[Dnsmasq-discuss] DHCP errors with vlans and multiple subnets

[David Joslin]

Thanks, Simon.

The access points are all on vlan 9 (the management vlan). They serve
wireless clients on vlans 3, 4 and 5 but they don't have interfaces on
these vlans.

What's really baffling me is that nothing has changed with the network
configuration. I've been through the configuration of the router (with it's
dnsmasq dhcp server) and the switches on the network over and over again
and I can't find anything that's changed.

Each subnet is defined on the router (with dhcp enabled for each subnet)
and each vlan is associated with a particular subnet. The way it used to
work was this: I simply reserved the IP addresses for the access points on
the router (and I can see that these reservations have gone into dnsmasq's
dhcp-hosts file) and the access points were given these addresses (on
vlan9) when they asked for them - simple! Now, when they request addresses,
they are being offered addresses from every vlan apart from vlan 9! And
when they request the offered address, dnsmasq pumps out the warning
messages that the requested address conflicts with the address in the hosts
file. I even disabled the dhcp server on all the subnets apart from that
assigned to vlan9 to see if that would force the dhcp server to assign the
correct 10.10.99... address but when I did this dnsmasq logged the
following messages over and over again:
no address range available for DHCP request via br0
no address range available for DHCP request via br1
no address range available for DHCP request via br2

It wouldn't offer a vlan9 address.

There appeared to be a simple solution to this which was to assign static
IPs to all vlan9 devices (switches, access points). But when I did, DHCP
requests from the access points still appeared at the router! This would
seem to be a fairly major bug (in the Ubiquti UniFi Wi-Fi system) but I'm
wondering if this is in some way related to this problem (I can't think
how, though).

The access points have only one network interface and don't seem to offer
any way to configure client-IDs so I still can't see a solution (apart from
reverting everything to factory settings and building it all from scratch
again - not something I want to do).

If you've got any advice I'd be grateful.

Cheers

David


On 4 March 2014 17:59, Simon Kelley <simon at thekelleys.org.uk> wrote:

> So, the same machine, with the same MAC address, seems to be talking to
> the dnsmasq DHCP server from (at least) three different subnets
> more-or-less simultaneously. This is not good, as the DHCP protocol (for
> IPv4, at least) assumes each interface will get _one_ address. Dnsmasq
> is chasing its tail, giving the machine one address, then abandoning
> that and giving it another, and so on.
>
> The APs have interfaces on multiple VLANS? If so you need to do one of
> two things
>
> 1) Get them to use different MAC addresses on each distinct VLAN,
> 2) Get them to use DHCP client-IDs and ensure that _those_ are distinct.
>
> As a unique identifier, client-ids override MAC addresses, so you should
> be OK leaving the interfaces with a single MAC address if you take the
> client-id route.
>
> Most DHCP clients have a way to configure which client-id they should use.
>
>
>
> Cheers,
>
> Simon.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.thekelleys.org.uk/pipermail/dnsmasq-discuss/attachments/20140304/15358053/attachment.html>