Configuration management systems (I'm thinking Ansible, Puppet etc) are used to declaratively execute configuration tasks at fairly slow timescales. You wouldn't use them them to say, make 1000s of VM/pod placement decisions per second.

Using DCM, we're building a Kubernetes scheduler (see the k8s-scheduler folder in the repo). It continuously listens for new pods being created and other events in k8s, and schedules these pods onto nodes using complex rules described in SQL. This is not the kind of problem that configuration management systems target.