Exporting images to QCOW2

Objective: Create a QCOW2 image from a PVC hosted in our namespace.
Overview: We will spin up a shared filesystem to store a QCOW2 image, generated by a worker pod from a mounted PVC in our namespace.

References:

• clone-to-file.yaml
• shared_data.yaml

Create a shared filesystem

Creating a shared filesystem gives us a destination for our worker pod to write to, as well as a volume that can be attached to a Virtual Server or Samba Pod to egress the exported QCOW2 file.

We'll deploy the following YAML with kubectl create -f shared_data.yaml.

shared_data.yaml

apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: shared-data
spec:
  storageClassName: shared-hdd-ord1
  accessModes:
    - ReadWriteMany
  resources:
    requests:
      storage: 1000Gi

note

Note we've created our shared filesystem in the ORD region. If our source disk exists in a different region - we'll want to change the shared data filesystem to match.

Identify source disk

Using kubectl get pvc, we'll identify a PVC in our namespace that we wish to export:

note

Note our source image exists in the ORD1 region - matching our shared data filesystem.

Deploy worker pod

Next, we'll create a worker pod that has both our source disk, and shared data file system mounted.

Using kubectl create -f clone-to-file.yaml:

clone-to-file.yaml

apiVersion: batch/v1
kind: Job
metadata:
  labels:
    app.kubernetes.io/component: clone-to-file
  name: clone-to-file
spec:
  template:
    metadata:
      labels:
        app.kubernetes.io/component: clone-to-file
    spec:
      affinity:
        nodeAffinity:
          requiredDuringSchedulingIgnoredDuringExecution:
            nodeSelectorTerms:
              - matchExpressions:
                  - key: failure-domain.beta.kubernetes.io/region
                    operator: In
                    values:
                      - ORD1
                  - key: node.coreweave.cloud/class
                    operator: In
                    values:
                      - cpu
                  - key: ethernet.coreweave.cloud/speed
                    operator: In
                    values:
                      - 10G
                      - 40G
      containers:
        - name: clone-to-file
          resources:
            requests:
              cpu: 1
              memory: 2Gi
          image: registry.gitlab.com/coreweave/utility-images/admin-shell:36f48c0d
          command:
            - sh
            - '-c'
            - 'qemu-img convert -f raw -O qcow2 /dev/xvda /shared-data/disk.qcow2 -c -p'
          volumeMounts:
            - name: shared-data
              mountPath: /shared-data
          volumeDevices:
            - devicePath: /dev/xvda
              name: source
      restartPolicy: OnFailure
      volumes:
        - name: shared-data
          persistentVolumeClaim:
            claimName: shared-data
            readOnly: false
        - name: source
          persistentVolumeClaim:
            claimName: winserver2019std-clone-20210701-ord1
            readOnly: true
      tolerations:
        - key: node.coreweave.cloud/hypervisor
          operator: Exists
        - key: is_cpu_compute
          operator: Exists

important

Note that while our shared filesystem can be mounted to multiple pods/Virtual Servers simultaneously, our source block filesystem winserver2019std-clone-20210701-ord1 must not be in use by any running pods when this job is deployed.

Progress can be monitored with kubectl get pods --watch:

Once the job status shows Completed, the job can be deleted with kubectl delete job clone-to-file.

The shared data filesystem, with its exported QCOW2 can be attached to a Virtual Server or a apps.coreweave.com based File Browser or SAMBA for further inspection.