Telusko Docs

Use namespaces to divide one cluster into isolated areas, create them with kubectl and YAML, deploy a pod into a namespace, query it with -n, and understand what deleting a namespace removes.

Every time you ran kubectl get pods on an empty cluster, the answer was No resources found in default namespace. That word, namespace, has been there from the start. Now it is time to use it.

The problem namespaces solve

Picture one company cluster shared by several teams: two development teams, a testing team, and production workloads. Each team wants a pod called webapp. Inside a single area that is impossible, because object names must be unique per kind within a namespace. Without some way to divide the cluster, teams would trip over each other's names, accidentally delete each other's pods, and nobody could tell which resources belong to whom.

A namespace is a logical partition inside one Kubernetes cluster. It groups and isolates resources. Think of it as a folder: dev, test and prod folders can each contain a file named webapp without any conflict.

cluster telusko-learning
 |- namespace default   : pods, services you created so far
 |- namespace dev       : webapp pod (dev team)
 |- namespace test      : webapp pod (test team)
 '- namespace kube-system : coredns, kube-proxy, aws-node ...

Namespaces that exist already

kubectl get namespaces      # short form: kubectl get ns

NAME              STATUS   AGE
default           Active   3h
kube-node-lease   Active   3h
kube-public       Active   3h
kube-system       Active   3h

Namespace	Purpose
`default`	Where your objects go when you do not specify a namespace. Everything we created so far lives here
`kube-system`	System components: CoreDNS, kube-proxy, the VPC CNI (`aws-node`), metrics-server. Do not delete or edit things here unless you know exactly why
`kube-public`	Readable by everyone, including unauthenticated users. Holds a small amount of public cluster info
`kube-node-lease`	One Lease object per node. The kubelet updates it as a heartbeat so the control plane knows the node is alive

Creating a namespace

With a command

kubectl create namespace dev
kubectl get ns

With a manifest

vi ns.yml

ns.yml

apiVersion: v1
kind: Namespace
metadata:
  name: alien

A Namespace needs no spec. Apply it:

kubectl apply -f ns.yml
kubectl get ns

Both dev (created by command) and alien (created from YAML) now appear with status Active. The YAML way is better for real projects, because the namespace is then stored in Git with the rest of the application.

Namespace names follow DNS label rules: lowercase letters, numbers and -, at most 63 characters. Avoid names starting with kube-, which is reserved for the system.

Deleting a namespace

kubectl delete namespace alien

Deleting a namespace deletes everything inside it

Every pod, Service, Deployment, ConfigMap, Secret and PersistentVolumeClaim in that namespace is deleted with it, and there is no undo. Before deleting, run kubectl get all -n <name> and make sure that is what you want. On a shared cluster, deleting the wrong namespace is one of the most damaging mistakes you can make.

Deletion can stay in Terminating status for a while as Kubernetes deletes the contents first.

Creating a pod in a namespace

Edit the pod manifest from earlier and add namespace under metadata:

k8s-pod.yml

apiVersion: v1
kind: Pod
metadata:
  name: webapp
  namespace: dev
  labels:
    app: telusko-webapp
spec:
  containers:
    - name: webapp-container
      image: <your-dockerhub-username>/springboot-app:latest
      ports:
        - containerPort: 8080

kubectl apply -f k8s-pod.yml
kubectl get pods

No resources found in default namespace.

The pod is not missing. kubectl looks in default unless told otherwise. Ask for dev:

kubectl get pods -n dev

NAME     READY   STATUS    RESTARTS   AGE
webapp   1/1     Running   0          12s

Option	Meaning
`-n dev` or `--namespace dev`	Run this command against the `dev` namespace
`-A` or `--all-namespaces`	List objects from every namespace, for example `kubectl get pods -A`

You can also leave namespace out of the YAML and choose it at apply time with kubectl apply -f k8s-pod.yml -n dev. If the file says one namespace and -n says another, kubectl refuses with an error, which protects you from deploying to the wrong place.

Stop typing -n every time

If you work in one namespace for a while, make it the default for your kubectl context:

kubectl config set-context --current --namespace=dev
kubectl config view --minify | grep namespace     # confirm
kubectl config set-context --current --namespace=default   # switch back

Now delete the namespace and watch the pod go

kubectl delete namespace dev
kubectl get pods -n dev

No resources found in dev namespace.

The pod went with the namespace, exactly as the warning said.

What namespaces do and do not isolate

Namespaces are about organisation and access control, not network walls.

Isolated by namespace	Not isolated by namespace
Object names (two `webapp` pods can exist in two namespaces)	Network traffic. A pod in `dev` can call a Service in `prod` using `svc-name.prod` unless a NetworkPolicy blocks it
RBAC permissions (give a team rights to only their namespace)	Nodes. Pods from all namespaces share the same worker nodes
Resource limits per team (ResourceQuota, covered later in the course)	Cluster-scoped objects such as Nodes, Namespaces themselves, PersistentVolumes, StorageClasses

To list which kinds live inside namespaces and which are cluster-wide:

kubectl api-resources --namespaced=true
kubectl api-resources --namespaced=false

Try It Yourself

Create namespaces dev and test, one with a command and one with YAML.
Create a pod named webapp in both namespaces at the same time. Confirm with kubectl get pods -A that two pods share the same name.
Set your current context's namespace to test, run kubectl get pods without -n, then switch back to default.
Delete the test namespace and confirm its pod disappeared.

Official references

Written By: Shramik Masti

Namespaces in Kubernetes