[oceanplexian]

> The old-skool approach is depending on a team of SREs or sysadmins to provision hardware for you

This assumes that K8s won't require a "team of SREs". My experience is you need the same amount of SREs to maintain Kubernetes, probably more, because now you have a complicated control plane, a networking nightmare, then you layer that on top of resource-contention issues, security issues, cloud provider compatibility issues, buggy controllers, the list goes on.

The only thing K8s is great for is the maintainers, the consultants, and highly experienced SREs that inevitably have to be hired to clean up the mess that was created. This is my experience working in two similar sized environments, one with >1M containers, and another with an equivalent scale of bare metal servers.

[nova22033]

then you layer that on top of resource-contention issues, security issues, and the list goes on.

applications running on bare metal don't have resource contention issues? or security issues?