Clustering¶

Kubernetes clustering allows you to manage a group of hosts as a single entity. This enables applications to be deployed across a cluster of machines without tying them specifically to one host, providing high availability and scalability. In Canonical Kubernetes the addition of k8sd to the Kubernetes ecosystem introduces enhanced capabilities for cluster coordination and management.

Kubernetes Cluster Topology¶

A Kubernetes cluster consists of at least one control plane node and multiple worker nodes. Each node is a server (physical or virtual) that runs Kubernetes components. In Canonical Kubernetes, the components are bundled inside the k8s-snap. The cluster’s topology divides responsibilities between the control plane node(s) and the worker nodes, ensuring efficient management and scheduling of workloads.

This is the overview of a Canonical Kubernetes cluster:

$@startuml set separator none title Canonical Kubernetes Cluster 'left to right direction !include <C4/C4> !include <C4/C4_Context> !include <C4/C4_Container> !include <C4/C4_Component> System_Boundary("K8sSnapDistribution.ControlPlane", "Control Plane", $tags="") { Container(K8sSnapDistribution.ControlPlane.K8sSnap, "K8s Snap", $techn="Snap Package", $descr="", $tags="", $link=""){ Component(K8sSnapDistribution.ControlPlane.K8sSnap.K8sd, "K8sd", $techn="", $descr="K8sd Instance", $tags="", $link="") Component(K8sSnapDistribution.ControlPlane.K8sSnap.Services, "Kubernetes Services", $techn="", $descr="kube-apiserver, kubelet, kube-proxy, kube-scheduler, kube-controller-manager, and container runtime.", $tags="", $link="") Component(K8sSnapDistribution.ControlPlane.K8sSnap.State, "State", $techn="dqlite", $descr="Datastore holding the cluster state", $tags="", $link="") Component(K8sSnapDistribution.ControlPlane.K8sSnap.DataStore, "Dqlite Datastore", $techn="dqlite", $descr="Datastores holding the Kubernetes state", $tags="", $link="") } Container(K8sSnapDistribution.ControlPlane.Pod, "Workloads", $techn="Pod", $descr="", $tags="", $link="") } System_Boundary("K8sSnapDistribution.Worker", "Worker", $tags="") { Container(K8sSnapDistribution.Worker.K8sSnap, "K8s Snap", $techn="Snap Package", $descr="", $tags="", $link=""){ Component(K8sSnapDistribution.Worker.K8sSnap.K8sd, "K8sd", $techn="", $descr="K8sd Instance", $tags="", $link="") Component(K8sSnapDistribution.Worker.K8sSnap.Services, "Kubernetes Services", $techn="", $descr="kubelet, kube-proxy, and container runtime.", $tags="", $link="") } Container(K8sSnapDistribution.Worker.Pod, "Workloads", $techn="Pod", $descr="", $tags="", $link="") } Rel(K8sSnapDistribution.ControlPlane.K8sSnap.K8sd, K8sSnapDistribution.ControlPlane.K8sSnap.Services, "Configures", $techn="", $tags="", $link="") Rel(K8sSnapDistribution.ControlPlane.K8sSnap.K8sd, K8sSnapDistribution.ControlPlane.K8sSnap.State, "Keeps state in", $techn="", $tags="", $link="") Rel(K8sSnapDistribution.ControlPlane.K8sSnap.Services, K8sSnapDistribution.ControlPlane.K8sSnap.DataStore, "Keeps state in", $techn="", $tags="", $link="") Rel(K8sSnapDistribution.Worker.K8sSnap.K8sd, K8sSnapDistribution.Worker.K8sSnap.Services, "Configures", $techn="", $tags="", $link="") Rel(K8sSnapDistribution.ControlPlane.K8sSnap.Services, K8sSnapDistribution.ControlPlane.Pod, "Manages", $techn="", $tags="", $link="") Rel(K8sSnapDistribution.Worker.K8sSnap.K8sd, K8sSnapDistribution.ControlPlane.K8sSnap.K8sd, "Interacts with", $techn="", $tags="", $link="") Rel(K8sSnapDistribution.Worker.K8sSnap.Services, K8sSnapDistribution.Worker.Pod, "Manages", $techn="", $tags="", $link="") SHOW_LEGEND(true) @enduml$

The Role of `k8sd` in Kubernetes Clustering¶

k8sd plays a vital role in the Canonical Kubernetes architecture, enhancing the functionality of both the Control Plane and Worker nodes through the use of microcluster. This component simplifies cluster management tasks, such as adding or removing nodes and integrating them into the cluster. It also manages essential features like DNS and networking within the cluster, streamlining the entire process for a more efficient operation.

Integration into the Kubernetes Cluster Topology¶

For Canonical Kubernetes, the detailed view of the two types of node is as follows:

Control Plane Node¶

The control plane node orchestrates the cluster, making decisions about scheduling, deployment and management. With the addition of k8sd, the control plane node’s components include:

API Server (kube-apiserver): Acts as the front-end for the Kubernetes control plane. It exposes the Kubernetes API and is the central management entity through which all components and external users interact.
Scheduler (kube-scheduler): Responsible for allocating pods to nodes based on various criteria such as resource availability and constraints.
Controller Manager (kube-controller-manager): Runs controller processes that regulate the state of the cluster, ensuring the desired state matches the observed state.
k8s-dqlite: A fast, embedded, persistent in-memory key-value store with Raft consensus used to store all cluster data.
k8sd: Implements and exposes the operations functionality needed for managing the Kubernetes cluster.

Worker Node¶

Worker nodes are responsible for running the applications and workloads. Worker nodes, can interact with the k8sd API, gaining capabilities to manage its entire life-cycle. Their components include:

Local API Server Proxy: This component forwards requests to the control plane nodes.
Kubelet: Communicates with the control plane node and manages the containers running on the machine according to the configurations provided by the user.
Kube-Proxy (kube-proxy): Manages network communication within the cluster.
Container Runtime: The software responsible for running containers. In Canonical Kubernetes the runtime is containerd.