Controllers

• watch the resources.
• reconcile the actual state and desired state.

Deployment

• Supervise and schedule programs.
• has
  • the name of the container image,
  • the number of replicas
  • and whatever else to start the container.
• creates an associated object ReplicaSet.

$ kubectl get deploy
$ kubectl describe deploy xyz

Pods

• represents a group of one or more containers.
• schedules a set of containers together, running on the same node, and communicating locally.

ReplicaSets

• manages a group of identical pods, or replicas.

• starts containers if the specified replica not met.

• restarts containers if crashed.

• Deployment creates ReplicaSets.

• ReplicaSet create Pods.

Scheduler

The Kubernetes scheduler watches for a Pod that isn’t yet running on any node, finds a suitable node for it, and instructs the kubelet on that node to run the Pod.

How does the Kubernetes scheduler work?

Services

• provides a single, unchanging IP address or DNS name for Pods.
• defines a list of ports, plus a selector and a type. Requests will be forwarded to any Pods matching the specified set of labels.

• Resources like Deployments are represented by records in Kubernetes’s internal database. Externally, these resources can be represented by text files (known as manifests) in YAML format. The manifest is a declaration of the desired state of the resource.
• kubectl is the main tool for interacting with Kubernetes, allowing you to apply manifests, query resources, make changes, delete resources, and do many other tasks.