Kubernetes - StatefulSets

Prerequisites

• Kubernetes
• kubectl

Assignment

• Create a StatefulSet with the help of a Deployment file(No easy of creatinga statefulset from CLI):

kubectl create deployment nginx --image=nginx --replicas=3 --dry-run=client -o yaml | tee statefulset.yaml

• Modify the file to change it to a StatefulSet:

cat <<EOF > statefulset.yaml
apiVersion: apps/v1
kind: StatefulSet
metadata:
  creationTimestamp: null
  labels:
    app: nginx
  name: nginx
spec:
  replicas: 3
  selector:
    matchLabels:
      app: nginx
  serviceName: nginx
  template:
    metadata:
      labels:
        app: nginx
    spec:
      containers:
      - image: nginx
        name: nginx
EOF

kubectl apply -f statefulset.yaml

• Review the statefulset and the created pods:

kubectl get statefulset

kubectl get pods -o wide

• Note: If a pod in a statefulset was removed, causing a restart, whilst the ip may change, the name will stay consistent.

• Delete pod/nginx-2

kubectl delete pod/nginx-2 --now

kubectl get pods -o wide

• Using statefulsets service functionality, we’ll create a headless service called nginx:

kubectl create service clusterip --clusterip=None nginx

kubectl get service

• Review the our endpoint slices:

kubectl get endpoints

kubectl get endpoints/nginx -o yaml

• Let’s check this from the viewpoint of a pod, we’ll run a curlimages/curl container:

kubectl run --rm -i --tty curl --image=curlimages/curl:8.4.0 --restart=Never -- sh

• And let’s curl our statefulset service name -

curl nginx-1.nginx.default.svc.cluster.local

exit

• Take a look at the statefulset yaml, you’ll notice that the updateStrategy/rollingUpdate has a partition value of 0:

kubectl get statefulset/nginx -o yaml | more

• Edit the existing statefulset to change the partition value to 2 but, the following command will do the equivalent:

kubectl patch statefulset/nginx -p '{"spec":{"updateStrategy":{"rollingUpdate":{"partition":2}}}}'

• We will set the image to nginx:alpine and watch the rollout:

kubectl set image statefulset/nginx nginx=nginx:alpine && kubectl rollout status statefulset/nginx

• Because of the partition value, the rollout only took place on nginx-2, we can verify this with the following commands (check the image) -

kubectl describe pod/nginx-1 | more

kubectl describe pod/nginx-2 | more

StatefulSet Persistent Storage

• Update our statefulset to include a dynamic volume:

cat <<EOF > statefulset.yaml
apiVersion: apps/v1
kind: StatefulSet
metadata:
  labels:
    app: nginx
  name: nginx
spec:
  selector:
    matchLabels:
      app: nginx
  serviceName: nginx
  replicas: 3
  template:
    metadata:
      labels:
        app: nginx
    spec:
      containers:
      - image: nginx
        name: nginx
        volumeMounts:
        - name: nginx
          mountPath: /data
  volumeClaimTemplates:
  - metadata:
      name: nginx
    spec:
      accessModes: [ "ReadWriteOnce" ]
      storageClassName: "local-path"
      resources:
        requests:
          storage: 1Gi
EOF

• And we will recycle our statefulset:

kubectl delete -f statefulset.yaml && kubectl apply -f statefulset.yaml

• Watch our pods start, press ctrl-c to exit:

watch --differences kubectl get pods -o wide

• And we can verify that we have an individual pvc and pv for each pod -

kubectl get pvc

kubectl get pv

• We’ll execute a shell into nginx-0:

kubectl exec -it nginx-0 -- bash

• Create some data:

cd /data; touch this_is_nginx-0; ls -l

exit

• Let’s delete the pod and watch it be recreated:

kubectl delete pod/nginx-0 --now; watch --differences kubectl get pods -o wide

• And if we execute a shell again:

kubectl exec -it nginx-0 -- bash

• We can verify that our data still exists -

ls /data

exit

• And we can take this further, we can delete the entire statefulset:

kubectl delete -f statefulset.yaml

• Our pvc and pv will still exist:

kubectl get pvc; kubectl get pv

• And if we apply the statefulset again:

kubectl apply -f statefulset.yaml

• Check our pods are running:

kubectl get pods -o wide

• And exec back into the pod:

kubectl exec -it nginx-0 -- bash

• And check our data, we can confirm that it is available as expected:

ls /data

exit

• Cleanup:

kubectl delete statefulset/nginx --now; for i in 0 1 2; do kubectl delete pvc/nginx-nginx-$i --now; done; kubectl delete service/nginx; rm statefulset.yaml