Has anyone else noticed that the ATT BGW320 won't delegate an IPv6 prefix until there's a link state change on its LAN ports? ATT reboots the BGW320 at night every few weeks. That takes my IPv6 connections down until I manually intervene. Unplugging and replugging my home router's WAN interface from the BGW320 is enough to restore prefix announcements from the BGW320, which restores IPv6 functionality for my network.
The WAN interface of my home router (Ubiquiti Dream Machine, "UDM") is connected to a LAN port of the BGW320 in DHCPS-Fixed passthrough mode. The UDM is the only device connected to the BGW320 and the BGW's WiFi is disabled. The BGW320 is currently running 3.18.5, but the issue has been going on for a while.
If the UDM is rebooted, or its WAN port unplugged then replugged, while the BGW320 is up, the UDM gets a delegated /64 prefix from the BGW320. The UDM uses that prefix to assign IPv6 addresses to clients on its LAN side (combination of SLAAC and DHCPv6). IPv6 "works" for my home machines at that point. However, if the BGW320 is rebooted while the UDM is up, the BGW320 does not resume advertising prefixes until the UDM is rebooted or its WAN cable unplugged+replugged. IPv6 is down for my home machines until then.**
I've attached packet captures from the UDM's WAN interface that demonstrate the problem. I started in the state where the UDM had a prefix from the BGW320 (i.e., IPv6 working). Then I rebooted the BGW. The first pcap is taken after the BGW comes back up: the BGW is advertising no available prefixes. After ~5 minutes, I unplugged the UDM's WAN interface, waited 30s and then plugged it back in. The second pcap is taken immediately afterwards and shows the BGW has resumed advertising a prefix (i.e., at which time I again have in-home v6).
Thanks in advance.
** IPv6 "going down" is nuanced. Depending on the client and when the BGW reboots, loss of IPv6 can take up to an hour, which is the lifetime of prefixes delegated by the BGW.
The WAN interface of my home router (Ubiquiti Dream Machine, "UDM") is connected to a LAN port of the BGW320 in DHCPS-Fixed passthrough mode. The UDM is the only device connected to the BGW320 and the BGW's WiFi is disabled. The BGW320 is currently running 3.18.5, but the issue has been going on for a while.
If the UDM is rebooted, or its WAN port unplugged then replugged, while the BGW320 is up, the UDM gets a delegated /64 prefix from the BGW320. The UDM uses that prefix to assign IPv6 addresses to clients on its LAN side (combination of SLAAC and DHCPv6). IPv6 "works" for my home machines at that point. However, if the BGW320 is rebooted while the UDM is up, the BGW320 does not resume advertising prefixes until the UDM is rebooted or its WAN cable unplugged+replugged. IPv6 is down for my home machines until then.**
I've attached packet captures from the UDM's WAN interface that demonstrate the problem. I started in the state where the UDM had a prefix from the BGW320 (i.e., IPv6 working). Then I rebooted the BGW. The first pcap is taken after the BGW comes back up: the BGW is advertising no available prefixes. After ~5 minutes, I unplugged the UDM's WAN interface, waited 30s and then plugged it back in. The second pcap is taken immediately afterwards and shows the BGW has resumed advertising a prefix (i.e., at which time I again have in-home v6).
Thanks in advance.
** IPv6 "going down" is nuanced. Depending on the client and when the BGW reboots, loss of IPv6 can take up to an hour, which is the lifetime of prefixes delegated by the BGW.