How is DHCP6 suppose to work with dynamic prefixes?

Question

Let's consider a simple tree topology. There is an ISP router, there is my router / DHCP6 server (which will update a DNS zone) and some other downstream routers. According to RIPE690 it is advised that ISPs provide static prefixes to customers, but let's suppose that doesn't happen here. In this scenario, the ISP router provides a /48 prefix which is subject to change.

The way I see it, the ISP router is supposed to hand a /48 prefix to my router/DHCP6 server, which in turn can take it, break it into smaller (let's say /56) subnets and hand them to the routers downstream which can assign IPv6 addresses to the hosts in those respective subnets.

There are a few things I don't quite get. How can kea-dhcp6-server fulfill such a task, since the configuration file seems to need a pd-pool definition? This suggests that the prefix has to be static, and if it changes I need to manually update the configuration file. Another thing I cannot figure out is what address each of my router's interfaces is going to get, and how? And within which subnet are those addresses going to be?

I suppose I am making quite a few false assumptions here, maybe the biggest one being that the prefix provided by the ISP is going to change often (or each time the power fails).

Answer 1

If your DHCPv6 server cannot learn a dynamic prefix, this is a failing of the software, not the protocol.

When the delegated prefix changes, it gets removed. With SLAAC, that's an instant process. With DHCP that can be complicated as the server can't always send an async message to it's clients. But a smart server won't hand out leases longer than it's own lease. (i.e. if the lease for prefix A expires in 2 hours, don't tell others it's good for a day.)

In any case, unless things on your end change (MAC, DUID, etc.), your prefix should be unchanging. It's not "static", but it's close enough. Otherwise, that's a conversation to have with your ISP.