How does Kubernetes work in general terms: what happens inside the cluster?

Kubernetes is a container orchestration system that manages deployment, scaling, and maintenance of applications within a cluster.

In general, the following occurs within a cluster:

• Master nodes (Control Plane) manage the state of the cluster. They contain components:

  • API Server — accepts commands and requests.
  • Scheduler — distributes pods across nodes.
  • Controller Manager — monitors the state of objects and maintains the desired state.
  • etcd — a distributed configuration and state storage.

• Worker nodes (Nodes) run containers in pods. Each node runs kubelet, which monitors the state of pods, and kube-proxy, which provides network access.

• The user or system sends a declarative description of the desired state (e.g., Deployment with a number of replicas).

• The Scheduler assigns pods to nodes considering resources and constraints.

• Kubelet on nodes starts containers via the container runtime.

• Controllers ensure that the number of running pods matches the desired count.

• If a pod fails, Kubernetes automatically restarts it or creates a new one.

Thus, Kubernetes provides automatic management of container lifecycle, load balancing, scaling, and recovery from failures.