This made me laugh…
Canonical do have “live kernel” for Ubuntu where you can allegedly, in theory, patch, without rebooting (my arse)… Oracle have the UEK (Unbreakable Enterprise Kernel) which breaks if you try to run it on HyperV or in Azure …
I’ve worked on stuff like RHEL 8 running mission critical services, patching a fleet of servers with no outage, using the simplicity of round robin DNS to provide failover… But there’s not getting around NOT rebooting if you patch the running kernel, and rebootless patching of the O/S kernel makes no sense…
Serverless and containerised solutions - yeah… in theory should be possible… But - in the case of containerised, you’re seeing no outage, but the rug is pulled out from one container and another one surrepticiously deployed onto the rug - quicker than you can notice… I guess that’s more like a magician pulling the linen off a table without disturbing the crockery and the silver…
The serverless and containerised (docker and kubernetes) solutions are a new way of thinking, there’s no such thing as a “server” - whatever’s running those, should be agnostic, the instances are ephemeral, actual “servers” are stood up, and torn down, on demand…