Hello, Atlas helped me today first to see a hard-to-detect bug: F-root anycast algo works fine for v4 but fails miserably for v6, as instead of AMSterdam it sends all traffic to some remote island in the pacific. ;-) Seems to be the reverse true for M-root: ipv4 sends me off to nowhere land while v6 is great. A-root follows the F-suit. (C-root is not a good avertisement valie for Cogent, with v4 only address and considerable ongoing pkt loss.) Thanks! Peter
On Tue, Nov 13, 2012 at 12:57 PM, Peter Gervai <grinapo+ripeatlas@gmail.com> wrote:
Atlas helped me today first to see a hard-to-detect bug: F-root anycast algo works fine for v4 but fails miserably for v6, as instead of AMSterdam it sends all traffic to some remote island in the pacific. ;-)
Just to follow myself up, most of the private email was about questions easily answered by my humble probe No. 3338. :-) It's all public, as far as my setup skills shine. :-] Peter ps: reminds me to always attach the probe# to my email to back up my claims.
In a message written on Tue, Nov 13, 2012 at 12:57:22PM +0100, Peter Gervai wrote:
Atlas helped me today first to see a hard-to-detect bug: F-root anycast algo works fine for v4 but fails miserably for v6, as instead of AMSterdam it sends all traffic to some remote island in the pacific. ;-)
The good folks at ISC would be quite happy if you opened a ticket with them on this issue. Drop an e-mail to noc@isc.org to do so, including traceroutes and such would help a lot. As you noticed with both F and M, there's always a routing misconfiguration somewhere for someone. IPv6 is still a little less mature than IPv4 in that sense. Lots of good people are working on the problem. -- Leo Bicknell - bicknell@ufp.org - CCIE 3440 PGP keys at http://www.ufp.org/~bicknell/
On Tue, Nov 13, 2012 at 4:34 PM, Leo Bicknell <bicknell@ufp.org> wrote:
In a message written on Tue, Nov 13, 2012 at 12:57:22PM +0100, Peter Gervai wrote:
Atlas helped me today first to see a hard-to-detect bug: F-root anycast algo works fine for v4 but fails miserably for v6, as instead of AMSterdam it sends all traffic to some remote island in the pacific. ;-)
The good folks at ISC would be quite happy if you opened a ticket with them on this issue. Drop an e-mail to noc@isc.org to do so, including traceroutes and such would help a lot.
I did.
As you noticed with both F and M, there's always a routing misconfiguration somewhere for someone. IPv6 is still a little less mature than IPv4 in that sense. Lots of good people are working on the problem.
We have to watch ipv6 with much more attention since we really have to prevent people saying that it's "worse" than ipv4, especially now that ripe handled our last v4 block ever and no more. Later or possibly sooner the endusers get it, and I mean the END endusers, those gravediggers, spoon-binders and dishwashers. :-) And DNS is the first entry, so I'd prefer working and reachable roots. And Atlas is a great way to assure that. Peter
On Nov 13, 2012, at 9:09 AM, Peter Gervai <grinapo+ripeatlas@gmail.com> wrote:
On Tue, Nov 13, 2012 at 4:34 PM, Leo Bicknell <bicknell@ufp.org> wrote:
In a message written on Tue, Nov 13, 2012 at 12:57:22PM +0100, Peter Gervai wrote:
Atlas helped me today first to see a hard-to-detect bug: F-root anycast algo works fine for v4 but fails miserably for v6, as instead of AMSterdam it sends all traffic to some remote island in the pacific. ;-)
The good folks at ISC would be quite happy if you opened a ticket with them on this issue. Drop an e-mail to noc@isc.org to do so, including traceroutes and such would help a lot.
I did.
And we appreciate it :-) We'll look into the issue for F. - Jim (who's Ops at ISC, but is subscribed to this list from a personal account)
Jim Martin wrote: [...]
The good folks at ISC would be quite happy if you opened a ticket with them on this issue. Drop an e-mail to noc@isc.org to do so, including traceroutes and such would help a lot.
Actually, I started today to have a closer look (again) at connectivity to F from our neck in the woods[1]. Interestingly, the RTT for v4 and v6 are almost identical, but the packet drop over v4 sometimes is *significant*, while v6 seems to work like a charm :-)
I did.
I may do the same, after better understanding what the situation here is. It may be a local issue.
And we appreciate it :-) We'll look into the issue for F.
- Jim (who's Ops at ISC, but is subscribed to this list from a personal account)
Wilfried. [1] AA P-ID=6009 a-vie-as1853
participants (4)
-
Jim Martin -
Leo Bicknell -
Peter Gervai -
Wilfried Woeber