Thanks! Love it.

Because the address is global and is routed to the nearest Cloudflare PoP. If you are in New Zealand then you get routed to Auckland, if you are in Scotland then you get routed to Edinburgh – and then Cloudflare transfer the data on their backbone to the GRE tunnels CCP has with them. See https://www.cloudflare.com/network/. My route to Tranquility, when not logged onto CCP’s work-at-home VPN, looks like this:

3 VFIS-GR-1.c.is [193.4.254.160]
4 fo-0-1-1-0-CCR01-Sidumuli.c.is [217.151.183.98]
5 rix-tg-gw.cloudflare.com [195.130.211.50]
6 172.65.201.188

where c.is is the backbone name for Vodafone Iceland and rix-tg-gw is at the Reykjavik Internet Exchange (RIX) where Vodafone Iceland and Cloudflare peer directly.

Also forwarded.

I’ve forwarded this.

For now many Virgin Media customer have used VPN (from the UK to the UK) to bypass the fault. Long-term then either Virgin Media finds that fault and fixes it (I have a relatively good confidence in that they do) or you should consider looking for another ISP that offers a high-quality, stable internet that is suitable for long-lived TCP connections. Honestly then Virgin Media did appear to offer such services before 22 o’clock on 23 February.

I don’t know if you can get compensated by Virgin Media for having to use VPN to bypass the fault.

We can’t fix internet network routing with a patch, and the server is fine.

Didn’t get DC’d between 23:50 to 00:20, was running traceroute the entire time and this was the route.

 5     *        *        *     Request timed out.
  6    31 ms    60 ms    50 ms  m674-mp2.cvx1-b.lis.dial.ntli.net [62.254.42.162]
  7    22 ms    33 ms    22 ms  213.46.182.162
  8    24 ms    21 ms    37 ms  adm-bb3-link.ip.twelve99.net [62.115.136.194]
  9    27 ms    38 ms    27 ms  adm-b10-link.ip.twelve99.net [62.115.120.227]
 10    29 ms    22 ms    23 ms  cloudflare-ic342028-adm-b10.ip.twelve99-cust.net [62.115.172.231]
 11    27 ms    32 ms    32 ms  172.65.201.188

Trace complete.

Last 16 hours; happens shortly before hour:

image1578×536 52.3 KB

I would prefer to get the full details in support tickets to us, please, and then post the ticket number here.

Yeah, this something hourly. I think Virgin Media is possibly fetching BGP routes from the internet every hour and applying them to some equipment. But then cost-saving changes are applied on top (that routing to Amsterdam). But for a brief moment before the cost-saving routes are applied then the traffic goes a different route and the equipment on that route sends back RST packets.

There is a grain of truth here on why it may feel like this matters. What matters is if the client is active or not, and it always active in space (receiving physics simulation updates) while it may be less active in station and therefore a brief issue like this may pass it by.

The scenario we think we are looking (this is a conjecture/hypothesis; but not a theory yet, for that we are waiting on more data from players and replies from Virgin Media and Cloudflare) are regular but very brief BGP flaps / routing changes. Your EVE Client initially established a TCP connection to Cloudflare’s Amsterdam PoP (which is incorrect, but that’s how Virgin Media routes their traffic ) and the firewalls and proxies there will have the state of your TCP connection.

That state is not sy...

Forwarded to Cloudflare; they were quick to point out they only got half of the data in our first reply.

Is it a count-down?!?

We will know when either they tell us or the disconnects stop happening… There was a disconnect spike at 19:00 this time and nothing at 19:55. But then again, there were minor spikes hourly even there were larger spikes every 4 hours.

Our test server network is in a different geographical location to the production network and is not behind Cloudflare’s systems. If you trace the route, then you will get very different paths.

% nslookup tranquility.servers.eveonline.com
Name: d638e439e07f413c97200d057e5ebd05.pacloudflare.com
Address: 172.65.201.188
Aliases: tranquility.servers.eveonline.com
tranquility.servers.eveonline.com.cdn.cloudflare.net

% nslookup singularity.servers.eveonline.com
Name: singularity.servers.eveonline.com
Address: 87.237.38.2

(And this IP address above for TQ is not its “correct” destination IP address; only Cloudflare can access that one.)

В связи с ограничением валютного обмена с банками Российской Федерации наш поставщик платежных сервисов больше не обрабатывают платежи в рублях, из-за чего приём оплаты в рублях в магазине EVE Online приостановлен на неопределённый срок.

Игроки из Российской Федерации смогут по-прежнему оплачивать товары на нашем сайте, но списание будет производиться в долларах США, как происходит в других странах, валюты которых технически не поддерживаются.

Indeed, we think the trace route is “corrupt” in the sense that you started it when the new route was in effect and as the trace route was running then new old route came back into effect; so the result is a mix of the two. We are still hoping it provides insight to Cloudflare and Virgin Media, along with the discovery of the regularity, but more trace routes would help us and them piece together what that new route is.

We are amazed that you caught it; Cloudflare had been asking for this and we had been telling them this would be next to impossible.

BTW, we have already sent this to Cloudflare and Virgin Media. VM has already replied and escalated the case further.

Hello all Virgin Media UK EVE players!

Fantastic engagement in this thread here as we at CCP continue to feed information to Virgin Media UK and Cloudflare on this issue and great team work!!! Special thanks to @Jander_Massek for catching the route change and to @Harvy_Powney, @Cade102 and everyone else for spotting the regularity and sending us details.

We have a favour to ask you. There appear to be large disconnect spikes every 4 hours, most notably at 11:55, 15:55, 19:55, 23:55 (between midnight and noon then UK players are more offline so it’s difficult to tell) but possibly there it looks like there are also minor hourly disconnects (but sometimes none). This looks a bit like some regular processing that is happening in the approx. 10 minutes before the hour, most notable every 4 hours.

We would want to get more trace routes when the disconnects happens, so if you could run them at 11:55, 15:55, 19:55, 23:55 and if you disconnect to see if results are different than your regular ...

I would love to get the full data; if you don’t want to post that here, then please file a support ticket and name drop me and GM Mechanic.

Thanks; the network people are now chewing on this.

BTW, 162.158.32.0/22 is AS13335, which is Cloudflare. So that last route is VM => CF => Telia => CF.