18 Mar
2020
18 Mar
'20
10:52 p.m.
On 18/Mar/20 22:22, Nick Hilliard wrote:
Yeah. I was thinking more for the case of customer-facing anycast resolvers, in which case BGP down means that the network is down, and if the network is down it doesn't matter than DNS is also down because their shared fate means that when BGP is back up, DNS will start working again.
As much as possible, I'll always choose to have the most basic infrastructure available under abnormal conditions, regardless of the service. ME3600X's and ASR920's, for example, will install 0/0 and ::/0 in FIB last. If your access to the core depends entirely on BGP in such scenarios, you will be unable to access the it for as much as 10 minutes. Mark.