Deploy a stateful workload with Filestore

Stay organized with collections Save and categorize content based on your preferences.

This tutorial shows how to deploy a simple reader/writer stateful workload using a Persistent Volume (PV) and a Persistent Volume Claim (PVC) on Google Kubernetes Engine (GKE). Follow this tutorial to learn how to design for scalability using Filestore, Google Cloud's managed network filesystem.

Background

By nature, Pods are ephemeral. This means that GKE destroys the state and value stored in a Pod when it is deleted, evicted, or rescheduled.

As an application operator, you may want to maintain stateful workloads. Examples of such workloads include applications that process WordPress articles, messaging apps, and apps that process machine learning operations.

By using Filestore on GKE, you can perform the following operations:

  • Deploy stateful workloads that are scalable.
  • Enable multiple Pods to have ReadWriteMany as its accessMode, so that multiple Pods can read and write at the same time to the same storage.
  • Set up GKE to mount volumes into multiple Pods simultaneously.
  • Persist storage when Pods are removed.
  • Enable Pods to share data and easily scale.

Objectives

This tutorial is for application operators and other users that want to set up a scalable stateful workload on GKE using PVC and NFS.

Stateful workload GKE diagram

This tutorial covers the following steps:

  1. Create a GKE cluster.
  2. Configure the managed file storage with Filestore using CSI.
  3. Create a reader and a writer Pod.
  4. Expose and access the reader Pod to a Service Load Balancer.
  5. Scale up the writer.
  6. Access data from the writer Pod.

Costs

This tutorial uses the following billable components of Google Cloud:

Use the Pricing Calculator to generate a cost estimate based on your projected usage.

When you finish this tutorial, you can avoid continued billing by deleting the resources you created. For more information, see Clean up.


For step-by-step guidance for this task directly in the Google Cloud console, click Guide me:

Guide me


The following sections take you through the same steps as clicking Guide me.

Before you begin

Set up your project

  1. Sign in to your Google Cloud account. If you're new to Google Cloud, create an account to evaluate how our products perform in real-world scenarios. New customers also get $300 in free credits to run, test, and deploy workloads.
  2. In the Google Cloud console, on the project selector page, click Create project to begin creating a new Google Cloud project.

    Go to project selector

  3. Make sure that billing is enabled for your Cloud project. Learn how to check if billing is enabled on a project.

  4. Enable the Compute Engine, GKE, and Filestore APIs.

    Enable the APIs

  5. In the Google Cloud console, on the project selector page, click Create project to begin creating a new Google Cloud project.

    Go to project selector

  6. Make sure that billing is enabled for your Cloud project. Learn how to check if billing is enabled on a project.

  7. Enable the Compute Engine, GKE, and Filestore APIs.

    Enable the APIs

Set defaults for the Google Cloud CLI

  1. In the Google Cloud console, start a Cloud Shell instance:
    Open Cloud Shell

  2. Download the source code for this sample app:

    git clone https://github.com/GoogleCloudPlatform/kubernetes-engine-samples
    cd kubernetes-engine-samples/stateful-workload-filestore
    
  3. Set the default environment variables:

    gcloud config set project PROJECT_ID
    gcloud config set compute/region COMPUTE_REGION
    gcloud config set compute/zone COMPUTE_ZONE
    gcloud config set filestore/zone COMPUTE_ZONE
    gcloud config set filestore/region COMPUTE_REGION
    

    Replace the following values:

Create a GKE cluster

  1. Create a GKE cluster named stateful-cluster:

    gcloud container clusters create-auto stateful-cluster --region COMPUTE_REGION
    

    The outcome is similar to the following once the cluster is created:

      gcloud container clusters describe stateful-cluster
      NAME: stateful-cluster
      LOCATION: northamerica-northeast2
      MASTER_VERSION: 1.21.11-gke.1100
      MASTER_IP: 34.130.255.70
      MACHINE_TYPE: e2-medium
      NODE_VERSION: 1.21.11-gke.1100
      NUM_NODES: 3
      STATUS: RUNNING
    

    Where the STATUS is RUNNING for the stateful-cluster.

Configure the managed file storage with Filestore using CSI

GKE provides a way to automatically deploy and manage the Kubernetes Filestore CSI driver in your clusters. Using Filestore CSI allows you to dynamically create or delete Filestore instances and use them in Kubernetes workloads with a StorageClass or a Deployment.

You can create a new Filestore instance by creating a PVC that dynamically provisions a Filestore instance and the PV, or access pre-provisioned Filestore instances in Kubernetes workloads.

