hostpath-provisioner-operator

The Kubernetes operator for managing KubeVirt hostpath provisioner deployment. Leverages the operator-sdk.

Any version after v0.9.0 requires Kubernetes version >= 1.21. This is because the operator manages the CSIDriver object and tries to update it with fields that only exist in kubernetes >= 1.21

How to deploy

As of version 0.11 the hostpath provisioner operator now requires cert manager to be installed before deploying the operator. This is because the operator now has a validating webhook that verifies the contents of the CR are valid. Before deploying the operator, you need to install cert manager:

$ kubectl create -f https://github.com/cert-manager/cert-manager/releases/download/v1.7.1/cert-manager.yaml

Please ensure the cert manager is fully operational before installing the hostpath provisioner operator:

$ kubectl wait --for=condition=Available -n cert-manager --timeout=120s --all deployments

Next, you need to create the hostpath provisioner namespace:

$ kubectl create -f https://raw.githubusercontent.com/kubevirt/hostpath-provisioner-operator/main/deploy/namespace.yaml

Followed by the webhook:

$ kubectl create -f https://raw.githubusercontent.com/kubevirt/hostpath-provisioner-operator/main/deploy/webhook.yaml -n hostpath-provisioner

And then you can create the operator:

$ kubectl create -f https://raw.githubusercontent.com/kubevirt/hostpath-provisioner-operator/main/deploy/operator.yaml -n hostpath-provisioner

If you want to change the namespace in which you create the provisioner, make sure to update the ClusterRoleBinding and RoleBinding namespaces in the operator.yaml to match your namespace. Also change the namespace by changing the -n argument

Once you have installed the operator, you need to create an instance of the Custom Resource to deploy the hostpath provisioner in the hostpath-provisioner namespace.

Custom Resource with storage pool (CR)

apiVersion: hostpathprovisioner.kubevirt.io/v1beta1
kind: HostPathProvisioner
metadata:
  name: hostpath-provisioner
spec:
  imagePullPolicy: Always
  storagePools:
    - name: "local"
      path: "/var/hpvolumes"
  workload:
    nodeSelector:
      kubernetes.io/os: linux

This example names the storage pool 'local' and the path used is '/var/hpvolumes'. No pvc template is defined so the directories will be created on the node filesystem in the specified path. The matching storage class looks like this:

apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: hostpath-csi
provisioner: kubevirt.io.hostpath-provisioner
reclaimPolicy: Delete
volumeBindingMode: WaitForFirstConsumer
parameters:
  storagePool: local

Notice the storagePool parameter. This lets the provisioner know which pool to use. You can define multiple storage pools each with a different name.

Custom Resource with PVCTemplate storage pool

Example CR allows you specify the storage pool you wish to use as the backing storage for the persistent volumes. You specify the path to use to create volumes on the node, and the name of the storage pool. The name of the storage pool is used in the storage class to identify the pool. You also specified the PVC template to use. This causes the operator to create PVCs for each node that match the workload nodeSelector and a pod that mounts that PVC on to the node at the path specified. The hpp csi driver will then use the PVC to create directories on. If the storageClassName is not specified the default storage class will be used.

apiVersion: hostpathprovisioner.kubevirt.io/v1beta1
kind: HostPathProvisioner
metadata:
  name: hostpath-provisioner
spec:
  imagePullPolicy: Always
  storagePools:
    - name: "local"
      pvcTemplate:
        accessModes:
          - ReadWriteOnce
        resources:
          requests:
            storage: 5Gi
      path: "/var/hpvolumes"
  workload:
    nodeSelector:
      kubernetes.io/os: linux

This example names the storage pool 'local' and the path used is '/var/hpvolumes'. The matching storage class looks like this:

apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: hostpath-csi
provisioner: kubevirt.io.hostpath-provisioner
reclaimPolicy: Delete
volumeBindingMode: WaitForFirstConsumer
parameters:
  storagePool: local

Notice the storagePool parameter. This lets the provisioner know which pool to use. You can define multiple storage pools each pointing to a different path.

Legacy CR

If you are using a previous version of the hostpath provisioner operator your CR will look like this:

apiVersion: hostpathprovisioner.kubevirt.io/v1beta1
kind: HostPathProvisioner
metadata:
  name: hostpath-provisioner
spec:
  imagePullPolicy: IfNotPresent
  pathConfig:
    path: "/var/hpvolumes" #The path of the directory on the node
    useNamingPrefix: false #Use the name of the PVC bound to the created PV as part of the directory name.

The operator will continue to create the legacy provisioner in addition to the CSI driver. If you use the legacy format of the CR, you can use the legacy CSI storage class to create the storage class for the CSI driver.

To create the CustomResource

$ kubectl create -f https://raw.githubusercontent.com/kubevirt/hostpath-provisioner-operator/main/deploy/hostpathprovisioner_cr.yaml -n hostpath-provisioner

Once the CustomResource has been created, the operator will deploy the provisioner and CSI driver as a DaemonSet on each node.

Storage Class

The hostpath provisioner supports two volumeBindingModes, Immediate and WaitForFirstConsumer. In general WaitForFirstConsumer is preferred however this requires Kubernetes >= 1.12 and if one is running an older kubernetes that volumeBindingMode will not work. Immediate binding mode is now deprecated and may be removed in the future. For this reason the operator will not create the StorageClass for you and you will have to do it yourself. Example storageclass yamls are available in deploy directory in this repository.

SELinux (legacy only)

On each node you will have to give the directory you specify in the CR the appropriate selinux rules by running the following (assuming you pick /var/hpvolumes as your PathConfig path):

$ sudo chcon -t container_file_t -R /var/hpvolumes

Another way to configure SELinux when using OpenShift is using a MachineConfig.

Deployment in OpenShift

The operator will create the appropriate SecurityContextConstraints for the hostpath provisioner to work and assign the ServiceAccount to that SCC. This operator will only work on OpenShift 4 and later (Kubernetes >= 1.12).

TLS Crypto Configuration

The operator deploys a webhook server;
Security-minded cluster administrators might want to set specific TLS ciphers/minimum version that clients may use when connecting to this server.
The operator will poll for cluster-wide crypto policy (via OpenShift's APIServer) and comply to those.
On non-OpenShift installs this is allowed via environment variables on the operator deployment, for example for Modern TLS spec:

        - name: TLS_CIPHERS_OVERRIDE
          value: TLS_AES_128_GCM_SHA256,TLS_AES_256_GCM_SHA384,TLS_CHACHA20_POLY1305_SHA256
        - name: TLS_MIN_VERSION_OVERRIDE
          value: VersionTLS13

OVERRIDEs will take precedence.