Certain destinations are reachable with difficulty from some OVH IPs.
We are investigating.
Date: 2014-03-21 15:44:36 UTC The sessions have been restored. According to the timing of events, the configuration issue was there from the start, but did not have an impact until the end of yesterday when there was a cut in the fibre on the Level3 network. The cut caused this from one of the intercos and thus a change in the routing which updated the config issue.
Everything has now been corrected on Level3's side.
Date: 2014-03-21 15:40:59 UTC Level3 has identified the source of the problem - a bad configuration on their side. We have remounted the interfaces and we are about to restore the BGP sessions on our routers in Roubaix (\"nitro\" infrastructure).
Date: 2014-03-20 23:47:00 UTC We have isolated the issue in 2 BGP sessions with Level3 forwarded to routers in Roubaix via our dwdm network. We have cut them and recovered the normal routing with Level3 in Paris and on the range 220.127.116.11/16. No prefixes were announced on these sessions that we're only using for outgoing traffic and therefore we do not explain the reason why at this level, the traffic to 18.104.22.168 was blackholed. We will continue troubleshooting with Level3 in order to understand the exact origin of the issue.
Date: 2014-03-20 21:06:09 UTC We identified an issue with Level3 in Paris who is impacting network 22.214.171.124/16. The traffic seem to be dropped by Level3 equipment in Paris. We have shutdown the BGP session with L3 in Paris. As emergency plan, we forced traffic going trough Londons but had to split the announcement of 126.96.36.199/16 in 2 /17. We are working with Level3 to find a permanent solution.