By the time you get to need to learn Kubernetes, people are usually too far deep in their engineering careers and already have this 'not even more shit' fatigue.

Yup.

K8s is worth having a dedicated platform engineering team around to keep running. It's a complicated system, and having feature devs waste cycles on it gets you a half assed unstable cluster, and half assed features.

Good news, most people don't need it, Ave the ones that do can afford a dedicated team

In my experience, the problem always comes from salesmen wanting to parrot shit about being "cloud native", "high availability", "scalable", "state of the art", and other bullshit.

And they end paying absurd amounts of money and man hours for something they could easily run in a couple computers gathering dust in the corner.

It's a different type of engineering, if it feels unnatural to learn then don't do it and take your career in a different path. As said in a different comment, companies that want to use k8s at scale, should use dedicated teams to manage this, or alternatively look into managed solutions that you have to pay for. For smaller systems and software, serverless is likely a better value to host on, that scales just as well but has a lower learning curve and is less flexible in the nitty gritty details like chipset architecture, networking, policies, PKI, DNS etc.

@CoreFusionX that's how I learned it.

CTO demanded we move from two ec2 instances and an ALB, to a rancher managed multi cluster deployment.

We had a dozen MAUs, but he wanted to use the big key words when wooing future partners. It didn't work. I learned a fuck tonne though