Glossary · Definition
Kubernetes
Kubernetes (k8s) orchestrates many containers across many machines. You declare the desired state — 'run 3 replicas of this image, expose this port, restart on crash' — and the control plane keeps reality matching it.
Definition
Kubernetes (k8s) orchestrates many containers across many machines. You declare the desired state — 'run 3 replicas of this image, expose this port, restart on crash' — and the control plane keeps reality matching it.
What it means
Workloads (Pods, Deployments, StatefulSets, Jobs), networking (Services, Ingress), config (ConfigMaps, Secrets), and storage (PersistentVolumes) are all declared as YAML and submitted to the API server. The scheduler places pods onto nodes; controllers run reconciliation loops that maintain the declared state forever. Add-ons (Helm charts, operators, service meshes like Istio or Linkerd) layer higher-level abstractions on top. Managed services (GKE, EKS, AKS, DigitalOcean Kubernetes) handle the control plane so you only run worker nodes.
Advertisement
Why it matters
K8s is the de facto API for running containers at scale — every cloud provider has a managed offering, every CI/CD pipeline has k8s deploy targets, every observability tool ships with k8s integration. The cost is real: complexity. A small team running a single app on Cloud Run, Fly, Render, or Railway will be more productive than the same team learning Helm + Istio + ArgoCD just to deploy a Next.js site. Reach for k8s when you actually need its power: many services, many environments, fine-grained scaling, regulatory requirements that managed PaaS can't meet.
Frequently asked questions
Do I need Kubernetes?
Probably not. If your workload fits Vercel, Cloud Run, Fly, or Railway, those are simpler to operate. Kubernetes pays off when you have many services, dedicated platform engineers, or specific compliance requirements.
What's a pod?
The smallest deployable unit — one or more containers that share network/storage and are scheduled together. Most apps run one container per pod; sidecars (logging, mesh proxies) ride alongside.
kubectl vs Helm vs Kustomize?
kubectl applies raw YAML. Helm templates YAML with values files (package manager). Kustomize patches YAML with overlays (no templating). Most teams use either Helm or Kustomize, not both.
Related terms
- DefinitionDockerDocker packages an app plus its OS-level dependencies into a portable container — a lightweight, isolated process that runs the same way on every host. The image is the recipe; the container is a running instance.
- DefinitionCI/CDCI/CD is the automation that turns 'I pushed code' into 'it's tested, built, and deployed'. CI (continuous integration) runs tests + builds on every commit; CD (continuous delivery / deployment) ships the result to staging or production automatically.