OpenNebula CSI

Overview

This CSI driver is designed to be used by Kubernetes deployments inside OpenNebula virtual machines. The CSI driver provision persistent images with the requested size. It then attaches the images to the required virtual machine.

Deployment

The OpenNebula CSI follows the deployment pattern of having a single container image that contains all the required code.

Preparation

The CSI driver needs to map OpenNebula virtual machines to Kubernetes nodes in a one-to-one manner. To do so, it will try to find out the OpenNebula VM ID in the list of its arguments. If it fails, the driver will try to read the VM ID from a file located at /var/lib/cloud/vm-id.

The recommended approach is to add the contents of the contrib/one-context.sh script to the start-up script of the virtual machine. The script will take care of creating the /var/lib/cloud/vm-id file and populating its contents with the correct VM ID.

Note

After adding the code from contrib/one-context.sh to the start-up script of a virtual machine, the virtual machine must be restarted for the new start-up script to be executed.

Starting the CSI driver

Deploying the CSI driver is carried out by using kubectl:

kubectl apply -f manifests/

The above command will also create a StorageClass called storpool-nvme that will use the default OpenNebula datastore. Feel free to delete it and create your own StorageClasses as described in the next section.

Managing image datastores

The CSI driver is designed to map a Kubernetes StorageClass object to an OpenNebula image datastore in a one-to-one fashion. The mapping is done by setting the datastore_id parameter of the StorageClass object to match the desired OpenNebula datastore. Below is an example StorageClass object definition:

apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: one-datastore-1
provisioner: csi.opennebula.io
parameters:
  datastore_id: 1
mountOptions:
  - noatime
allowVolumeExpansion: true
volumeBindingMode: WaitForFirstConsumer

Building

The driver can be built like this:

docker build -t <repo_url>:<desired_tag> .

Note

If you decide to build a custom image, please remember to update the URL in the manifests