That is not always 100% the case. I've seen relay agents in ISP environments that never tell the client the location of the actual DHCP server. They masquerade as the DHCP server swapping their own address for the GIADDR in the response packet. Probably not proper RFC wise, but I have seen it.
----- Original Message -----
> From: "glenn satchell" <
[hidden email]>
> To: "Users of ISC DHCP" <
[hidden email]>
> Sent: Monday, April 29, 2019 2:14:02 AM
> Subject: Re: Separates VLANs with the same IP-Range
> Remember the DHCP relay only handles relaying broadcast traffic from the
> client. This is the DHCPDISCOVER and initial response packets. Once the
> client gets an IP address it talks directly to the DHCP server for
> checking, renewing and acknowledging, the relay is no longer involved.
> regards,
> -glenn
> On 2019-04-28 02:02, Simon Hobson wrote:
> >
[hidden email] wrote:
> >> So i remove the nat konstrukt and add a subnet for the second relay.
> > It's not entirely clear what your network topology is here. Are you
> > saying that there is NAT between the DHCP server and the clients ?
> > There must be no NAT between clients and server - the server needs to
> > be able to send unicast packets to the client and this cannot be done
> > if there is NAT in the path.
> > Also, you cannot have any overlap in IP addressing between networks -
> > normal packet routing doesn't handle this, and the ISC DHCP server
> > certainly doesn't.
> > Remember two two fundamental rules of IP are "addresses must be
> > globally unique"(1) and "any node should be able to address a packet
> > to any other node"(2). NAT breaks both of these.
> > 1 - In the context here, in the collection of networks your DHCP
> > server is to handle, all addresses must be unique.
_______________________________________________
dhcp-users mailing list
[hidden email]
https://lists.isc.org/mailman/listinfo/dhcp-users