RE: Two MAC addresses for a same IP Primary address
"Don Provan" <[email protected]>
| Newsgroups | gmane.ietf.vrrp |
|---|---|
| Message-ID | <[email protected]> |
> Here, what you call VR IP address is defined by *IP Address > Owner* into the RFC, isn't it ? There is a VR IP address. The VR IP address has an owner, but it is the VR master, whether currently the owner or not, that uses the VR MAC address. > > ARP should always advertise that mapping and no other. > > But this has no relevance to the selection of source > > MAC address when transmitting a packet. > > I agree that Address Resolution (ARP) of *IP Address Owner* must only > adverise the *Virtual Router MAC Address*. The *master* is the one that ARPs the VR IP address, regardless of whether it is the VR IP address owner. In any case, we agree with the important point: ARP always maps VR IP address to VR MAC address. > Please, here, when you write "by sending the advertisement > packets with > VR MAC address", please, what is the IPv4 source address ? > 1-> the *IP Address Owner* > 2-> xor the *Primary IP Address* The source is the primary IP address of the interface through which the advertisement is sent. For the VR IP address owner, this could be the VR IP address itself, but for any other router supporting this VR, it has to be an IP address unrelated to the VR. (In my implementation, even the VR IP address owner uses an unrelated IP address as it's primary IP address.) > Our issue is related to the fact that 2 IP addresses are defined into > the RFC *IP Address Owner* and *Primary IP Address*. It is > clear how the *IP Address Owner* must be used with the VR MAC address on > many gateways running VRRP on a same link, but it is unclear for the > *Primary IP Address*. The primary IP address uniquely identifies each router participating in the VR. It is the IP address assigned to the router because it is an IP node on that network; it was nothing to do with the fact that the router is now assigned to support VRRP. > I would be interested in better understanding how is supposed > to be used > the *Primary IP Address* on 3 gateways running VRRP on a same link: > -> There are 3 *Primary IP addresses* at least, aren't there ? Yes. Each router has its own, unique primary IP address on the network VRRP is running on. (The VR IP address owner is allowed but not required to use the VR IP address as its primary IP address.) > -> Are these 3 *Primary IP addresses* reachable at anytime even if > the gateway is in the backup state ? The primary IP address are assigned to the physical router. Its reachability is unrelated to the VRRP protocol in any way. One would expect the address to be reachable whenever that interface is up just as one would expect any other node's IP addresses to be reachable when the node is active. As it turns out, VRRP doesn't even require that much: VRRP never sends a packet *to* the Primary IP address (unless I'm forgetting something....), so for VRRP to work, the Primary IP address need not be actually reachable, although I don't think the protocol spec itself mentions this unimportant detail. -don _______________________________________________ vrrp mailing list [email protected] https://www1.ietf.org/mailman/listinfo/vrrp