addresses to make them unique.

The ONLY reason I see feasible to make the loopback the same is to implement Anycast 

Side note: Your statement of 

network 192.168.250.77 255.255.255.255

 would advertise everything. I think you meant 

network 192.168.250.77 0.0.0.0

 just as a heads up.

Hope this helps

-David

[jmaxwellUSAF]

Thank you for yoru reply.

I didn't mean

x.x.x.77

as an IP for all devices. That was an example of a single device. 

By the way, there are already explicit RID's for all these devices.

So is discussed situation good strategy?

[David Ruess]

My apologies. I misunderstood your network intention. As you have it making the

/32

IPs as loopback address and being advertised into EIGRP then yes, you will be able to

SSH/Ping

into the device as long as there is connectivity on a port and nothing is being filtered out.

I was thrown off by your statement of making a discontinuous network. If everything is  a

/32

its not discontinuous. Its only discontinuous if you have the same network in multiple areas (not to be confused with OSPF Areas), such as part of a

/24

network on one device and another part of the same

/24

on another device but both routers are advertising the

/24

network as a whole.

-David

[paul driver · Accepted solution]

Hello

---

@jmaxwellUSAF wrote:
I've studied the idea of loopback interface. Now it's time i actually implement loopback interfaces network wide

Just take into consideration the current router id (RID) of a rtr running eigrp.

An eigrp rtr primarily prefers an hardcoded RID in its eigrp process over anything else, However if that isnt configured then the rtr will secondarily choose the highest active loopback interface as its eigrp RID,

So one problem that could occur if you append a new loopback on a eigrp rtr that had a non hardcoded RID and had a lower/no loopback or higher physical interface ip address was unpredictable routing issues.

An example of this would be any route advertisement/filtering that was presently using extended access-lists that matched on current advertised rids and prefix(s) in the network, the addition of a new loopback could result in those extended acls eventually not being match correctly.

Hence why it is best practice to hardcode the routing process of a rtr (if applicable) so to negate any unwarranted or unforeseen issues as/when new ip addressing is appended to an active router running dynamic routing.

[ammahend]

Some basic checks, If you have passive interface then make sure it's disabled on loopback, no auto-summary, when you do this across multiple sites then you are adding 100 routes in routing table to each device which might be ok for your network, not a bad thing, just something to keep in mind. Based on how your network is setup, you can also consider possibility of adding a loopback mgmt address per site from local pool itself so it’s already included in site summary instead of adding additional resource in routing table, again I don’t know how feasible it is for your network but something to think about.