Components constituting the Kubernetes (component) mainly three types, Master assembly, Node assembly, Addons (auxiliary plug type). Respectively undertake different work together to build a kubernetes.
Master node
Master assembly provides a control plane cluster. Main assembly of the cluster to make global decisions (e.g., scheduling), and which detects and responds to cluster events (e.g., when a copy controller replicasfield unmet started when a new pod ).
The main assembly can be run on any computer in the cluster. However, for simplicity, the setup script normally starts all the main components on the same computer, and the user is running on a vessel is not in this computer.
*** The official explanation from the above description, I understand that a console in the existing IT infrastructure in many hardware or software in this way, Master and node way. Issued under the Master load strategy, management node running the life cycle.
Equivalent Kubernetes brain, responsible for the management and control of the entire cluster. It is often impossible to deploy function node does not run container (container) on this machine.
Run the following set of critical processes on the node Master:
kube-apiserver: providing a key service process interface HTTP Rest ful style, where all resources are Kubernetes add, delete, change, and other operations of the only entrance, the entrance process management is also Kubernetes cluster.
kube-controller-manager: Kunernetes all the resources in the automation control center,
These controllers include:
- Node Controller: responsible for attention and response when a node fails.
- Copy Controller: is responsible for maintaining the correct amount of pod for each copy controller objects in the system.
- Endpoint controller: filling endpoint object (i.e., connection service and pods).
- Services account and token controller: Creating default accounts and API access token for the new namespace.
kube-scheduler: factors responsible pod scheduling process, scheduling decisions to be considered include individual and collective resource requirements between hardware / software / policy constraints, affinity and anti-affinity specifications, data location, interference workloads and deadlines
etcd: Kubernetes responsible for all storage resources processes.