Can't connect to cogentco.com

[annodomini]

I can't seem to get packets to cogentco.com via IPv6. It appears to be reachable according to http://www.berkom.blazing.de/tools/ping.cgi and http://www.ipv6-test.com/validate.php, and I can get to other IPv6 hosts via my tunnel, so this appears to be a localized routing problem.

Code Select Expand

$ traceroute6 cogentco.com
traceroute6 to cogentco.com (2001:550:1::cc01) from 2001:470:1f07:189:21c:b3ff:febc:c615, 64 hops max, 12 byte packets
1  2001:470:1f07:189:9284:dff:fed3:15d4  2.044 ms  3.095 ms  0.677 ms
2  annodomini-1.tunnel.tserv4.nyc4.ipv6.he.net  78.759 ms  78.918 ms  78.096 ms
3  gige-g3-8.core1.nyc4.he.net  78.151 ms  77.643 ms  78.305 ms
4  * * *
5  * * *
6  * * *
7  * * *
8  * * *
^C


$ ping6 cogentco.com
PING6(56=40+8+8 bytes) 2001:470:1f07:189:21c:b3ff:febc:c615 --> 2001:550:1::cc01
Request timeout for icmp_seq=0
Request timeout for icmp_seq=1
Request timeout for icmp_seq=2
Request timeout for icmp_seq=3
Request timeout for icmp_seq=4
Request timeout for icmp_seq=5
Request timeout for icmp_seq=6
^C
--- cogentco.com ping6 statistics ---
8 packets transmitted, 0 packets received, 100.0% packet loss


[cconn]

because Cogent and HE are kicking sand at each other and not peering, or at least, Cogent is, HE says they have offerred to fix this.  I don't know the whole story.

I am fortunate enough (i guess?) to have transit with AS174, so I peer IPv6 with them.  But they have almost 1000 less v6 prefixes in their "full view" than I get from HE.

[donbushway]

ping gives me a timeout. However the website does com up.

[cconn]

ping gives me a timeout. However the website does com up.

are you sure you are not defaulting back to IPv4???

[donbushway]

yes it is ipv6.
Firefox reports address 2001:550:1::cc01
cogento.com brings up => http://cogentco.com/us/

[cconn]

yes it is ipv6.
Firefox reports address 2001:550:1::cc01
cogento.com brings up => http://cogentco.com/us/

hello,

Not sure what magic is happening but the HE looking glass doesn't have a route for that IP or prefix.

They are announcing it as a /32 according to my router;

v6#show ipv6 route 2001:550:1::cc01
Routing entry for 2001:550::/32
  Known via "bgp xxxxx", distance 20, metric 122021, type external
  Route count is 1/1, share count 0
  Routing paths:

Interesting you can get to it.   If you manually telnet to 2001:550:1::cc01 port 80, the socket opens?
 

[donbushway]

Yes the connection opens takes a while like with the browser but it does open. Trace route does show stars after HE.

[snarked]

Short story:  Cogent is not IPv6 competent.  Find another transit provider if you can.

[cconn]

Short story:  Cogent is not IPv6 competent.  Find another transit provider if you can.

I wouldn't go that far.....but Cogent does historically have a somewhat agressive, if not to say menacing, habit of de-peering or not peering with other backbones.  But they do have competence.  HE has the opposite mentality, I suspect to attract significant IPv6 traffic to their network to better position themselves as a heavyweight.

[snarked]

I would go that far and HAVE gone that far.

In a recent IPv6 traceroute earlier this week by someone I know who peers with cogent, the IPv6 addresses returned by cogent networking routers included addresses in the ::/80 range (IPv4 compatible IPv6 addresses).  That should not happen.

...
6  te0-0-0-5.mpd21.ord01.atlas.cogentco.com (::ffff:154.54.27.61)  63.292 ms  63.354 ms  63.423 ms
7  te0-0-0-0.mpd22.ord01.atlas.cogentco.com (::ffff:154.54.31.150) 63.590 ms  63.619 ms  63.773 ms
8  te0-2-0-0.ccr21.ord01.atlas.cogentco.com (::ffff:154.54.25.177) 63.274 ms te0-4-0-0.mpd22.mci01.atlas.cogentco.com (::ffff:154.54.30.178) 63.473 ms te0-3-0-1.mpd22.mci01.atlas.cogentco.com (::ffff:154.54.7.165) 63.208 ms
9  te0-1-0-0.ccr22.mci01.atlas.cogentco.com (::ffff:154.54.30.153) 63.417 ms te0-1-0-3.mpd22.sfo01.atlas.cogentco.com (::ffff:154.54.24.105) 63.363 ms te0-3-0-3.mpd22.sfo01.atlas.cogentco.com (::ffff:154.54.7.222) 63.352 ms
10   (2001:550:2:3::15)  62.775 ms te0-1-0-cr22.sfo01.atlas.cogentco.com (::ffff:154.54.41.77)  63.291 ms 63.486 ms
...

If they were competent, then explain this crap.  Compatible addresses are not native addresses.

[cconn]

I would go that far and HAVE gone that far.

In a recent IPv6 traceroute earlier this week by someone I know who peers with cogent, the IPv6 addresses returned by cogent networking routers included addresses in the ::/80 range (IPv4 compatible IPv6 addresses).  That should not happen.

