Alexander Gall gall at switch.ch
Wed Aug 23 10:05:18 CEST 2006

On Tue, 22 Aug 2006 17:45:29 -0400, "James Jun" <james at towardex.com> said:

>> I checked with source routing (yes, this still works in IPv6 and it's
>> wonderful for debugging!) that the path is symmetrical.  My guess is
>> that there is an MTU mismatch on the interconnection between GBLX and
>> Equinix (probably a tunnel).

[oops, s/Equinix/OCCAID/]


> Thanks for the insight.  I checked the ticket db and I still have a ticket
> opened by a user back in 2005 regarding tcp sessions hanging when
> downloading from HEAnet, or any other paths going thru OCCAID<->GBLX in
> Ashburn.

> For the time being, I've temporarily rerouted our east-coast traffic with
> GBLX to Chicago and Palo Alto interconnects which are not affected by
> mtu-mismatch somewhere to my knowledge.  The path may be less optimal in
> latency-wise but let me know if you can establish a tcp session and transfer
> data now.[1]

Thanks.  However, GBLX is now blackholing the entire 2001:500:4::/48
for us in Paris.  I have filtered the /48 from them and can reach ARIN
through Level3 just fine.


More information about the ipv6-ops mailing list