logo for NFS Provisioner Operator

A persistent volume (PV) is a common way to preserve data in case a container is accidentally lost in Kubernetes or Red Hat OpenShift. This article shows you how to manage persistent volumes with the NFS Provisioner Operator that I developed.

Difficulties with PVs

In the early days of container-based development, each user had to ask an administrator to create a PV for that user's containers. Usually the administrator created 100 PVs in advance when the cluster was installed. The administrator also had to clean up the used PVs when they were released. Obviously, this process was inefficient and really burdensome.

Dynamic provisioning using StorageClass was developed to solve this problem. With StorageClass, you no longer have to manually manage your PVs—a provisioner manages them for you. Sounds good, right?

But the next question is how to set up the StorageClass on the cluster without cost. If you can afford it, the easiest way is to use Red Hat OpenShift Dedicated, which provides the default gp2 StorageClass. But it is not free.

Let's say you want to play around with an OpenShift cluster installed on your laptop using Red Hat CodeReady Containers. The environment is absolutely free and under your control. Wouldn't it be great if this cluster had a StorageClass? With such an environment, you could test most scenarios without charge.

The NFS Provisioner Operator is open source and available on OperatorHub.io, which means that it can be easily installed via OpenShift's OperatorHub menu. The Operator uses the Kubernetes NFS subdir external provisioner from kuberentes-sigs internally.

Set up persistent volumes anywhere

To start, you need to have an OpenShift cluster, version 4.9.15 or later, and to log into the cluster with the cluster-admin role user.

Begin by installing the NFS Provisioner Operator:

# Login
oc login -u kubeadmin -p kubeadmin https://api.crc.testing:6443

# Create a new namespace
oc new-project nfsprovisioner-operator

# Deploy NFS Provisioner operator in the terminal (You can also use OpenShift Console
cat << EOF | oc apply -f -  
apiVersion: operators.coreos.com/v1alpha1
kind: Subscription
  name: nfs-provisioner-operator
  namespace: openshift-operators
  channel: alpha
  installPlanApproval: Automatic
  name: nfs-provisioner-operator
  source: community-operators
  sourceNamespace: openshift-marketplace

Next, create a directory in the node and add a label to that node:

# Check nodes
oc get nodes
NAME                 STATUS   ROLES           AGE   VERSION
crc-8rwmc-master-0   Ready    master,worker   54d   v1.22.3+e790d7f

# Set Env variable for the target node name
export target_node=$(oc get node --no-headers -o name|cut -d'/' -f2)
oc label node/${target_node} app=nfs-provisioner

# ssh to the node
oc debug node/${target_node}

# Create a directory and set up the Selinux label.
$ chroot /host
$ mkdir -p /home/core/nfs
$ chcon -Rvt svirt_sandbox_file_t /home/core/nfs
$ exit; exit

Now you need to deploy an NFS server using the created folder on a HostPath volume. Note that you could use an existing persistent volume claim (PVC) for the NFS server as well.

# Create NFSProvisioner Custom Resource
cat << EOF | oc apply -f -  
apiVersion: cache.jhouse.com/v1alpha1
kind: NFSProvisioner
  name: nfsprovisioner-sample
  namespace: nfsprovisioner-operator
    app: nfs-provisioner
  hostPathDir: "/home/core/nfs"

# Check if NFS Server is running
oc get pod
NAME                               READY   STATUS    RESTARTS   AGE
nfs-provisioner-77bc99bd9c-57jf2   1/1     Running   0          2m32s

Finally, you need to make the NFS StorageClass the default:

# Update annotation of the NFS StorageClass
oc patch storageclass nfs -p '{"metadata": {"annotations":{"storageclass.kubernetes.io/is-default-class":"true"}}}'

# Check the default next to nfs StorageClass
oc get sc
nfs (default)   example.com/nfs   Delete          Immediate           false                  4m29s

Congratulations—you have a StorageClass. Verify it as follows:

# Create a test PVC
oc apply -f https://raw.githubusercontent.com/Jooho/jhouse_openshift/master/test_cases/operator/test/test-pvc.yaml
persistentvolumeclaim/nfs-pvc-example created

# Check the test PV/PVC
oc get pv, pvc

NAME                                                        CAPACITY   ACCESS MODES   RECLAIM POLICY   STATUS      CLAIM                                                 STORAGECLASS   REASON   AGE
persistentvolume/pvc-e30ba0c8-4a41-4fa0-bc2c-999190fd0282   1Mi        RWX            Delete           Bound       nfsprovisioner-operator/nfs-pvc-example               nfs                     5s

NAME                                    STATUS   VOLUME                                     CAPACITY   ACCESS MODES   STORAGECLASS   AGE
persistentvolumeclaim/nfs-pvc-example   Bound    pvc-e30ba0c8-4a41-4fa0-bc2c-999190fd0282   1Mi        RWX            nfs            5s

The output shown here indicates that the NFS server, NFS provisioner, and NFS StorageClass are all working fine. You can use the NFS StorageClass for any test scenarios that need PVC.

CodeReady Containers allow you to do quite a bit with a local installation of RedHat OpenShift on your own hardware. For more, read my earlier article on Red Hat Developer, Configure CodeReady Containers for AI/ML development.