I’m actually doing the inverse networking where everything is on wifi. It ended up being cheaper than wired.
I don’t like the single points of failure, but other than that it looks cool.
It’s not meant as a highly-reliable cluster :-)
Isn’t that the purpose of a cluster though?
Of a production one, yes for sure. As I mentioned, the goal isn’t to build a production cluster, but to learn.
Also, removing all SPoF here would mean 2 masters, and 2 load balancers, with load balancing on both. On
raspberry pis at home, this could get very messy quite easily.