Save the Date - KBE Insider with Kaslin Fields on December 12th

Namespaces

Namespaces provide a scope for Kubernetes resources, carving up your cluster in smaller units. You can think of it as a workspace you're sharing with other users. Many resources such as pods and services are namespaced. Others, such as nodes, are not namespaced, but are instead treated as cluster-wide. As a developer, you'll usually use an assigned namespace, however admins may wish to manage them, for example to set up access control or resource quotas.

Like other resources, the get subcommand displays a list of all namespaces a user has access to in a cluster (both the full resource type name namespace and the abbreviation ns can be used):

kubectl get ns

On a simple minikube installation, the result shows:

NAME              STATUS   AGE
default           Active   17h
kube-node-lease   Active   17h
kube-public       Active   17h
kube-system       Active   17h

You can learn more about a namespace using the describe verb:

kubectl describe ns default

If no changes were made to the minikube cluster, the output should look like the following:

Name:         default
Labels:       <none>
Annotations:  <none>
Status:       Active

No resource quota.

No LimitRange resource.
Let's now create a new namespace called "test":
kubectl apply -f https://raw.githubusercontent.com/openshift-evangelists/kbe/main/specs/ns/ns.yaml

Once the namespace is created, it will appear in the list of available namespaces:

kubectl get ns
Alternatively, we could have created the namespace using the kubectl create namespace test command.
 
To launch a pod in the newly created namespace test, run:
kubectl apply --namespace=test -f https://raw.githubusercontent.com/openshift-evangelists/kbe/main/specs/ns/pod.yaml

Note that using above method the namespace becomes a runtime property. In other words, you can deploy the same pod or service into multiple namespaces (for example, dev and prod). Hard-coding the namespace directly in the metadata section as shown in the following is possible, but causes less flexibility when deploying your apps:

apiVersion: v1
kind: Pod
metadata:
name: podintest
namespace: test

To list namespaced objects, such as our pod podintest, pass the --namespace variable to the get call:

kubectl get pods --namespace=test

You can remove the namespace (and everything inside of it) with:

kubectl delete ns test

If you're an admin, you might want to check out the docs for more info how to handle namespaces.