DHCP by definition works by the client using the broadcast address to find all responding DHCP servers. It then chooses one of them (based on an algorithm, and AFAIK this is mostly to do with response times) and asks for an address. This address is then bound to the interface that is being configured.
Again, by definition, broadcasts are sent on all network interfaces. So, unless you are willing to write your own DHCP client (the sources are freely available), there is not much you can do about that.
However, you could use a different port. Read the man page for dhclient for details. This would then also require a DHCP server that listens on that port. The standard ISC DHCP server supports this with the -p option. Again, read the man page for dhcpd.
I talked to a friend of mine who has dealt with this type of thing before and here is what we did to get the environment to the end-state.
First, I never had to use dnsmasq -y, this worked as soon as I restarted the dnsmasq service and the test VM
The first thing we did was fix the dnsmasq server. In /etc/network/interfaces, you specifie the ip addr you want for that nic, use a netmask of 255.255.255.0, and define the network for that nic. It will look like this:
auto eth1
iface eth1 inet static
address 10.2.2.10
netmask 255.255.255.0
network 10.2.2.0
broadcast 10.2.2.255
The only change is to eth0, in which these three lines were already there. Not sure if they are needed (hopefully someone else can clarify that) but this is what I had added:
up route add default gw 10.2.1.1
dns-search my.lab
dns-nameservers <Corp-DNS-server>
I also removed the two post-up route lines, since they were not needed.
Finally, I needed to fix dnsmasq.conf; here I added the lo interface and commented out the no-dhcp-interface:
interface=lo
interface=eth0
interface=eth1
interface=eth2
#no-dhcp-interface=eth0
#no-dhcp-interface=eth1
#no-dhcp-interface=eth2
That fixed the server. The next thing to achieve the end-state is to configure your router to allow inter-vlan communication. I had already done that prior to this question, but here is an article to do that configuration: http://www.cisco.com/c/en/us/support/docs/lan-switching/inter-vlan-routing/41860-howto-L3-intervlanrouting.html
Lastly, I found that the test vm was not adding in the nameserver to its resolv.conf, so I added a file called tail to /etc/resolvconf/resolv.conf.d/ and simply added the dns server's ip for that subnet:
nameserver 10.1.1.10
Once I finished that, I was able to run apt-get update and ping my dns names. I also then changed the test vm's interface file so it would get a dhcp addr and once I rebooted, it was able to recieve dhcp addrs.
I hope this helps everyone! Feel free to leave any comments if you wish for me to clarify.
Best Answer
You can do everything you want with one instance. This is an example based on a configuration I have used.
This is covered in the man page.