Expanding an Experiment with Modern IP in a Home/Lab Network
This is the second article in a series on getting started playing with some IPv6 experiments in your Home/Lab Network. The first article had us pinging between two machines over a direct wire. Now we're ready to experiment a little bit more.
Find an Ethernet switch that you can borrow for this experiment. It's
perfectly fine to be only 10/T or 100/T and not 1000T or higher. In fact,
for some purposes the older the better. Hook up your two test machines
to this switch, in the same fd01:2:3:4:5:6:7::/112 subnet from last time.
(Those with deeper IPv6 knowledge -- calm down about the /112 mask, the
4-think on this subnet will go away soon. This is an experiment for
those coming from legacy IP.)
This is best (for now) to be an isolated network, again just an experiment. We should be able to ping and SSH from A->B and B->A. It should go smoothly and without hiccup.
The first network switch I tried this with, it worked fine. The next one, it didn't. Unfortunately, I sank several hours into diagnosing it. For now, if a network switch seems to not work (in "dumb" modes) with IPv6, put a label on it and plan to replace it. In the end, this is almost certainly what you'll wind up doing. In my (and most cases), some older, dumb network switches have problems with the way that IPv6 crafts multicast packets and broadcast ethernet frames for its equivalent of ARP.
IPv6 uses a protocol called NDP to do the equivalent of ARP in IPv4.
The mechanical details are irrelevant, except that because its slightly
different, some older gear might think it's some kind of network
gremlins and try to suppress it. I discovered that two machines on this
faulty switch couldn't ping each other, but if they pinged each other
via a direct connection first and then the switch was patched in...
the cached information from the NDP (ARP equivalent) would be available
for a few minutes and the ping would go thru. If you have a switch
which demonstrates this behaviour, you'll almost certainly have to
retire it once you start to embrace Modern IP.
You can expand this test to perhaps include a few other machines or a few daisy chained switches. It's perfectly acceptable. This phase of your experiments is just to help you gain awareness of what equipment (if any) in your network needs replacement. You'll have to retire that equipment eventually, as IPv6 becomes more widespread and more mandatory.
Once you have a few machines that you know are capable of talking Modern IP, a few switches that can as well, and you've setup an IPv6 network... you're ready to move onto making your first IPv6 network go "live", even if internal only.