Pacman on the Cloud: how to deploy a Kubernetes application on zCloud using a fully integrated EKS distribution

Written by Osmar R Leão

SA @Zadara

This article is the second part of my first (How to Install and Use an EKS Distribution on the Zadara Cloud) showing how to deploy an application and how integrated the EKS is with Zadara’s Cloud Orchestrator.

My first step was finding one application that uses data persistence (storage) and a load balancer. I found one that uses all of these and is also fun: Pacman! This app is part of one GitHub repository of Kubernetes examples: https://github.com/font/k8s-example-apps.

In this repository, you can find how to download, build, and push your images of this application to one docker container registry.

The Node.js Pacman uses MongoDB to keep the game’s high scores and user data, one internal load balancer to be the MongoDB frontend, and another load balancer to be the frontend of the Pacman game itself. For this example, I will use MongoDB and 2 Pacman pods.

The First step is to create a namespace to be used:

$ kubectl create namespace pacman

Now, create the MongoDB PVC (using 8GB of storage):

kind: PersistentVolumeClaimapiVersion: v1metadata:  name: mongo-storagespec:  accessModes:    - ReadWriteOnce  resources:    requests:      storage: 8Gi

$ kubectl -n pacman apply -f mongo-pvc-new.yamlpersistentvolumeclaim/mongo-storage created

The MongoDB service (load balancer):

apiVersion: v1kind: Servicemetadata:  labels:    name: mongo  name: mongospec:  type: LoadBalancer  ports:    - port: 27017      targetPort: 27017  selector:    name: mongo

$ kubectl -n pacman apply -f ../services/mongo-service.yamlservice/mongo created

And the deployment of the MongoDB:

apiVersion: apps/v1kind: Deploymentmetadata:  labels:    name: mongo  name: mongospec:  replicas: 1  selector:    matchLabels:      name: mongo  template:    metadata:      labels:        name: mongo    spec:      containers:      - image: mongo        name: mongo        ports:        - name: mongo          containerPort: 27017        volumeMounts:          - name: mongo-db            mountPath: /data/db      volumes:        - name: mongo-db          persistentVolumeClaim:            claimName: mongo-storage

Verify that the PV is mounted as /data/db inside the MongoDB pods.

$ kubectl -n pacman apply -f mongo-deployment-new.yamldeployment.apps/mongo created

Now, checking the EKS integration with Zadara zCompute, storage:

$ kubectl -n pacman get pvNAME                                       CAPACITY   ACCESS MODES   RECLAIM POLICY   STATUS   CLAIM                  STORAGECLASS   VOLUMEATTRIBUTESCLASS   REASON   AGEpvc-d3a8b520-af36-41a1-95e0-d32cc071b6a1   8Gi        RWO            Delete           Bound    pacman/mongo-storage   ebs-sc         <unset>                          5h9m

$ kubectl -n pacman describe pv pvc-d3a8b520-af36-41a1-95e0-d32cc071b6a1Name:              pvc-d3a8b520-af36-41a1-95e0-d32cc071b6a1Labels:            <none>Annotations:       pv.kubernetes.io/provisioned-by: ebs.csi.aws.comvolume.kubernetes.io/provisioner-deletion-secret-name:volume.kubernetes.io/provisioner-deletion-secret-namespace:Finalizers:        [kubernetes.io/pv-protection external-attacher/ebs-csi-aws-com]StorageClass:      ebs-scStatus:            BoundClaim:             pacman/mongo-storageReclaim Policy:    DeleteAccess Modes:      RWOVolumeMode:        FilesystemCapacity:          8GiNode Affinity:Required Terms:Term 0:        topology.ebs.csi.aws.com/zone in [symphony]Message:Source:Type:              CSI (a Container Storage Interface (CSI) volume source)Driver:            ebs.csi.aws.comFSType:            ext4VolumeHandle:      vol-30ec110a21c94a929be9b76988c43ffdReadOnly:          falseVolumeAttributes:      storage.kubernetes.io/csiProvisionerIdentity=1732818873883-5119-ebs.csi.aws.comEvents:                <none>

Showing the volume at the zCompute GUI:

And the load balancer:

$ kubectl -n pacman get svcNAME     TYPE           CLUSTER-IP     EXTERNAL-IP                                                             PORT(S)           AGEmongo    LoadBalancer   10.102.59.17   elb-ce73483f-d64b-4f9f-aac3-bd5f0e45018e.zadara.demo                    27017:30376/TCP   5h22m

And now, the Pacman part! First the service (internet facing):

apiVersion: v1kind: Servicemetadata:  name: pacman  labels:    name: pacman  annotations:    service.beta.kubernetes.io/aws-load-balancer-scheme: internet-facingspec:  type: LoadBalancer  ports:    - port: 80      targetPort: 8080      protocol: TCP  selector:    name: pacman

The annotations part of this YAML is needed to signal to the Zadara Cloud Orchestrator to create a public load balancer.

$ kubectl -n pacman apply -f pacman-service.yamlservice/pacman created

It is possible to check the load balancer created in the Kubernetes and Zadara’s GUI:

$ kubectl -n pacman get svcNAME     TYPE           CLUSTER-IP     EXTERNAL-IP                                                             PORT(S)           AGEmongo    LoadBalancer   10.102.59.17   elb-ce73483f-d64b-4f9f-aac3-bd5f0e45018e.zadara.demo                    27017:30376/TCP   8m20spacman   LoadBalancer   10.110.35.73   elb-3a3bfac2-7c58-46f6-87fc-0a5bdf1ebb92.elb.services.symphony.public   80:30510/TCP      2m11s

Now, the deployment of the Pacman app itself:

apiVersion: apps/v1kind: Deploymentmetadata:  labels:    name: pacman  name: pacmanspec:  replicas: 2  selector:    matchLabels:        name: pacman  template:    metadata:      labels:        name: pacman    spec:      containers:      - image: zosmarleao/pacman-nodejs-app:latest        name: pacman        ports:        - containerPort: 8080          name: http-server

The image is located on my account in the docker container registry. I had already uploaded into it.

Checking the pods:

$ kubectl -n pacman get podsNAME                      READY   STATUS             RESTARTS          AGEmongo-57695f7544-5zl6n    1/1     Running            0                 21hpacman-5555f97879-47b24   1/1     Running            0                 21hpacman-5555f97879-wddjc   1/1     Running            0                 21h

Refreshing the Internet Browser, the load balancer will change the backend pod: