[Dnsmasq-discuss] DHCP and DNS regestration

[support | IT visibility]

On 3/24/25 09:47, support | IT visibility via Dnsmasq-discuss wrote:
> Hi *,
> 
> I’m using dnsmasq with a few vlan interfaces and each interface is tagged.
> 
> And running tests with pihole as well as native dnsmasq.
> 
> An example on how this is setup in both cases (.i.e. pihole as well as 
> native dnsmsaq):
> 
> =====
> 
> # Wired - vlan 210
> 
> interface=eth0.210
> 
> domain=wired.lan,192.168.210.0/24,local
> 
> dhcp-range=set:vlan-210,192.168.210.11,192.168.210.99,168h
> 
> dhcp-option=tag:vlan-210,option:router,192.168.210.241
> 
> dhcp-option=tag:vlan-210,option:dns-server,192.168.210.235
> 
> dhcp-option=tag:vlan-210,option:ntp-server,192.168.210.235
> 
> dhcp-option=tag:vlan-210,option:domain-name,wired.lan
> 
> =====
> 
> The idea behind this is that devices with multiple vlan interfaces 
> have their IP’s registered with dhcp and dns.
> 
> The dhcp part seems to work – all interfaces have the correct IP details.
> 
> However, the log has numerous wrong-address errors like this (i.e. 
> pihole as well as native dnsmasq):
> 
> =====
> 
> Mar 24 09:02:03 dnsmasq-dhcp[49]: DHCPREQUEST(eth0.210) 
> 192.168.210.201
> bc:24:11:87:be:e5
> 
> Mar 24 09:02:03 dnsmasq-dhcp[49]: DHCPNAK(eth0.210) 192.168.210.201
> bc:24:11:87:be:e5 wrong address
> 
> Mar 24 09:02:03 dnsmasq-dhcp[49]: DHCPDISCOVER(eth0.210) 
> bc:24:11:87:be:e5
> 
> Mar 24 09:02:03 dnsmasq-dhcp[49]: DHCPOFFER(eth0.210) 192.168.210.201
> bc:24:11:87:be:e5
> 
> Mar 24 09:02:03 dnsmasq-dhcp[49]: DHCPREQUEST(eth0.210) 
> 192.168.210.201
> bc:24:11:87:be:e5
> 
> Mar 24 09:02:03 dnsmasq-dhcp[49]: DHCPACK(eth0.210) 192.168.210.201
> bc:24:11:87:be:e5 osiris
> 
> =====
> 
> *Question-1*: where are these wrong-address messages coming from?
> 
> If I look at the dhcp.leases (Pihole) or dnsmasq.leases (native 
> dnsmasq) there is always one entry like this:
> 
> 1743408124 bc:24:11:87:be:e5 192.168.230.201 osiris *
> 
> *Question-2*: why is there only an entry from the last vlan interface 
> of (in this example) osiris? And not one for every vlan interface?
> 
> *Question-3*: the same applies for dns regestrations where there is 
> only an entry for the last vlan interface?
> 
> I have tried to force this dns regestration for all fqdn’s for each 
> vlan using the following settings (i.e. pihole as well as native dnsmasq):
> 
> dhcp-ignore-clid
> 
> dhcp-fqdn
> 
> dhcp-client-update
> 
> expand-hosts
> 
> domain-needed
> 
> domain=lan
> 
> local=/lan/
> 
> Does the above ring any bells? Any suggestion?
> 
> With warm regards – Will
> 
> 

>Answer 1: Because a lease exists for the relevant MAC address with a different IP address to the address it's requesting.
>
>Answer 2: Because the device is using the same MAC address on all the VLANS it's talking through. The MAC address is supposed to be unique: a DHCP 
>server won't allow a MAC address to have more than one lease.
>
>Answer 3: Partly the same answer as 2, but even if you use unique MAC addresses (or client-ids) dnsmasq will only associate a DNS name with at most one
> DHCP lease. That's a design decision.

=====

Thank you for the detailed and to-the-point answers... which makes me wundering:
If the mac-address is unique and clid is set via the duid option, all interfaces/vlan-domains are registered in DNS.

In addition, the leases file shows all the registrations - including the clid for each interface.
This showed me that both are unique across all interfaces.

If I then remove the custom mac addresses (leaving the unique clid's in place), only the last one is registered in dns.
Which is not what I would expect based on what the manual says about the way things are expected to work with clid's.

This was tested with a Debian LXC-container where I replaced ifupdown and isc-dhcp-client with systemd-networkd.

What am I missing/overlooking here?


With warm regards - Will

=====