As we can see in the following diagram, the client uses kubectl, which is a command-line interface, to send requests to the API Server. The API Server, a hub between the master components, will respond to the client requests and push and pull the object information from etcd. If a new task is created, such as run pods, the scheduler will determine which node should be assigned to do that task. The Controller Manager monitors the running tasks and responds if any undesired state occurs.

The API Server fetches the logs from pods via the kubelet: