Skip to content
Back to glossary
Container Orchestration

What is Kubernetes?

Kubernetes is an open-source container orchestrator that automates deployment, scaling and operation of containerised apps across server clusters.

What Kubernetes does

Kubernetes (often abbreviated K8s) takes containerised workloads, typically Docker images, and runs them across a fleet of servers as a single logical resource. It handles three jobs that, before its release in 2014, were either manual or stitched together with bespoke scripts:

  • Scheduling : placing each container on a node that has enough CPU, memory and the right hardware (GPU, SSD, network zone).
  • Self-healing : restarting crashed containers, replacing unhealthy nodes, rescheduling pods when a host disappears.
  • Service discovery & load-balancing : exposing groups of containers behind a stable virtual IP and DNS name, distributing traffic across replicas.

Why teams adopt it

The honest answer is that Kubernetes turns infrastructure into a declarative API. You describe the desired state, "3 replicas of this app, 500 MB RAM each, behind this domain, talking to this database", and the cluster reconciles reality to match. That contract works on AWS, Azure, GCP, OVH, on-premise and on a laptop with k3s, which is why it became the de-facto standard for portable cloud-native workloads.

In our experience at Hidora, the teams that succeed with Kubernetes already have:

  1. Stateless services that can be killed and restarted without losing data.
  2. Decent CI/CD to build and tag images on every commit.
  3. A monitoring story : at minimum metrics + logs aggregated outside the cluster.

Without those three, Kubernetes adds operational weight without unlocking the benefits.

Common production components

A real cluster rarely runs vanilla Kubernetes alone. The typical stack we deploy includes:

  • Ingress controller (NGINX, Traefik, or Gateway API) to expose HTTP traffic.
  • CNI plugin (Cilium, Calico) for pod networking and network policies.
  • Storage CSI for persistent volumes backed by Ceph, EBS or Longhorn.
  • GitOps controller (ArgoCD, Flux) so the cluster state is reconciled from Git.
  • Observability (Prometheus, Loki, Grafana) for metrics, logs and dashboards.

When not to use it

Single-container apps, simple internal tools, or workloads with very specific kernel requirements often run more cheaply on a managed PaaS or a single VM. Kubernetes earns its keep when you have at least 4–5 services, multiple environments, and a team that can absorb the learning curve.

Related Hidora services

  • Managed Services : outsourced operation of your existing or new K8s clusters with 24/7 monitoring.
  • Hikube : Swiss sovereign managed Kubernetes built and operated in Geneva.
  • Consulting : audit, migration and training when you adopt Kubernetes for the first time.