[afnog] Another Perspective - Kentik's View on the Facebook Outage

Nishal Goburdhan nishal at controlfreak.co.za
Thu Oct 7 12:19:50 UTC 2021


On 7 Oct 2021, at 13:33, Markus Akena Wipfler wrote:


> either way it's bad design and I doubt that FB
> makes that kind of mistake.

since this is a technical forum, perhaps a useful post would be to 
explain how _you_ would design this better?  i admit, i have zero 
insider knowledge of the incident, nor of FB’s infrastructure, so this 
might be instructional to - at least - me.


> Also no one is connecting the dots what is happening to FB as a whole 
> atm..

we try to focus on the tech here..


> Further more the /19 supernet was still in GRT during whole outage and 
> only
> the more specific DNS prefixes were missing.

there is _at least_ one credible source that says that it was more than 
just their DNS prefixes that went missing:
https://twitter.com/ryan505/status/1445072241256013828?s=12

do you have BGP data that shows otherwise?


> Further more I assume without checking, that the DNS IPs are anycasted 
> so
> it would mean that the automation script  has access globally to all 
> BGP
> speakers. lol
>
> If you want I can sell you a nice fridge on the north pole...

thank you, but no.  i’d rather hear your design for improvement.
after all, we are all here to learn.

-n.



More information about the afnog mailing list