New instance

Create the Storage Class

apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: filestore-sc
provisioner: filestore.csi.storage.gke.io
volumeBindingMode: Immediate
allowVolumeExpansion: true
parameters:
  tier: standard
  network: default
  • volumeBindingMode is set to Immediate, which allows the provisioning of the volume to begin immediately.
  • tier is set to standard for faster Filestore instance creation time. If you need higher available NFS storage, snapshots for data backup, data replication over multiple zones and other enterprise level features, set tier to enterprise instead. Note: The reclaim policy for dynamically created PV defaults to Delete if the reclaimPolicy in the StorageClass is not set.
  1. Create the StorageClass resource:

    kubectl create -f filestore-storageclass.yaml
    
  2. Verify that the Storage Class is created:

    kubectl get sc
    

    The output is similar to the following:

    NAME                     PROVISIONER                    RECLAIMPOLICY   VOLUMEBINDINGMODE      ALLOWVOLUMEEXPANSION   AGE
    filestore-sc             filestore.csi.storage.gke.io   Delete          Immediate              true                   94m
    

Pre-provisioned instance

Create the Storage Class

apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: filestore-sc
provisioner: filestore.csi.storage.gke.io
volumeBindingMode: Immediate
allowVolumeExpansion: true

When volumeBindingMode is set to Immediate, it allows the provisioning of the volume to begin immediately.

  1. Create the StorageClass resource:

      kubectl create -f preprov-storageclass.yaml
    
  2. Verify that the Storage Class is created:

      kubectl get sc
    

    The output is similar to the following:

      NAME                     PROVISIONER                    RECLAIMPOLICY   VOLUMEBINDINGMODE      ALLOWVOLUMEEXPANSION   AGE
      filestore-sc             filestore.csi.storage.gke.io   Delete          Immediate              true                   94m
    

Create a Persistent Volume for the Filestore instance

apiVersion: v1
kind: PersistentVolume
metadata:
  name: fileserver
  annotations:
    pv.kubernetes.io/provisioned-by: filestore.csi.storage.gke.io
spec:
  storageClassName: filestore-sc
  capacity:
    storage: 1Ti
  accessModes:
    - ReadWriteMany
  persistentVolumeReclaimPolicy: Delete
  volumeMode: Filesystem
  csi:
    driver: filestore.csi.storage.gke.io
    # Modify this to use the zone, filestore instance and share name.
    volumeHandle: "modeInstance/<FILESTORE_ZONE>/<INSTANCE_NAME>/<FILESTORE_SHARE_NAME>"
    volumeAttributes:
      ip: <IP_ADDRESS> # Modify this to Pre-provisioned Filestore instance IP
      volume: <FILESTORE_SHARE_NAME> # Modify this to Pre-provisioned Filestore instance share name
  1. Verify that the pre-existing Filestore instance is ready:

      gcloud filestore instances list
    

    The output is similar to the following, where the STATE value is READY:

      INSTANCE_NAME: stateful-filestore
      LOCATION: us-central1-a
      TIER: ENTERPRISE
      CAPACITY_GB: 1024
      FILE_SHARE_NAME: statefulpath
      IP_ADDRESS: 10.109.38.98
      STATE: READY
      CREATE_TIME: 2022-04-05T18:58:28
    

    Note the INSTANCE_NAME, LOCATION, FILE_SHARE_NAME, and IP_ADDRESS of the Filestore instance.

  2. Populate the Filestore instance console variables:

      INSTANCE_NAME=INSTANCE_NAME
      LOCATION=LOCATION
      FILE_SHARE_NAME=FILE_SHARE_NAME
      IP_ADDRESS=IP_ADDRESS
    
  3. Replace the placeholder variables with the console variables obtained above to the file preprov-pv.yaml:

      sed "s/<INSTANCE_NAME>/$INSTANCE_NAME/" preprov-pv.yaml > changed.yaml && mv changed.yaml preprov-pv.yaml
      sed "s/<LOCATION>/$LOCATION/" preprov-pv.yaml > changed.yaml && mv changed.yaml preprov-pv.yaml
      sed "s/<FILE_SHARE_NAME>/$FILE_SHARE_NAME/" preprov-pv.yaml > changed.yaml && mv changed.yaml preprov-pv.yaml
      sed "s/<IP_ADDRESS>/$IP_ADDRESS/" preprov-pv.yaml > changed.yaml && mv changed.yaml preprov-pv.yaml
    
  4. Create the PV

      kubectl apply -f preprov-pv.yaml
    
  5. Verify that the PV's STATUS is set to Bound:

      kubectl get pv
    

    The output is similar to the following:

      NAME        CAPACITY   ACCESS MODES   RECLAIM POLICY   STATUS   CLAIM                STORAGECLASS    REASON   AGE
      fileserver  1Ti        RWX            Delete           Bound    default/fileserver   filestore-sc             46m
    

Use a PersistentVolumeClaim to access the volume

The following pvc.yaml manifest references the Filestore CSI driver's StorageClass named filestore-sc.

