This is part of a series on creating a Kubernetes cluster. In the previous post, we installed Helm to make deployments easier. In this post, we’ll start configuring our new Kubernetes (k8s) cluster by setting up persistent storage.
In any containerized system, there is likely the need for data to persistent after the container is gone. Commonly, that data is some configuration for the application inside the container, but it may also be files that are used by the container. For example, in a container for a photo management app, when a user uploads a photo to the app, it gets stored on a volume somewhere. If that container is stopped or deleted, the users would appreciate all their photos not being lost.
Kubernetes further increases the need for persistent storage because of the ephemeral nature of the cluster. A node may be taken out of service, resource constraints may require containers to be shifted around, or an application might be scaled to multiple instances.
Kubernetes uses volumes to provide a container with persistent storage. A volume is mounted in a container at a specified path, then the container uses that path as if it were the local file system. In k8s, this is accomplished via a Persistent Volume and Persistent Volume Claim.
A [Persistent Volume](https://kubernetes.io/docs/concepts/storage/persistent-volumes/) (PV) is a volume that has either been provisioned by a cluster administrator for future use or has been dynamically provisioned by a Storage Class. Various cloud providers and other plugins provide the ability to dynamically provision storage upon request. For our cluster, we’re going to be using NFS storage, so we will be statically provisioning it.
A [Persistent Volume Claim](https://kubernetes.io/docs/concepts/storage/persistent-volumes/) (PVC) is a request for storage provided by a Persistent Volume. There can only be one PVC bound to a PV at any given time. Once the PVC releases its claim, the PV can be released or deleted. The PVC will be used by the container to mount a path to the volume.
For our purposes, we will setup a PV that is attached to an NFS share on a storage server. Creating an NFS share is beyond the scope of this article, but you will need one available before continuing.
You configure objects in k8s by using the kubectl command to apply a yaml file. Sometimes these yaml files are on your server, and sometimes they are on the internet. You can apply them either way. In fact, you already applied one when you configured the cluster with the flannel network. We’ll be writing our own yaml file on the master node to create our PV.
Login to the master node using ssh.
Save the following yaml into a file named nfs-pv.yaml, replacing <nfs-server> and <nfs-path> with your appropriate share details. This will create a PV with 1 GB of capacity.
apiVersion: v1
kind: PersistentVolume
metadata:
name: nfs
spec:
capacity:
storage: 1Gi
volumeMode: Filesystem
storageClassName: slow
persistentVolumeReclaimPolicy: Retain
accessModes:
- ReadWriteMany
nfs:
server: <nfs-server>
path: "<nfs-path>"
Apply the yaml file to the cluster: kubectl apply -f nfs-pv.yaml
The PV is now ready to be bound to a PVC.
Save the following yaml into a file named nfs-pvc.yaml. Note the volumeName parameter matches the name parameter of the above PV. If you create additional PV and PVC objects later, be sure to use the corresponding name.
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
name: nfs
spec:
accessModes:
- ReadWriteMany
storageClassName: slow
resources:
requests:
storage: 1Gi
volumeName: "nfs"
Apply the yaml file to the cluster: kubectl apply -f nfs-pvc.yaml
Now that we have persistent storage provisioned and available for use, we need to configure our networking.