Post

That said, daemons that all watch and update the same resources has got to be an anti-pattern. Thats like trying to horizontally scale a controller to infinity.

Great write up - very interesting and informative. Thank you

Interesting! Makes me wonder what drove them away from sending node details via grpc. Occasional failure would be a more localised (per node) and recoverable. (Admitted: Hindsight always easy).