I also noticed that IPv6 disappeared from the network and reported it to the
NOC. I think they figured out the problem at least in one of the APs or
whatever it was. I've requested to know the reason but got no information at
the time being.
Jordi,
At the heart of this problem was that we were using the hotel's existing
switch infrastructure, since they already had ports all over the place,
and it also allowed us to use their existing APs as well as our own.
Unbeknownst to us, their switches were configured with the "security"
options, 'switchport block unicast' and 'switchport block multicast'.
The first meant that if the switch forgot a MAC address before the end
device's ARP table did (e.g., because there are lots of MAC addresses
flying around the network at a big meeting), that connectivity between
the two systems would be blackholed. This caused a great deal of trouble
with our monitoring station and also with the printers. The second meant
that the IPv6 ND / RA packets, sent to arbitrary multicast addresses,
were not forwarded since the switch didn't think that the multicast
should go to these ports.
After asking the hotel's provider to remove these restrictions, IPv6
worked again. There were still some isolated incidents which we were
unable to completely debug but could be explained by some lingering
'switchport block' commands.
This was the first time (to my knowledge) that we used a venue's existing
infrastructure so heavily. It certainly taught us a few things.
Bill
_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
https://www1.ietf.org/mailman/listinfo/ietf