In order to have multiple Pods reading and writing to the volume, the accessMode is set to ReadWriteMany.

kind: PersistentVolumeClaim
apiVersion: v1
metadata:
  name: fileserver
spec:
  accessModes:
  - ReadWriteMany
  storageClassName: filestore-sc
  resources:
    requests:
      storage: 1Ti
  1. Deploy the PVC:

    kubectl create -f pvc.yaml
    
  2. Verify that the PVC is created:

    kubectl get pvc
    

    The output is similar to the following:

    NAME         STATUS   VOLUME                                     CAPACITY   ACCESS MODES   STORAGECLASS        AGE
    fileserver   Bound    pvc-aadc7546-78dd-4f12-a909-7f02aaedf0c3   1Ti        RWX            filestore-sc        92m
    
  3. Verify that the newly created Filestore instance is ready:

    gcloud filestore instances list
    

    The output is similar to the following:

    INSTANCE_NAME: pvc-5bc55493-9e58-4ca5-8cd2-0739e0a7b68c
    LOCATION: northamerica-northeast2-a
    TIER: STANDARD
    CAPACITY_GB: 1024
    FILE_SHARE_NAME: vol1
    IP_ADDRESS: 10.29.174.90
    STATE: READY
    CREATE_TIME: 2022-06-24T18:29:19
    

Create a reader and a writer Pod

Create the reader Pod

The reader Pod will read the file that is being written by the writers Pods. The reader Pods will see what time and which writer Pod replica wrote to the file.

apiVersion: apps/v1
kind: Deployment
metadata:
  name: reader
spec:
  replicas: 1
  selector:
    matchLabels:
      app: reader
  template:
    metadata:
      labels:
        app: reader
    spec:
      containers:
      - name: nginx
        image: nginx:stable-alpine
        ports:
        - containerPort: 80
        volumeMounts:
        - name: fileserver
          mountPath: /usr/share/nginx/html # the shared directory 
          readOnly: true
      volumes:
      - name: fileserver
        persistentVolumeClaim:
          claimName: fileserver

The reader Pod will read from the path /usr/share/nginx/html which is shared between all the Pods.

  1. Deploy the reader Pod:

    kubectl apply -f reader-fs.yaml
    
  2. Verify that the reader replicas are running by querying the list of Pods:

    kubectl get pods
    

    The output is similar to the following:

    NAME                      READY   STATUS    RESTARTS   AGE
    reader-66b8fff8fd-jb9p4   1/1     Running   0          3m30s
    

Create the writer Pod

The writer Pod will periodically write to a shared file that other writer and reader Pods can access. The writer Pod records its presence by writing its host name to the shared file.

The image used for the writer Pod is a custom image of Alpine Linux, which is used for utilities and production applications. It includes a script indexInfo.html that will obtain the metadata of the most recent writer, and keep count of all the unique writers and total writes.

apiVersion: apps/v1
kind: Deployment
metadata:
  name: writer
spec:
  replicas: 2 # start with 2 replicas
  selector:
    matchLabels:
      app: writer
  template:
    metadata:
      labels:
        app: writer
    spec:
      containers:
      - name: content
        image: us-docker.pkg.dev/google-samples/containers/gke/stateful-workload:latest
        volumeMounts:
        - name: fileserver
          mountPath: /html # the shared directory
        command: ["/bin/sh", "-c"]
        args:
        - cp /htmlTemp/indexInfo.html /html/index.html;
          while true; do
          echo "<b> Date :</b> <text>$(date)</text> <b> Writer :</b> <text2> ${HOSTNAME} </text2> <br>  " >> /html/indexData.html;
          sleep 30;  
          done
      volumes:
      - name: fileserver
        persistentVolumeClaim:
          claimName: fileserver

For this tutorial, the writer Pod writes every 30 seconds to the path /html/index.html. Modify the sleep number value to have a different write frequency.

  1. Deploy the writer Pod:

    kubectl apply -f writer-fs.yaml
    
  2. Verify that the writer Pods are running by querying the list of Pods:

    kubectl get pods
    

    The output is similar to the following:

    NAME                      READY   STATUS    RESTARTS   AGE
    reader-66b8fff8fd-jb9p4   1/1     Running   0          3m30s
    writer-855565fbc6-8gh2k   1/1     Running   0          2m31s
    writer-855565fbc6-lls4r   1/1     Running   0          2m31s
    

Expose and access the reader workload to a Service Load Balancer

To expose a workload outside the cluster, create a Service of type LoadBalancer. This type of Service creates an external load balancer with an IP address reachable through the internet.

