Has anyone implemented lab or production IPv4 next hops over an IPv6 only IX vlan with BIRD as the route server? I'm interested to hear your experiences. Specifically router vendor interop. Jay
* Jay Hanke
Has anyone implemented lab or production IPv4 next hops over an IPv6 only IX vlan with BIRD as the route server?
I'm interested to hear your experiences. Specifically router vendor interop.
I haven't, but I cannot begin to fathom how that could possibly work. How can the receiving router possibly resolve an IPv4 next-hop address to an destination Ethernet MAC address, if the interface facing the IX does not have any IPv4 addresses assigned? What would the next-hop IPv4 address be set to, and how is that determined? Tore
I haven't, but I cannot begin to fathom how that could possibly work.
https://www.ietf.org/archive/id/draft-chroboczek-int-v4-via-v6-01.html
How can the receiving router possibly resolve an IPv4 next-hop address to an destination Ethernet MAC address, if the interface facing the IX does not have any IPv4 addresses assigned?
A router only needs to know the l2 address to forward a frame. So there would be no ipv4 address on the transitory network just an ipv6 that resolves to a Mac address. Kind of like ipv4 unnumbered but with ipv6 addresses on the segment. Multiprotocol bgp would be used for the next hops.
I have used rfc5549 extensively in production, but not with bird. Is that the same setup you're talking about here? On Tue, Apr 23, 2024, 09:31 Jay Hanke via Bird-users <bird-users@network.cz> wrote:
I haven't, but I cannot begin to fathom how that could possibly work.
https://www.ietf.org/archive/id/draft-chroboczek-int-v4-via-v6-01.html
How can the receiving router possibly resolve an IPv4 next-hop address to an destination Ethernet MAC address, if the interface facing the IX does not have any IPv4 addresses assigned?
A router only needs to know the l2 address to forward a frame. So there would be no ipv4 address on the transitory network just an ipv6 that resolves to a Mac address.
Kind of like ipv4 unnumbered but with ipv6 addresses on the segment. Multiprotocol bgp would be used for the next hops.
Probably better question. Is RFC5549 supported by BIRD? -Chris
On Apr 23, 2024, at 9:59 AM, Jay Hanke via Bird-users <bird-users@network.cz> wrote:
On Tue, Apr 23, 2024 at 8:45 AM Darren O'Connor <mellow.drifter@gmail.com> wrote:
I have used rfc5549 extensively in production, but not with bird. Is that the same setup you're talking about here?
Yes. Are there any IXPs doing this right now with bird2?
* Chris Malayter
Probably better question. Is RFC5549 supported by BIRD?
Yes, RFC 8950 is on the list of supported standards, and can be configured with the «[require] extended next hop» configuration option. https://bird.network.cz/?get_doc&v=20&f=bird-6.html#ss6.4 No reason whatsoever this wouldn't work as well on an IX as it would on any other Ethernet link. IX-es aren't that special, except maybe needing to cater to customers running BGP implementations not updated since the 1990s. Tore
Thank you!
On Apr 23, 2024, at 10:28 AM, Tore Anderson <tore@fud.no> wrote:
* Chris Malayter
Probably better question. Is RFC5549 supported by BIRD?
Yes, RFC 8950 is on the list of supported standards, and can be configured with the «[require] extended next hop» configuration option.
https://bird.network.cz/?get_doc&v=20&f=bird-6.html#ss6.4
No reason whatsoever this wouldn't work as well on an IX as it would on any other Ethernet link. IX-es aren't that special, except maybe needing to cater to customers running BGP implementations not updated since the 1990s.
Tore
* Jay Hanke
I haven't, but I cannot begin to fathom how that could possibly work.
https://www.ietf.org/archive/id/draft-chroboczek-int-v4-via-v6-01.html
How can the receiving router possibly resolve an IPv4 next-hop address to an destination Ethernet MAC address, if the interface facing the IX does not have any IPv4 addresses assigned?
A router only needs to know the l2 address to forward a frame. So there would be no ipv4 address on the transitory network just an ipv6 that resolves to a Mac address.
Kind of like ipv4 unnumbered but with ipv6 addresses on the segment. Multiprotocol bgp would be used for the next hops.
