luc.run

Exercise

Check the taints of the cluster’s nodes. What does that mean ?
Create the specification of a Deployment named nginx with 6 replicas of Pods based on the nginx:1.20 image, then create the Deployment.
Where are the Pods scheduled ?
Change the specification of the Deployment so the Pods tolerate the taint present on the controlplane node. Update the Deployment resource.
Where are the Pods scheduled ?
Delete the Deployment

Documentation

https://kubernetes.io/docs/concepts/scheduling-eviction/taint-and-toleration/

Solution

Check the taints of the cluster’s nodes. What does that mean ?

Only the controlplane node has a Taint, this one prevents the application’s Pods from being scheduled on the controlplane.

k get nodes controlplane -o jsonpath={.spec.taints}
[{"effect":"NoSchedule","key":"node-role.kubernetes.io/control-plane"}]

The NoSchedule taints prevent Pods which do not tolerate the taints to be scheduled on that node

Create the specification of a Deployment named nginx with 6 replicas of Pods based on the nginx:1.20 image, then create the Deployment.

Specification:

k create deploy nginx --replicas 6 --image=nginx:1.20 --dry-run=client -o yaml > deploy.yaml

Creation of the Deployment

k apply -f deploy.yaml

Where are the Pods scheduled ?

k get po -o wide
NAME                     READY   STATUS    RESTARTS   AGE   IP          NODE      NOMINATED NODE   READINESS GATES
nginx-6d777db949-2kxhp   1/1     Running   0          8s    10.38.0.5   worker2   <none>           <none>
nginx-6d777db949-6d7ms   1/1     Running   0          8s    10.32.0.7   worker1   <none>           <none>
nginx-6d777db949-8vl6j   1/1     Running   0          8s    10.32.0.6   worker1   <none>           <none>
nginx-6d777db949-rqql4   1/1     Running   0          8s    10.38.0.6   worker2   <none>           <none>
nginx-6d777db949-tfbjb   1/1     Running   0          8s    10.32.0.2   worker1   <none>           <none>
nginx-6d777db949-vs6wl   1/1     Running   0          8s    10.38.0.1   worker2   <none>           <none>

The Pods are deployed either on worker1 or on worker2. None are deployed on the controlplane node because of the NoSchedule taints that the Pods do not tolerate.

Change the specification of the Deployment so the Pods tolerate the taints present on the controlplane node. Update the Deployment resource.

apiVersion: apps/v1
kind: Deployment
metadata:
  labels:
    app: nginx
  name: nginx
spec:
  replicas: 6
  selector:
    matchLabels:
      app: nginx
  template:
    metadata:
      labels:
        app: nginx
    spec:
      containers:
      - image: nginx:1.20
        name: nginx
      tolerations:
      - key: node-role.kubernetes.io/control-plane
        effect: NoSchedule

Update the resource:

k apply -f deploy.yaml

Where are the Pods scheduled ?

Due to the toleration of the taints, the Pods can now be scheduled on the controlplane node as well

NAME                     READY   STATUS    RESTARTS   AGE   IP          NODE            NOMINATED NODE   READINESS GATES
nginx-7b788fb97d-2vrv4   1/1     Running   0          59s   10.32.0.8   worker1         <none>           <none>
nginx-7b788fb97d-bjlvk   1/1     Running   0          57s   10.40.0.2   controlplane    <none>           <none>
nginx-7b788fb97d-gdn6x   1/1     Running   0          59s   10.38.0.1   worker2         <none>           <none>
nginx-7b788fb97d-qkp78   1/1     Running   0          59s   10.40.0.1   controlplane    <none>           <none>
nginx-7b788fb97d-rb74s   1/1     Running   0          57s   10.32.0.6   worker1         <none>           <none>
nginx-7b788fb97d-tj7f9   1/1     Running   0          57s   10.38.0.7   worker2         <none>           <none>

Delete the Deployment

k delete deploy/nginx