Deploy an NFS server and client

This tutorial describes how to deploy two VMs in different namespaces, where the first VM acts as an NFS server and the other as the client, able to mount the exported share via a VirtualMachineService.

Prerequisites

This tutorial assumes the following:

There are two namespaces in which the user is able to deploy VirtualMachine resources. In this tutorial, the namespaces shall be referred to as ns-1 and ns-2.
Both namespaces have access to a VirtualMachineClass. In this tutorial, the class used is my-vm-class.
Both namespaces have access to a VirtualMachineImage (VMI) with Ubuntu 22.04+. In this tutorial, the VMI used is vmi-0a0044d7c690bcbea.
Both namespaces have access to a StorageClass. In this tutorial, the class used is my-storage-class.

Deploy the server

The first step is to deploy a VM that will act as the NFS server, which is achieved by applying the following YAML:

apiVersion: vmoperator.vmware.com/v1alpha3
kind: VirtualMachine
metadata:
  name: nfs-server
  namespace: ns-1
  labels:
    app: nfs-server
spec:
  className:    my-vm-class
  imageName:    vmi-0a0044d7c690bcbea # Ubuntu 22.04 (jammy)
  storageClass: my-storage-class
  bootstrap:
    cloudInit:
      rawCloudConfig:
        name: nfs-server-bootstrap-data
        key:  user-data

---

apiVersion: v1
kind: Secret
metadata:
  name: nfs-server-bootstrap-data
  namespace: ns-1
stringData:
  user-data: |
    #cloud-config
    ssh_pwauth: true
    users:
    - name: user
      primary_group: user
      sudo: ALL=(ALL) NOPASSWD:ALL
      groups: users
      lock_passwd: false
      # Password is secret
      passwd: '$1$SaltSalt$YhgRYajLPrYevs14poKBQ0'
      shell: /bin/bash
    runcmd:
    - apt-get update --assume-no
    - apt-get install --assume-yes nfs-kernel-server sudo
    - mkdir -p /mnt/share
    - echo 'world.' >/mnt/share/hello
    - echo '/mnt/share *(rw,sync,no_subtree_check,insecure)' >>/etc/exports
    - exportfs -a

Create service to access NFS exports

Great, an NFS server is up and running in namespace ns-1! To ensure workloads outside of that namespace are able to mount the NFS server's exports, a VirtualMachineService is used to provide a load balanced IP address that can be accessed across namespaces.

apiVersion: vmoperator.vmware.com/v1alpha3
kind: VirtualMachineService
metadata:
  name: nfs-server
  namespace: ns-1
spec:
  selector:
    app: nfs-server
  type: LoadBalancer
  ports:
  - name: tcp-ssh
    port: 22
    protocol: TCP
    targetPort: 22
  - name: tcp-rpcbind
    port: 111
    protocol: TCP
    targetPort: 111
  - name: tcp-nfs
    port: 2049
    protocol: TCP
    targetPort: 2049

Use kubectl to make note of the service's external IP address, ex.:

$ kubectl -n ns-1 get service nfs-server
NAME         TYPE           CLUSTER-IP      EXTERNAL-IP   PORT(S)                   AGE
nfs-server   LoadBalancer   172.24.116.43   192.168.0.4   22/TCP,111/TCP,2049/TCP   2m3s

Deploy the client

With the NFS server up and running, it is time to deploy the VM that will act as the client by applying the following YAML:

apiVersion: vmoperator.vmware.com/v1alpha3
kind: VirtualMachine
metadata:
  name: nfs-client
  namespace: ns-2
  labels:
    app: nfs-client
spec:
  className:    my-vm-class
  imageName:    vmi-0a0044d7c690bcbea # Ubuntu 22.04 (jammy)
  storageClass: my-storage-class
  bootstrap:
    cloudInit:
      rawCloudConfig:
        name: nfs-client-bootstrap-data
        key:  user-data

---

apiVersion: v1
kind: Secret
metadata:
  name: nfs-client-bootstrap-data
  namespace: ns-2
stringData:
  user-data: |
    #cloud-config
    ssh_pwauth: true
    users:
    - name: user
      primary_group: user
      sudo: ALL=(ALL) NOPASSWD:ALL
      groups: users
      lock_passwd: false
      # Password is secret
      passwd: '$1$SaltSalt$YhgRYajLPrYevs14poKBQ0'
      shell: /bin/bash
    runcmd:
    - apt-get update --assume-no
    - apt-get install --assume-yes nfs-common sudo
    - mkdir /var/share
    - chmod 0777 /var/share

Create service to access NFS client via SSH

Just like a service is required to access the NFS exports across namespaces, the following YAML creates a service that allows external actors to access the NFS client via SSH:

apiVersion: vmoperator.vmware.com/v1alpha3
kind: VirtualMachineService
metadata:
  name: nfs-client
  namespace: ns-2
spec:
  selector:
    app: nfs-client
  type: LoadBalancer
  ports:
  - name: tcp-ssh
    port: 22
    protocol: TCP
    targetPort: 22

Use kubectl to make note of the service's external IP address, ex.:

$ kubectl -n ns-2 get service nfs-client
NAME         TYPE           CLUSTER-IP     EXTERNAL-IP   PORT(S)   AGE
nfs-client   LoadBalancer   172.24.204.9   192.168.0.5   22/TCP    51s

Alright, it is time to see if everything works as intended!

With the external IP address from the nfs-client service, SSH into the VM acting as the NFS client:
```
ssh user@<NFS_CLIENT_SERVICE_EXTERNAL_IP>
```
Print the contents of the file /var/share/hello to the screen:
```
cat /var/share/hello
```
Oops! An error occurred, because that file does not exist!
```
cat: /var/share/hello: No such file or directory
```

Mount the export from the NFS server:

sudo mount -t nfs <NFS_SERVER_SERVICE_EXTERNAL_IP>:/mnt/share /var/share

Print the contents of the file /var/share/hello one more time:
```
cat /var/share/hello
```
If the text world. appeared on the screen, then congratulations, the NFS export was successfully mounted to the client!