We only have argo enabled on one account, and as far as I understand Argo would only, potentially, affect the routing from the nearest eyeball colo to the origin via it’s nearest colo. I don’t think this has anything to do with Argo tbh.
Thanks for looking into this @amayorga
Where’s it UP:
1.|-- www.adsl.vf 0.5% 200 412.6 183.2 2.2 2234. 369.5 2.|-- ??? 100.0 200 0.0 0.0 0.0 0.0 0.0 3.|-- 10.183.69.17 1.5% 200 1768. 248.9 4.7 2271. 452.8 4.|-- 172.29.176.113 62.5% 200 1816. 249.6 4.0 1987. 497.8 5.|-- 126.96.36.199 95.0% 200 21.4 19.2 13.8 24.9 4.2 6.|-- 188.8.131.52 1.0% 200 2868. 273.4 13.0 2868. 510.5 7.|-- ??? 100.0 200 0.0 0.0 0.0 0.0 0.0 8.|-- 184.108.40.206 3.0% 200 2645. 403.6 130.8 2645. 507.8
Unfortunately it makes no sense for me to upgrade to a CF paid plan for just 0.5GB of monthly traffic app.
But even if it made sense financially speaking we have the following:
This I do understand.
This I can’t. Specially when the costs in the USA and EU are more or less the same. Why not route to another EU colo?
There was a time recently where EU traffic was so off the charts that Cloudflare had to route a lot of lower plan traffic to the US. They didn’t say what was causing it, but it did clear up.
If you consider that, say…80% of the domains on Cloudflare are on free plans, and EU POPs are saturated, that’s a lot of traffic to distribute to neighboring regions. If you look at the judge.sh test @soldier_21 posted, you can usually see how saturated a POP/region is by which plan level gets shunted off to some other place.
That’s very interesting @sdayman… Perhaps COVID restrictions are preventing them from expanding their EU POPs capacity, I know Equinix, which I guess CF are tenants of in many European cities, has some strict requirements at the moment for on-site (if it’s even possible to get someone out there - I imagine the Lisbon staff aren’t going anywhere soon)
Just an update. A few minutes ago CF started routing to BCN and MRS. That makes more sense.
Still I had to start working on an alternative in case this happens again.
I will start monitoring /cdn-cgi/trace and measuring latency from a few machines and whenever it starts routing to an undesired location I will programatically stop proxying through CF and change the DNS to an NGINX box in EU acting as reverse proxy.
Nope. False alarm. CF was routing my mobile ISP data to BCN and MRS. Got home to my fiber provider and again to MIA.
May I ask if it is possible and if it could depend to with whom does (y)our ISP have peers? Meaning, the route should choose depending on the ISP “by default”, but moreover if the user is using Cloudflare DNS like 220.127.116.11, which has all over the World POPs, could it be possible to have different routing due to the packet trying to go “by default route” for (y)our home IP address provided by (y)our ISP rather than following the “exit node” or zig-zag on the Cloudflare network?
I ask because an example where one local ISP provider in my country (90% optical, fiber) which has 6-7 peers within the USA, and only the one peer to within our home country.
Just wondering …
@amayorga were you able to check if there is a routing problem? I can’t believe that you have a peering problem with a major ISP like Vodafone that routes requests more than 6000 kms away.
I also find that I have less traffic when going through cloudflare but not sure if this is related.
Can you tell me which city has a colo = MRS?
Thank you very much !
What do you mean? What does MRS stand for or from which city am I being routed to MRS?
I mapped out the servers and here is the test:
Loc France ok but I don’t know the city of the server which is MRS
uag=Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/88.0.4324.150 Safari/537.36
They follow the IATA airport codes.
Ok ok, thank you very much for your response!
This topic was automatically closed 24 hours after the last reply. New replies are no longer allowed.