IPv6 link-local address on point-to-point interface

[CSCO12350797]

Hello,

I'm a little bit confused. I know that there are two ways of assigning an IPv6 link-local address on interface. We can add it manually, or it is by default created automatically using modified EUI-64. But how is this address made when there is no MAC address on point-to-point link?

Thanks.

Martin

[Harold Ritter · Accepted solution]

Hi Martin,

Since link local addresses are only locally significant, the router can borrow a mac address from any local ethernet port.

Regards

[CSCO12350797]

Thanks I though that it willl work like this but does it have some rules or it is choosen randomly? Aren´t there some problems, for example if there are two routers connected with two links? Can IGP and EGP routing protocols handle this if there are two links, lets say with the same link-local address?

[Harold Ritter]

Hi Martin,

This is not an issue. As I mentioned, the LLA is context based (per interface), so you could basically use the same LLA on all interfaces of the same routers. The only important thing is for that LLA to be unique on a given link. Some service provider configure the LLA manually and set them all to fe80::1 on one side of the link and fe80::2 on the other.

Regards

[CSCO12350797]

Hi Harold,

thanks I really wasn't sure about this. My last question to you is: What is it good for when providers set these addresses manually? Why they don't let them like they are?

Martin

[Harold Ritter]

Hi Martin,

The advantage of using manually is that it are

1. easier for someone to remember and type ffe80::1 vs something like fe80::aabb:ccff:feaa:bbcc.

2. the LLA does not change if the ethernet port (module) is replaced on the router. This would be an important point if some configuration (BGP, ACL) use LLA.

Regards

[CSCO12350797]

Hello,

sorry but I don't think that these reasons are reasonable. Why do they need to remember these addresses when are they used only on a link with internal processes? They cannot be used for management purposes, so I would say that it is easier for me to configure one loopback with global address at each router which would be easy to remember, announce it to IGP and it can be used for management, so as for BGP..and I don't need to care about all link local addresses I'm using in my network, which might contain hundreds of links..

[Harold Ritter]

Hi Martin,

If you do not think they are reasonable reasons then do not use manually configurred LLA in your network ;-)

Seriously, LLAs are not only used internally. I used them very frequently to test connectivity on a given subnet. And I have also see people used them for BGPsession, as I previously mentioned.

At then end of the day, you this is not something you like or are confortable with, you always have the option of not doing it.

Regards and best wishes for the holidays

[CSCO12350797]

Yes, you can try connectivity with it, but if CDP runs, which by default is, you have other options how to test it..personally I think that there are some situations where you can find it usable, but it all depends on your design..I thnik that network can be fully functional even if LLC are not configured manually..

However, it was nice to write with you, wish you merry christmas and happy new year too..

Martin

[Harold Ritter]

Hi Martin,

Bear in mind that CDP is normally turned off in SP networks. Network design is not a "one size fits all" and what is good for one is not necessarily good for the other.

It was nice discussing with you and I hope you can get some rest and peaceful time with family during the holidays.

Cheers

[CSCO12350797]

Yes, I know about it, there are some security issues with CDP and its safer not to use it, but now I was thinking more generally..of course it depends on designer of the network, that's the reason why I'm so curious..:)

Byee