Some debian IPv6 addresses nullrouted by TATA (prefix leak?)
I'm unable to reach some debian.org addresses in IPv6 today, like
backports.debian.org.
This is what I see:
skull@logger:~$ traceroute6 -n 2607:f8f0:610:4000:211:25ff:fec4:59ae
traceroute to 2607:f8f0:610:4000:211:25ff:fec4:59ae
(2607:f8f0:610:4000:211:25ff:fec4:59ae), 30 hops max, 40 byte packets
1 2a02:9a8:1::1 0.000 ms 4.000 ms 4.000 ms
2 2a02:9a8:1:64::32 8.000 ms 8.000 ms 8.000 ms
3 2a02:9a8:0:7f::1 8.000 ms 8.000 ms 8.000 ms
4 2001:450:2002:20e::1 8.000 ms 8.000 ms 8.000 ms
5 2001:450:2008:101::2e 27.999 ms 27.999 ms 27.999 ms
6 2001:450:2008:101::2e 27.999 ms 23.999 ms 23.999 ms
7 2a01:3e0:ff80:100::12 111.999 ms 115.999 ms 115.999 ms
8 2001:5a0:300:100::e 135.999 ms 135.999 ms 135.999 ms
9 2001:5a0:3100:100::1 187.998 ms 187.998 ms 187.998 ms
10 2001:5a0:3100:100::2 187.998 ms 191.999 ms 191.999 ms
11 2001:5a0:3100:100::1 243.999 ms 243.999 ms 243.999 ms
12 2001:5a0:3100:100::2 243.999 ms 239.998 ms 239.998 ms
13 2001:5a0:3100:100::1 291.997 ms 291.997 ms 291.997 ms
14 2001:5a0:3100:100::2 291.997 ms 291.998 ms 291.998 ms
15 2001:5a0:3100:100::1 343.997 ms 343.997 ms 343.997 ms
16 2001:5a0:3100:100::2 343.997 ms 339.997 ms 343.997 ms
17 2001:5a0:3100:100::1 391.996 ms 391.996 ms 455.996 ms
18 2001:5a0:3100:100::2 447.996 ms 447.997 ms 427.997 ms
[...]
The last couple of hops are on TATA network.
Looking at the prefix for 2607:F8F0:600::/40 as it reaches my network:
mi2#sh bgp ipv6 unic 2607:F8F0:600::/40
BGP routing table entry for 2607:F8F0:600::/40, version 614297
Paths: (1 available, best #1, table Default)
Advertised to update-groups:
3
3549 6453 271, (aggregated by 271 207.23.253.1)
2001:450:2002:20E::1 (metric 20) from 2A02:9A8::FFFA:1 (147.123.0.65)
Origin IGP, metric 10000, localpref 100, valid, internal, best
Community: 3549:2623 3549:31250 6734:201
As you can see, GlobalCrossing is taking the path through TATA to reach
that BCnet prefix, maybe because that's the right path but maybe because
of TATA leaking out the prefix (not the first time...).
Whatever the reason, that prefix ends up in a loop.
I'm trying to open a case with GlobalCrossing (if it's a leak, they
should filter this prefix out on their side anyway), but you may be
interested in opening a case at BCnet too.
Hope this helps.
Thank you for all the great work.
Bye!
--
# Emanuele Balla # #
# System & Network Engineer # #
# Spin s.r.l. - AS6734 # Phone: +39 040 9869090 #
# Trieste # Email: balla@staff.spin.it #
Reply to: