Ask HN: How do you deal with multi-tenancy (in k8s)?

Curious how people deal with multi-tenancy in general and within Kubernetes in particular.

We run a B2B SaaS and have found configuration management to be an increasing pain, especially the fact that the resources that we dedicate to a single customer is a result of that configuration.

Right now we have clear separation between each tenant in the storage layer but share "compute resources" (yup, our service does some ML on behalf of our customers).

Since we already run on Kubernetes we are thinking of achieving that isolation through namespaces and maybe even through a custom resource/operator.