Strange problems crossing 6to4.lon1.he.net

[jbwisemo]

Since last night I am having strange problems crossing 6to4.lon1.he.net (London Docklands) from a server hosted elsewhere in the docklands (on peer1's network):

1. IPv4 ICMP traceroute to 192.88.99.1 succeeds in 6 hops (2ms) with 10gigabitethernet1-1.core1.lon1.he.net as hop 5

2. IPv6 over 6to4 ICMP traceroute to 2002:c058:6301::1 (the IPv6 of 6to4 relays) succeeds in one hop.

3. IPv6 over 6to4 ICMP traceroute to 2001:470:0:161::2 (6to4.lon1.he.net?, broken rDNS) succeeds in one hop.

4. IPv6 over 6to4 ICMP traceroute to he.net or any other non-6to4 address fails, with no replies beyond a few IPv4 hops.

5. Outside IPv6 ping of the 6to4 address of the server on peer1 reaches the server, which sends back an ICMPv6 reply that is lost in transit.

(Above tests were done from 2002:ipv4::1 on a Linux server).

So what is wrong with 6to4.lon1.he.net?

Is there something wrong with its routing tables, is it doing packet filtering contrary to RFCs or is it just plain crashed?

[rmosher]

This should be fixed.  Go ahead and retest.

[jbwisemo]

Yes, everything going through that relay came back online between 17:29 and 17:47 UTC according to my logs, thanks.

This was quite a confusing failure mode, because the relay responded to probes from all sides I initially blamed another router in a different AS for dropping the traffic.

Are you running automated monitoring on your 6to4 relays to detect failures early?

I am thinking of stuff like sending ping packets through each relay in both directions every 5 minutes, with the pings terminating in remote parts of the HE network to detect loss of routing from the relay on either side?