Infrastructure & Operations › Kubernetes & Orchestration

kubectl

The command-line tool for Kubernetes.

Also known as: kubectl commands, kubernetes cli, k8s cli

kubectl is the command-line client for Kubernetes. It talks to the cluster’s API server using a kubeconfig, which lists clusters, users and contexts, and tells kubectl which cluster and identity to use. Almost everything you do with a cluster starts here.

kubectl get pods -n payments          # list pods in a namespace
kubectl describe pod api-7d9f          # details, events, and status
kubectl logs api-7d9f --previous       # logs from the last attempt
kubectl exec -it api-7d9f -- sh        # open a shell in a container
kubectl apply -f deployment.yaml        # create or update from a file
kubectl get events --sort-by=.lastTimestamp

Declarative files with apply are the normal path; imperative commands like create and scale are handy for quick changes and experiments. Add -o wide or -o yaml to see more, and -n <namespace> to target one — many confusing “not found” errors are just the wrong namespace or context.

The classic mistakes:

• Being on the wrong context. kubeconfig can hold production and development side by side. Always know where you are; many people set the context to include the cluster name, and use separate configs per environment.
• kubectl edit and ad hoc apply in production. Changes made by hand drift from source control. Manage manifests as infrastructure as code — often via Helm — and use kubectl to inspect.
• Deleting without looking. kubectl delete is immediate. Preview with --dry-run=client and confirm the object and namespace.

For log-hunting a failing container, pair kubectl with container logs; describe and get events are usually where the real cause shows up.