It’s DNS far too often in large part because Linux has the default behaviour of having a singular (“the”) name server. If you configure multiple, this is not t

It’s DNS far too often in large part because Linux has the default behaviour of having a singular (“the”) name server. If you configure multiple, this is not the same as in Windows, MacOS, iOS, or even Android. It’s not a pair of redundant servers, it’s a sequential list of logically layered configurations, like an override. An outage of the primary will cause repeated timeouts until it comes back up. Contrast this with every other operating system that does not seek to emulate forever and ever the specific network setup of some computer lab at Berkeley in the 1970s, where failover is near instant (sub-second in Windows Vista and later) and persists, so that a small failure doesn’t become a big one. To compound things, thanks to response caching and the relatively good stability of typical DNS servers, this failure mode is rare enough that most admins have never encountered it and probably don’t even recognise the problem. Or worse, they’ll handwave it away saying things starting with “You can…” or “You should…”. Who’s this “You” person? It’s not me! It’s not most admins, none I’ve met are named You. None have changed this default in my experience. Not one. That would be a design decision requiring sign off, testing, rollout, etc… I’ve seen this take down corporations for a day. Defaults matter.