In your original message you said you wanted to advertise routes *IPv4* next-hops over an IPv6-only IX. The I-D you are linking to are not about that, but about advertising IPv4 routes with an *IPv6* next-hop. That is the exact opposite of what you asked about. Advertising IPv6 next-hops over an IPv6-only IX is of course completely unproblematic, since the destination MAC address can then be resolved with standard ICMPv6 ND. Advertising IPv4 prefixes with IPv6 next-hops is also entirely unproblematic, assuming the routers involved support extended next-hop encoding (RFC 5549/8950). What you did ask about initially, though, that is, using IPv4 next-hops across an IPv6-only IX, does seem impossible to me. Tore
I mistyped. Should be IPv6 hops for IPv4 routes. On Tue, Apr 23, 2024 at 9:10 AM Tore Anderson <tore@fud.no> wrote:
* Jay Hanke
I haven't, but I cannot begin to fathom how that could possibly work.
https://www.ietf.org/archive/id/draft-chroboczek-int-v4-via-v6-01.html
How can the receiving router possibly resolve an IPv4 next-hop address to an destination Ethernet MAC address, if the interface facing the IX does not have any IPv4 addresses assigned?
A router only needs to know the l2 address to forward a frame. So there would be no ipv4 address on the transitory network just an ipv6 that resolves to a Mac address.
Kind of like ipv4 unnumbered but with ipv6 addresses on the segment. Multiprotocol bgp would be used for the next hops.
In your original message you said you wanted to advertise routes *IPv4* next-hops over an IPv6-only IX.
The I-D you are linking to are not about that, but about advertising IPv4 routes with an *IPv6* next-hop. That is the exact opposite of what you asked about.
Advertising IPv6 next-hops over an IPv6-only IX is of course completely unproblematic, since the destination MAC address can then be resolved with standard ICMPv6 ND.
Advertising IPv4 prefixes with IPv6 next-hops is also entirely unproblematic, assuming the routers involved support extended next-hop encoding (RFC 5549/8950).
What you did ask about initially, though, that is, using IPv4 next-hops across an IPv6-only IX, does seem impossible to me.
Tore
Ciao Tore, if I am not mistaken this is rather easy and "standard": - Router knows a.b.c.d/x is supposed to be routed via 2001:db::1 - Router can lookup the mac address of 2001:db::1 (usual icmp6) - Router sends packet for a.b.c.d/x towards mac address of 2001:db::1 That's it. Cheers, Nico p.s.: The first few weeks it feels like magic, the following years that method feels like AMAZINGLY USEFUL magic ;-) Tore Anderson <tore@fud.no> writes:
* Jay Hanke
Has anyone implemented lab or production IPv4 next hops over an IPv6 only IX vlan with BIRD as the route server?
I'm interested to hear your experiences. Specifically router vendor interop.
I haven't, but I cannot begin to fathom how that could possibly work.
How can the receiving router possibly resolve an IPv4 next-hop address to an destination Ethernet MAC address, if the interface facing the IX does not have any IPv4 addresses assigned?
What would the next-hop IPv4 address be set to, and how is that determined?
Tore
-- Sustainable and modern Infrastructures by ungleich.ch
On Fri, Apr 19, 2024 at 02:29:44PM -0500, Jay Hanke via Bird-users wrote:
Has anyone implemented lab or production IPv4 next hops over an IPv6 only IX vlan with BIRD as the route server?
I'm interested to hear your experiences. Specifically router vendor interop.
Note that there is an Euro-IX working group targetting this issue: https://github.com/euro-ix/rfc8950-ixp -- Elen sila lumenn' omentielvo Ondrej 'Santiago' Zajicek (email: santiago@crfreenet.org) "To err is human -- to blame it on a computer is even more so."
To me, the better part is to imagine that IPv6 only Link-Layer(on IXPs and Transit providers) will make PMTUD better working. Em ter., 23 de abr. de 2024 11:57, Ondrej Zajicek via Bird-users < bird-users@network.cz> escreveu:
On Fri, Apr 19, 2024 at 02:29:44PM -0500, Jay Hanke via Bird-users wrote:
Has anyone implemented lab or production IPv4 next hops over an IPv6 only IX vlan with BIRD as the route server?
I'm interested to hear your experiences. Specifically router vendor interop.
Note that there is an Euro-IX working group targetting this issue:
https://github.com/euro-ix/rfc8950-ixp
-- Elen sila lumenn' omentielvo
Ondrej 'Santiago' Zajicek (email: santiago@crfreenet.org) "To err is human -- to blame it on a computer is even more so."
participants (7)
-
Chris Malayter -
Darren O'Connor -
Douglas Fischer -
Jay Hanke -
Nico Schottelius -
Ondrej Zajicek -
Tore Anderson