Re: [v6ops] draft-ietf-v6ops-slaac-renum and draft-ietf-v6ops-cpe-slaac-renum

[Richard Patterson <[e-mail]>]

On Wed, 20 May 2020 at 09:56, Philip Homburg <[e-mail]>
wrote:

>
> I think the need to relax L-15 comes from the fact that ISPs have an
> incentive to use very short lease times.
>
> I don't know the details, but what seems to happen is that when a
> customer swaps one CPE for anothor, the old CPE doesn't release the
> lease (i.e., people just poweroff or disconnect the old CPE).
>
> Then when the new CPE asks for a lease and the customer has a fixed
> prefix, the radius daemon finds that the lease is already in use by
> the old CPE and doesn't give any lease to the new CPE.
>
> The hacky quick fix for this is to set lifetimes in the lease to only
> a few hours at most and ask people to wait.
>

So this is indeed a concern for an Operator running IPoE services, but is a
problem to be tackled on the BNG rather than the CPE router.
One mitigation is as you suggest, the Operator uses short DHCP lease times
so that the session times out gracefully within a reasonable time, but this
is even more reason for L-15 to exist as a MUST, so the CPE doesn't
overpromise the prefix duration on the LAN.

There are also additional approaches depending on the BNG vendor, that
either reactively clear subscriber sessions when they detect a
duplicate/conflict, or more proactively check the aliveness of the CPE
Router periodically with ARP/ND.