Kubernetes - Components

Last Updated: 2024-02-06

How are components running?

Components (e.g. binaries) can run in different ways:

  • Not containerized: run as systemd services.
    • e.g. kubelet, containerd, docker.
    • config: systemd unit files.
    • start: systemctl start.
    • check status: systemctl status.
    • check logs: journalctl.
  • Containerized: static pods: managed directly by kubelet.
    • config: /etc/kubernetes/manifests.
    • start / update: by kubelet (watching /etc/kubernetes/manifests).
    • check status: kubectl get (mirror pods are automatically created, so they are visible on kube-apiserver, but not modifiable).
    • check logs: kubectl logs.
  • Containerized: (normal) pods: managed by apiserver.
    • config: any yaml files.
    • start / update: watched by kube-apiserver; kube-apiserver talks to kubelet to manage pods.
    • check status: kubectl get.
    • check logs: kubectl logs.

Control Plane Components

Must-have Kubernetes components.

Name Using static pods kubernetes-the-hard-way
kubelet systemd service (no kubelet)
kube-apiserver static pod systemd service
kube-controller-manager static pod systemd service
kube-scheduler static pod systemd service
etcd static pod systemd service
kube-proxy daemonset
container runtime (e.g. containerd) systemd service systemd service


  • kubelet
    • if kube-apiserver and other binaries are running as static pods, kubelet runs the static pods according to /etc/kubernetes/manifests. kubelet writes log to journald, check by journalctl -u kubelet.
    • if kube-apiserver and other binaries are deployed as systemd services, kubelet is not running on control plane nodes (as illustrated in kubernetes-the-hard-way).
  • kube-apiserver: the API Server.
  • etcd: Most of the kubernetes components are stateless and state of each component comes from the etcd db files.
  • kube-controller-manager: Controller Manager of the built-in controllers.
  • cloud-controller-manager: embeds cloud-specific control logic.
  • kube-scheduler: Scheduler.
  • containerd: The kubelet always directs your container runtime to write logs into directories within /var/log/pods.

Worker Node Components

Worker Node: virtual or physical machines, managed by the control plane and contains the services necessary to run Pods.

  • kubelet: Talks to API Server.
  • kube-proxy: responsible for implementing a virtual IP mechanism for Services of type other than ExternalName.
  • Container Runtime: e.g. containerd, a daemon on worker nodes. Manages the container lifecycle.
  • monitoring / logging: supervisord, fluentd

The Pod Lifecycle Event Generator or PLEG is a daemon on each node that ensures the cluster's current state matches the desired state of the cluster. This might mean restarting containers or scaling the number of replicas but its possible for it to encounter issues.

The kubelet monitors resources like memory, disk space, and filesystem inodes on your cluster's nodes.


  • kubelet talks to CSI to have the storage ready.
  • kubelet talks to container runtime by grpc. kubelet instructs the Container Runtime to spin up a container.
  • container runtime talks to CNI plugin to get network setup.
  • (if use static pods) Control Plane node kubelet runs API Server
  • API Server talks to worker node kubelets
  • API Server clients wihin Control Plane: controllers, scheduler, etcd.
  • between API Server and human users: kubectl, kubeadm, REST API, client libraries.
  • between API Server and Nodes: kubelet.
  • Other API Server clients: CI/CD (Jenkins), Dashboard / UI

Access management:

authentication -> authorization -> admission control ("mutating" / "validating" admission controllerss)

the API server implements a push-based notification stream of state changes (events), also known as Watch

One of the reasons why watches are so efficient is because they’re implemented via the gRPC streaming APIs.

apiserver <=> (worker node) kubelet

Communications between the apiserver and kubelet are bi-directional. For some functions the kubelet calls the apiserver and for others the apiserver calls the kubelet.

The kubelet has a REST API, the API exposes info about the pods running on a node, the logs from those pods, and execution of commands in every container running on the node; typically exposed on TCP port 10250, which the API server calls for some functions. These connections terminate at the kubelet's HTTPS endpoint.

apiserver => (worker node) kubelet for:

  • Fetching logs for pods.
  • Attaching (usually through kubectl) to running pods.
  • Providing the kubelet's port-forwarding functionality.

kubelet watches apiserver: apiserver supports a "watch" mode, which uses the WebSocket protocol. In this way the kubelet is notified of any change to Pods with the Hostname equal to the hostname of the kubelet.

  • new pods with specific node labels (the labels are added by the scheduler). When it sees a pod for the node it's running on , it co-ordinates with the Container runtime on the node (usually over a UNIX socket) to start the appropriate containers.


  • apiserver server: /etc/kubernetes/pki/apiserver.crt
  • apiserver client: /etc/kubernetes/pki/apiserver-kubelet-client.crt
  • kubelet server: /var/lib/kubelet/pki/kubelet-server-current.pem
  • kubelet client: /var/lib/kubelet/pki/kubelet-client-current.pem

Life of a deployment (Put everything together)

  • User submits a deployment.yaml to API Server.
  • A Deployment is stored in etcd; only API Server can access etcd.
  • kube-controller-manager sees the Deloyment from the API Server and create corresponding Pods.
  • kube-scheduler: assigns a Pod to a Node.
  • kubelet talks to the API Server and read the schedule, runs the Pods.
  • End-users calls the running pods through kube-proxy (kube-proxy calls API Server to get Services).