A persistent volume (PV) is a cluster-wide resource that you can use to store data in a way that it persists beyond the lifetime of a pod. The PV is not backed by locally-attached storage on a worker node but by networked storage system such as EBS or NFS or a distributed filesystem like Ceph.

In order to use a PV you need to claim it first, using a persistent volume claim (PVC). The PVC requests a PV with your desired specification (size, speed, etc.) from Kubernetes and binds it then to a pod where you can mount it as a volume. So let’s create such a PVC, asking Kubernetes for 1 GB of storage using the default storage class:

$kubectl apply -f https://raw.githubusercontent.com/openshift-evangelists/kbe/master/specs/pv/pvc.yaml$ kubectl get pvc
NAME      STATUS   VOLUME                                     CAPACITY   ACCESS MODES   STORAGECLASS    AGE
myclaim   Bound    pvc-27fed6b6-3047-11e9-84bb-12b5519f9b58   1Gi        RWO            gp2-encrypted   18m


To understand how the persistency plays out, let’s create a deployment that uses above PVC to mount it as a volume into /tmp/persistent:

$kubectl apply -f https://raw.githubusercontent.com/openshift-evangelists/kbe/master/specs/pv/deploy.yaml  Now we want to test if data in the volume actually persists. For this we find the pod managed by above deployment, exec into its main container and create a file called data in the /tmp/persistent directory (where we decided to mount the PV): $ kubectl get po
$kubectl exec -it pv-deploy-69959dccb5-jhxxw -- bash bash-4.2$ touch /tmp/persistent/data
bash-4.2$ls /tmp/persistent/ data lost+found  It’s time to destroy the pod and let the deployment launch a new pod. The expectation is that the PV is available again in the new pod and the data in /tmp/persistent is still present. Let’s check that: $ kubectl delete po pv-deploy-69959dccb5-jhxxw
$kubectl get po NAME READY STATUS RESTARTS AGE pv-deploy-69959dccb5-kwrrv 1/1 Running 0 16m$ kubectl exec -it pv-deploy-69959dccb5-kwrrv -- bash
bash-4.2$ls /tmp/persistent/ data lost+found  And indeed, the data file and its content is still where it is expected to be. Note that the default behavior is that even when the deployment is deleted, the PVC (and the PV) continues to exist. This storage protection feature helps avoiding data loss. Once you’re sure you don’t need the data anymore, you can go ahead and delete the PVC and with it eventually destroy the PV: $ kubectl delete pvc myclaim