  1. Create a Service of type LoadBalancer named reader-lb:

    kubectl create -f loadbalancer.yaml
    
  2. Watch the deployment to see that GKE assigns an EXTERNAL-IP for reader-lb Service:

    kubectl get svc --watch
    

    When the Service is ready, the EXTERNAL-IP column displays the public IP address of the load balancer:

      NAME         TYPE           CLUSTER-IP    EXTERNAL-IP     PORT(S)        AGE
      kubernetes   ClusterIP      10.8.128.1    <none>          443/TCP        2d21h
      reader-lb    LoadBalancer   10.8.131.79   34.71.232.122   80:32672/TCP   2d20h
    
  3. Press Ctrl+C to terminate the watch process.

  4. Use a web browser to navigate to the EXTERNAL-IP assigned to the load balancer. The page refreshes every 30 seconds. The more writers Pods and shorter the frequency, the more entries it will show.

To see more details about the load balancer service, see loadbalancer.yaml.

Scale up the writer

Because the PV accessMode was set to ReadWriteMany, GKE can scale up the number of Pods so that more writer Pods can write to this shared volume (or more readers can read to read them).

  1. Scale up the writer to five replicas:

    kubectl scale deployment writer --replicas=5
    

    The output is similar to the following:

    deployment.extensions/writer scaled
    
  2. Verify the number of running replicas:

    kubectl get pods
    

    The output is similar to the following:

    NAME                      READY   STATUS    RESTARTS   AGE
    reader-66b8fff8fd-jb9p4   1/1     Running   0          11m
    writer-855565fbc6-8dfkj   1/1     Running   0          4m
    writer-855565fbc6-8gh2k   1/1     Running   0          10m
    writer-855565fbc6-gv5rs   1/1     Running   0          4m
    writer-855565fbc6-lls4r   1/1     Running   0          10m
    writer-855565fbc6-tqwxc   1/1     Running   0          4m
    
  3. Use a web browser to navigate again to the EXTERNAL-IP assigned to the load balancer.

At this point, you configured and scaled your cluster to support five stateful writer Pods. Where multiple writer Pods are writing to the same file simultaneously. The reader Pods can also be easily scaled up.

Optional: Access data from the writer Pod

This section demonstrates how to use a command-line interface to access a reader or writer Pod. You can see the shared component that the writer is writing to and the reader is reading from.

  1. Obtain the writer Pod name:

    kubectl get pods
    

    The output is similar to the following:

    NAME                      READY   STATUS    RESTARTS   AGE
    writer-5465d65b46-7hxv4   1/1     Running   0          20d
    

    Note the hostname of a writer Pod (Example: writer-5465d65b46-7hxv4).

  2. Run the following command to access the writer Pod:

    kubectl exec -it WRITER_HOSTNAME -- /bin/sh
    
  3. See the shared component in the file indexData.html:

    cd /html
    cat indexData.html
    
  4. Clear the indexData.html file:

    echo '' > indexData.html
    

    Refresh the web browser hosting the EXTERNAL-IP address to see the change.

  5. Exit the environment:

    exit
    

Clean up

To avoid incurring charges to your Google Cloud account for the resources used in this tutorial, either delete the project that contains the resources, or keep the project and delete the individual resources.

Delete the project

  1. In the Google Cloud console, go to the Manage resources page.

    Go to Manage resources

  2. In the project list, select the project that you want to delete, and then click Delete.
  3. In the dialog, type the project ID, and then click Shut down to delete the project.

Delete the individual resources

  1. Delete the load balancer Service:

    kubectl delete service reader-lb
    

    Wait until the load balancer provisioned for the reader service is deleted

  2. Verify the list returns Listed 0 items:

    gcloud compute forwarding-rules list
    
  3. Delete the Deployments

    kubectl delete deployment writer
    kubectl delete deployment reader
    
  4. Verify the Pods are deleted and returns No resources found in default namespace.

    kubectl get pods
    
  5. Delete the PVC. This will also delete the PV and the Filestore instance due to the retention policy set to delete

    kubectl delete pvc fileserver
    
  6. Delete the GKE cluster:

    gcloud container clusters delete stateful-cluster --zone=COMPUTE_ZONE
    

    This deletes the resources that make up the GKE cluster, including the reader and writer Pods.

What's next