...
6  te0-0-0-5.mpd21.ord01.atlas.cogentco.com (::ffff:154.54.27.61)  63.292 ms  63.354 ms  63.423 ms
7  te0-0-0-0.mpd22.ord01.atlas.cogentco.com (::ffff:154.54.31.150) 63.590 ms  63.619 ms  63.773 ms
8  te0-2-0-0.ccr21.ord01.atlas.cogentco.com (::ffff:154.54.25.177) 63.274 ms te0-4-0-0.mpd22.mci01.atlas.cogentco.com (::ffff:154.54.30.178) 63.473 ms te0-3-0-1.mpd22.mci01.atlas.cogentco.com (::ffff:154.54.7.165) 63.208 ms
9  te0-1-0-0.ccr22.mci01.atlas.cogentco.com (::ffff:154.54.30.153) 63.417 ms te0-1-0-3.mpd22.sfo01.atlas.cogentco.com (::ffff:154.54.24.105) 63.363 ms te0-3-0-3.mpd22.sfo01.atlas.cogentco.com (::ffff:154.54.7.222) 63.352 ms
10   (2001:550:2:3::15)  62.775 ms te0-1-0-cr22.sfo01.atlas.cogentco.com (::ffff:154.54.41.77)  63.291 ms 63.486 ms
...

If they were competent, then explain this crap.  Compatible addresses are not native addresses.

I see your point;

Tracing route to ipv6.cogentco.com [2001:550:1::cc01]
over a maximum of 30 hops:

  1    <1 ms    <1 ms    <1 ms  2605:xxxx:xxxx:xxxx::1
  2     2 ms     2 ms     2 ms  2605:xxxx:xxxx:xxxx::1
  3     3 ms     2 ms     2 ms  2001:550:x:xx::x
  4    18 ms    18 ms    18 ms  ::ffff:154.54.0.37
  5    18 ms    18 ms    18 ms  ::ffff:154.54.7.57
  6    18 ms    18 ms    18 ms  ::ffff:154.54.5.253
  7    18 ms    18 ms     *     2001:550:1::1
  8    17 ms    17 ms    35 ms  2001:550:1::cc01

Trace complete.

the xxx obfuscation is mine.

I have them for now since in Canada IPv6 connectivity doesn't grow on trees.  HE gave me a free peering opportunity FFS, which I took and appreciate their generosity in offerring it.  The day we roll out with customers however I should have Peer1 and hopefully Tiscali up as well.  

Other than fugly traceroutes, and the fact they refuse to peer with HE, anything else?  ;D  Incidentally the topic of this thread was reachability to cogentco.com.

[cconn]

just as a followup to this, I have read that IPv4 mapped addresses in traceroutes are caused by MPLS routers in the path that do not have IPv6 native stacks.

AT&T would share the Cogent "incompetence", likely for the same reasons;

traceroute6 www.ietf.org
traceroute to www.ietf.org (2001:1890:123a::1:1e), 30 hops max, 80 byte packets
1  cconnv6.cconn.info (2605:2a00:ffff:fffd::1)  3.812 ms  6.499 ms  7.446 ms
2  v6.hautevitesse.net (2605:2a00:ffff:ffff::1)  4.545 ms  4.667 ms  4.783 ms
3  2001:550:2:16::9 (2001:550:2:16::9)  4.072 ms  4.165 ms  4.285 ms
4  2001:1890:1fff:105:192:205:36:77 (2001:1890:1fff:105:192:205:36:77)  11.845 ms 2001:550:3::2ce (2001:550:3::2ce)  57.363 ms 2001:1890:1fff:105:192:205:36:77 (2001:1890:1fff:105:192:205:36:77)  12.125 ms
5  n54ny21crs.ipv6.att.net (2001:1890:ff:ffff:12:122:80:226)  86.448 ms  86.486 ms  84.474 ms
6  cgcil22crs.ipv6.att.net (2001:1890:ff:ffff:12:122:1:2)  84.736 ms  84.832 ms  84.868 ms
7  cr1.cgcil.ip.att.net (::ffff:12.122.2.53)  86.010 ms  86.177 ms  86.343 ms
8  sffca21crs.ipv6.att.net (2001:1890:ff:ffff:12:122:4:121)  83.847 ms  84.078 ms  83.707 ms
9  cr81.sj2ca.ip.att.net (::ffff:12.122.1.118)  108.979 ms  108.959 ms  107.975 ms
10  sj2ca401me3.ipv6.att.net (2001:1890:ff:ffff:12:122:126:238)  81.250 ms  81.250 ms  81.379 ms
11  2001:1890:c00:3a00::11fb:8591 (2001:1890:c00:3a00::11fb:8591)  82.225 ms  81.968 ms  82.298 ms

[kasperd]

just as a followup to this, I have read that IPv4 mapped addresses in traceroutes are caused by MPLS routers in the path that do not have IPv6 native stacks.

That makes sense. I am trying to make intermediate IPv4 hops visible in my 6in4 tunnelling software, and they could show up as ::ffff:198.51.100.30 in an IPv6 traceroute. I couldn't figure out if ::/96 or ::ffff:0:0/96 is the best choice for such addresses. But if MPLS routers always use ::ffff:0:0/96 for that purpose, I should probably stick with that as well.

[matthewh]

Are there any other providers that offer an IPv6 BGP connection over a tunnel for free in addition to HE.net?

[kriteknetworks]

This thread is about cogent connectivity. please start a new thread, or google.