K8s Selectors and Labels

January 15, 2022  -   3 mins read time -   514 words -  garrardkitchen

Right, what’s the deal with all the labels and metadata in a Deployment manifest?!!!!

Take this example:

apiVersion: apps/v1
kind: Deployment
metadata:
  namespace: default
  name: nginx-deployment
  labels:
    app: nginx
spec:
  replicas: 3
  selector:
    matchLabels:
      app: nginx
      foo: baa
  template:
    metadata:
      labels:
        app: nginx
        foo: baa
    spec:
      containers:
      - name: nginx
        image: nginx:1.14.2
        ports:
        - containerPort: 80

Here, we see metadata twice, and also there’s mention of matchLabels in selector??? What does it all mean???

Ok, let me explain 👀 …

The first metadata reference

A deployment manifest kind is a manifest that describes the desired state of your application(s). I say applications here as a POD can contain more than one container (application). The desired part of this is found in a ReplicaSet kind manifest. For example, you’d use a ReplicaSet if you require to have 2 replicas (instances) of your POD running. A Deployment manifest is a short-hand way of stipulating this, ergo, saves you having to create 2 separate manifests. Makes sense? Good.

Behind the scenes, it is the Deployment Controller that monitors your deployment’s desired state and if it differs, it will return to it’s desired state.

So why is there two mentions of metadata? Ok, The first reference identifies this Deployment object itself:

apiVersion: apps/v1
kind: Deployment
metadata:
  namespace: default
  name: nginx-deployment
  labels:
    app: nginx

So for example, if you want to delete this object, you’d issue either of these kubectl commands:

kubectl delete deployments nginx-deployment -n default
OR
kubectl delete deployments -l app=nginx-deployment -n default

The latter delete example above uses an equality-based label selector condition. There’s also a set-based label selector condition. You can read about these here ➡ labels and selectors

The second metadata reference

The second metadata reference is inside the template that is being used to describe the POD to be created AND is separate from the Deployment itself.

...
template:
  metadata:
    labels:
      app: nginx
      foo: baa
  spec:
    containers:
    - name: nginx
    image: nginx:1.14.2
    ports:
    - containerPort: 80

To demonstrate this, let’s run the following command:

❯ kubectl.exe get pods -l app=nginx -n default --show-labels
NAME                                READY   STATUS         RESTARTS   AGE   LABELS
nginx-deployment-6494589cc9-242v8   1/1     Running        0          3s    app=nginx,foo=baa,pod-template-hash=6494589cc9   
nginx-deployment-6494589cc9-2fdf5   1/1     Running        0          3s    app=nginx,foo=baa,pod-template-hash=6494589cc9   
nginx-deployment-6494589cc9-n8vxb   0/1     ErrImagePull   0          3s    app=nginx,foo=baa,pod-template-hash=6494589cc9      

If you look at the LABELS column you can see labels that are not found in the Deployment metadata - eg foo=baa.

You can also use set-based selector; here’s an example that produces the same outcome:

❯ kubectl.exe get pods -l "app in (nginx)" -n default --show-labels
NAME                                READY   STATUS             RESTARTS   AGE     LABELS
nginx-deployment-6494589cc9-6sn7v   1/1     Running            0          6m10s   app=nginx,foo=baa,pod-template-hash=6494589cc9
nginx-deployment-6494589cc9-htxpx   0/1     ImagePullBackOff   0          6m10s   app=nginx,foo=baa,pod-template-hash=6494589cc9
nginx-deployment-6494589cc9-qh8nc   0/1     ErrImagePull       0          6m10s   app=nginx,foo=baa,pod-template-hash=6494589cc9

Binding deployment to pod

So, how do we couple the Deployment with the Pod? Well, this is where the selector comes into play. The selector instructs Kubernetes to match on the app label for those that have a value of nginx and that the foo label that has the value of baa.

spec:
  selector:
    matchLabels:
      app: nginx
      foo: baa

I hope this has made sense and has cleared up any confusion you may have had.