Used WS2016 on VM to implement PXE boot server. It was simple matter of enhancing IPv4 server.
The IPv6 was much different than IPv4 but noticed following difference in setting up on WS2016:
On DHCP snap-in from servermanager, it does not seem to allow to create scope with prefix other than /64. (In contrast, i can certainly set the net address static IPv6 with any prefix).
The way to create scope pool of IPv6 address was weird. I had to do by defining the excluded addresss:
i.e. wanted pool range: 2000::100 to 2000::1000 then had to create exclusion range of 2000::100 to 2000::0ff and 2000:1001 to 2000:ffff:ffff:ffff:ffff with /64
On the server itself, I assigned static address of 2000::1 to its net interface. But during installation using IPV6 over PXE, I see it has another address assigned to it: 2000::something which it got it from its own DHCP pool.
The client had also IPv6 from pool. They were all non-sequential, (does it pick it randomly?)
DHCP will pick randomly, yes.
I remember having to muck around with the "M" and "O" flags to get things working...that and something about RA's... it's been a while...
thx 4 input, CCNA book briefly mentioned about refer to section x.y for why net interfaces can possibly get more than one IPv6 address, lost it where it mentioned now.
Also I think the PXE client was associated at least 3-4 different IPv6 during and after installation.
Multiple IPv6 addresses is also a thing. Makes licensing based on IP addresses a hellacious thing.
yeah, one thing came to mind is all the observation made here are working as defined. RFC for IPV6 may be?