Kubernetes Workloads: Pods, Deployments, Services, And Declarative State

**Kubernetes** orchestrates containerized workloads across a cluster. The smallest schedulable unit is a **Pod**: one or more containers that share network and storage namespaces. Pods are ephemeral; you do not “fix” a pod in place. You replace it.

A **Deployment** declares desired replica count and pod template. The Deployment controller creates a **ReplicaSet**, which maintains running pods matching the template. Rolling updates change the pod template and gradually replace instances. The API is declarative: you submit desired state; controllers reconcile.

A **Service** provides a stable virtual IP and DNS name for a set of pods, usually selected by labels. **ClusterIP** is internal; **LoadBalancer** and **NodePort** expose traffic depending on cloud integration. **Ingress** (or Gateway API resources) routes HTTP at L7.

Official references: `kubernetes.io/docs/concepts` and the API reference for your cluster version. Always match `kubectl` client minor to server policy your platform documents.