Pool independent GI Address

Simon Hobson dhcp1 at thehobsons.co.uk
Mon Dec 15 12:25:22 UTC 2014

Stephan Sanders <stephan at sanders.koeln> wrote:

> Let me explain the solution.
> Each domain is located on a switch/router with a dhcp-relay.

Domain is not a sensible description of an entity in this context - as Niall has pointed out, we have no idea what you mean by it. Please stick to standard terminology and things might go a bit better !

We tend to use "network", "physical network", "LAN" as interchangeable terms - VLANs complicate things somewhat. Technically it's the broadcast domain that counts. One physical ethernet network (without VLANs), or one VLAN on a VLAN equipped network. What we're interested in is, "can a device get a packet to the appropriate interface of a relay agent without going through a router ?" - noting that there is no requirement for a relay agent to be in a router or have more than one interface.

> on each domain are two subnets configures (primary and secondary addresses)
> primary    GI and GW address:
> secondary           GW address:

Yep, that's perfectly OK. Are these on the same link/VLAN ?
Assuming they are, eg for Cisco kit you have an interface with :
  ip address
  ip address secondary
then you have a shared network. If you configure the server correctly, as I gave your a fragment for, then things will just work "automagically".

> The dhcp relay allows to use more than one dhcp server but only to define one GI address.
> So for the first server everything is fine. The GI address is within the pool.

There you go with "strange" terminology again. I think you mean the GI-Addr is within the **subnet** !

> But the Backup Server
> will receive discovers from the GI and has to offer addresses from Selection of which /29 pool
> should be done by option82.

No, Option82 is irrelevant. Just configure the shared subnet and appropriate subnets/pools as I suggested.

> option82 will be filled by the dhcp relay with a string for the remote-id.

Which will be the same regardless of which subnet the client needs to get an address in !

> Thats the way i want run my configuration.

OK, so you've asked how to do things the hard way, you've been told how to make it really easy (plus reliable, and understandable for the poor sucker you takes over after you leave/get knocked down by the proverbial bus). If you still want to do it the hard way then realistically you'll be on your own !
I still can't see any logical reason that both server sites can't route packets to the subnet. The fact that you are suggesting it's impossible suggests that someone involved in this network doesn't understand some fairly basic IP addressing and routing stuff.

... Lightbulb moment ... There aren't overlapping subnets somewhere are there - ie non-unique IP addresses across the network ?

More information about the dhcp-users mailing list