topic changing loopback0 lisp fabric problem in Software-Defined Access (SD-Access)

changing loopback0 lisp fabric problem

p11l — Wed, 11 Sep 2024 13:40:38 GMT

Hello together,

i have running multiple sda sites and already have devices with roles like BN, CP, E and so on.

Now i must change the loopback0 address to another IP.

I tested it and saw that all lisp sessions goes down and devices not able to update lisp configurations.

I believe devices get there lisp configurations whe they are joined to a fabric role.

Did someone know if there is a way how i can push the changed ip address from loopback0 to the new one without remove and readd the fabric device from fabric overview?

Some lines from config lisp is shown like that:

import publication publisher [old-ip]
itr map-resolver [old-ip]
etr map-server [old-ip] key 7 [key] domain-id [domain]
etr map-server [old-ip] proxy-reply
proxy-itr [old-ip]
itr map-resolver [old-ip]
etr map-server [old-ip] key 7 [key]
etr map-server [old-ip] proxy-reply

#show lisp session

Sessions for VRF default, total: 1, established: 0
Peer State Up/Down In/Out Users
[old-ip]:4342 Down never 0/0 16

many thanks and have a nice day!

Re: changing loopback0 lisp fabric problem

Andrii Oliinyk — Wed, 11 Sep 2024 14:15:38 GMT

u r in trouble. Lo0 is used as RLOC & referred to from many places... this makes it almost impossible to change it for BN|CP as u need to update configs on the ENs also.

Re: changing loopback0 lisp fabric problem

p11l — Wed, 11 Sep 2024 14:33:59 GMT

Thanks for that.

Is it pssoible to fix it when i remove roles and so on from devices and readd it to fabric?

Re: changing loopback0 lisp fabric problem

Andrii Oliinyk — Wed, 11 Sep 2024 14:59:50 GMT

never did it. basically Lo0 configs r made for us by LAN-automation. we never care about custom IPs for RLOCs unless it's BN|CP. if i had such task for already operational BN|CP i'd look in direction of removal/reinsertion. if i had task to renumber Lo0s on the ENs i'd write dismissal.

Re: changing loopback0 lisp fabric problem

p11l — Wed, 11 Sep 2024 16:12:28 GMT

ok thanks

did you know if there is a problem to change the lo0 address to a address from the lan automation pool from this site?

I did it and at the pool reservation i noticed under 'unassignable' that it seems to be the address is blocked for automation sessions.

Or is there any other best practice to set lo0 for as you sayed right BN/CP?

Lan-auto does it automatically for devices behind BN/CP and i dont worry about that.

Re: changing loopback0 lisp fabric problem

Andrii Oliinyk — Wed, 11 Sep 2024 16:26:01 GMT

our approach is to leave to DNAC/LAN-automation to assign Lo0s from site's automation pool
we use totally separate IP-ranges not coded in DNAC for Lo0s for BN|CPs for manual assignment

Re: changing loopback0 lisp fabric problem

p11l — Thu, 12 Sep 2024 09:01:23 GMT

thanks for that!

A small draw attached to my post.

Between Site:Transit and Site:A I've configured SDA-Transit on BN.

So I looked with command #show lisp session on the devices and wondering about the Edge Switch in Site:A

BN|CP-1
Peer
172.16.0.3 (TCP-1)
172.16.0.4 (TCP-2)
172.16.1.1 (BN|CP-1)
172.16.1.2 (BN|CP-2)

BN|CP-2
Peer
172.16.0.3 (TCP-1)
172.16.0.4 (TCP-2)
172.16.1.1 (BN|CP-1)
172.16.1.2 (BN|CP-2)

Edge-1
Peer
172.16.0.1 (BN-1)
172.16.0.2 (BN-2)
172.16.1.1 (BN|CP-1)
172.16.1.2 (BN|CP-2)

So I would have expected that the Edge-1 is only connected to BN|CP-1 and BN|CP-2 in his own Site Site:A
And when he wanted a session to the transit site then he should connected to the TCP-1 und TCP-2 (like the BN|CP-1 nd BN|CP-2 did).

Is that like it is expected or is something wrong?