Re: [v6ops] What problem are we trying to solve? (Re: A broken promise - "You said PD Prefix Valid Lifetime is going to be X" (Re: SLAAC renum: Problem Statement & Operational workarounds))

[Richard Patterson <[e-mail]>]

On Wed, 13 Nov 2019 at 05:08, Fred Baker <[e-mail]> wrote:

> Change "would" to "could", and you're probably close. In general, I would
> expect ISPs to not want to randomly change their routing, and so it would
> not be deterministic - if the customer is no longer using a prefix (e.g.,
> the CPE crashed and routing has changed anyway), when the CPE comes back up
> they might give it a different prefix. However, I'd be surprised if they
> did it "just because".
>

Depends on the vendor.  For example, one of our BNGs (that's operating as a
DHCPv6 server) will keep a lease in a "held" state for a configurable
duration after the Valid Lifetime expires.  We set that value to 7 days,
allowing a CPE to come back and get the same PD within a week.

A different BNG vendor of ours however, doesn't have this configurable hold
time concept; once the Valid LIfetime expires, the MAC+DUID binding to the
PD is cleared.
They do have setting to retain the MAC+DUID->PD lease binding indefinitely,
however they didn't satisfactorily answer my question about cleanup if
these bindings are exhausted (CPEs that use dynamic DUID values perhaps?
Even though even the DUID-LLT method is *supposed* to be stable).