Tomek Mrugalski tomasz at isc.org
Tue Sep 23 11:39:38 UTC 2014

On 23.09.2014 01:24, Agustín Bertamoni wrote:
> According to docsis scenario, we need to assign IA_NA and  IA_PD to
> erouters at the same time, but the NA (non-temporary addresses)
> candidate and the IAPREFIX not necessary must be in the same prefix, so
> the “pool” and “pd-pool” are not in the same subnet.
> As a result of the tests, this seems not to be supported and we see that
> first selects a class-based available subnet and PD that is within it.
> Is this analysis correct ?
Yes. The assignment process more or less looks like following:
1. parse incoming packet
2. assign packet to a class if appropriate. Right now it simply takes
content of the vendor-class identifier (DHCPv4) or vendor class
(DHCPv6). The class is specified on a per packet basis.
3. a subnet is selected for incoming packet. Normally it is done based
on relay IP address and/or interface-id inserted by relay, but you can
also allow/forbid specific client class from a given subnet.
4. Leases are assigned from the selected subnet. Each lease type (IA_NA
or IA_PD) is processed separately, but to assign both of them, you need
to have both pool and pd-pool defined.

So if you want to have both IA_NA and IA_PD assigned in the same
transaction, it is currently not supported. But are you completely sure
that's what you really want? One of those resources assigned (either
address or prefix) will be topologically incorrect.


More information about the kea-dev mailing list