Kitz Forum
Internet => General Internet => Topic started by: broadstairs on October 04, 2021, 05:57:41 PM
-
In case anyone is interested both of these fail to resolve using DNS, at least for me!
Stuart
-
WhatsApp and Oculus too:
https://www.bbc.co.uk/news/technology-58793174
Has Facebook purchased a DNS company yet?
-
In case anyone is interested both of these fail to resolve using DNS, at least for me!
Stuart
When I checked I could resolve Facebook.com but not www.facebook.com or m.facebook.com. That was using Opendns.
-
Looks like facebook BGP routes have been removed. FB appears to be an internet blackhole at the moment. No traffic can get routed to them.
Andrew
-
Our lasses mates will be gutted they wont be able to share what they had for tea. ;)
-
One interesting side effect of this was the load Facebook apps put on recursive DNS servers.
Since Facebook's DNS servers were down, apps kept making DNS requests, essentially DDoSing recursive DNS servers. Cloudflare's 1.1.1.1 reported loads higher than normal, Google's 8.8.8.8 became slower in some places, and I assume many ISPs had the same problem.
This resulted in other sites being "slow" or "offline" because the DNS servers were too slow or didn't reply to queries at all...
-
The Technology Section (https://www.bbc.co.uk/news/technology) of the BBC web-site now has a "What happened to Facebook, WhatsApp, and Instagram? (https://www.bbc.co.uk/news/technology-58800670)" article.
-
Andrews and Arnold have taken some measures against disaster with the separate independent domain name aastatus.net which I think is hosted by their friend-company Watchfront, and they also iirc use Watchfront’s network for disaster-redundancy [or whatever?] where needed. Watchfront is their collaborator in firebrick.co.uk. I have this very vague dubious memory about AA using Zen’s network for some disaster measures - is that utter nonsense that I just dreamed up?
I was thinking that Facebook need a completely independent, second, parallel management network, which also has a route into the main management network and their normal AS. But they need second interfaces (maybe not physical ones) into switches, routers, critical servers, so they can monitor and configure them even when the main AS is dead. It would be a nightmare to implement and test and would cost an absolute fortune, but they can afford it.