With the Operator approach, Kubernetes functionality can be extended in a friendly way. Operator = CRD + Controller. First you generate the CRD from a yaml definition, then the Controller continuously watches the data in etcd and performs the corresponding actions. There is a lot of tedious and repetitive work involved in developing an Operator.

Helm is the package manager for Kubernetes; Operator is used to manage stateful distributed applications on Kubernetes. This article mainly describes how to use Helm and Operator to quickly deploy Prometheus on a Minikube cluster and use Grafana to view monitoring data. For Minikube installation, see Building a Remote Kubernetes Development Environment; for Helm configuration, see The Package Manager for Kubernetes – Helm; for Operator, see Kubernetes Complex Stateful Application Management Framework – Operator.

1. Origins Originally, two CoreOS employees, in order to make deploying etcd clusters easier, bound corresponding logical operations to the create, delete, and update events of the etcdCluster object, using Kubernetes to automate etcd cluster management. A few months later, at a KubeCon conference, they shared this approach, which they called Operator, and it received a strong response from the community.