Cloning block volumes

Objective: Clone a Virtual Server block volume between storage classes, or regions.
Overview: This guide details using the clone_block_volume function from the Kubernetes Cloud repository to clone between two example datacenter regions. 
This guide requires kubectl and a valid access token. View Getting Started for more details.
Volume Cloning
Storage Classes
​Storage on CoreWeave Cloud is delineated by storage classes, in the notation of <storage type block|shared>-<storage medium hdd|nvme>-<region ord1|las1|etc>. For example:
block-nvme-ewr1
block-nvme-las1
block-nvme-lga1
block-nvme-ord1
​CSI Volume Cloning allows an existing volume to be duplicated:
Cloning via CLI within the same Storage Class
Cloning a block volume within the same storage class is easily done using our PVC cloning script, or with a simple PVC manifest:
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: destination-pvc
spec:
  accessModes:
  - ReadWriteOnce
  storageClassName: block-nvme-ord1
  volumeMode: Block
  resources:
    requests:
      storage: 1234Gi
  dataSource:
    kind: PersistentVolumeClaim
    name: source-pvc
Cloning via Web UI within the same Storage Class
Cloning within the same storage class can also be accomplished via the CoreWeave Cloud Storage UI: 
Cloning via CLI between Storage Classes
Cloning a block volume between storage classes, as done when cloning between regions, requires data within the volume to be manually duplicated, as it cannot be zero-copied. 
To automate the process of cloning a Virtual Server root disk block volume, we provide a block volume cloning script. This script creates a Kubernetes Job along with a new volume in the destination region, and uses dd to clone between source as destination.
First, we identify the Virtual Server root disk volume we'd like to clone:
kubernetes-cloud/clone_block_volume.sh at master · coreweave/kubernetes-cloud
GitHub
$ kubectl get pvc vs-example
NAME         STATUS   VOLUME                                     CAPACITY   ACCESS MODES   STORAGECLASS      AGE
vs-example   Bound    pvc-78d4950a-e1ba-476f-815e-b9c32e9c8b27   40Gi       RWO            block-nvme-ewr1   3m11s
Ensure it's not running:
$ kubectl get vs vs-example
NAME         STATUS                 REASON                 STARTED   INTERNAL IP   EXTERNAL IP
vs-example   VirtualServerStopped   VirtualServerStopped   False 
Download and source the script:
$ curl https://raw.githubusercontent.com/coreweave/kubernetes-cloud/master/virtual-server/clone_block_volume.sh -o ~/clone_block_volume.sh
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100  4109  100  4109    0     0  27211      0 --:--:-- --:--:-- --:--:-- 27211
​
$ source ~/clone_block_volume.sh
Initiate a clone:
$ clone_block_volume  --source vs-example --destination vs-example --region lga1
persistentvolumeclaim/vs-example-lga1 created
job.batch/clone-vs-example-lga1 created
No resources found in namespace.
pod/clone-vs-example-lga1-rn55n condition met
job.batch/clone-vs-example-lga1 condition met
42945478656 bytes (43 GB, 40 GiB) copied, 54 s, 795 MB/s 
40960+0 records in
40960+0 records out
42949672960 bytes (43 GB, 40 GiB) copied, 54.0091 s, 795 MB/s
job.batch "clone-vs-example-lga1" deleted
​
Script Workflow
In the clone example above, the following actions were performed:
A new PVC in the destination namespace was created - vs-example-lga1
A cloning job was created - clone-vs-example-lga1
The script pauses until a worker pod is spun up - clone-vs-example-lga1-rn55n
The script pauses until the job reports a completed status
Stats from the copy job are printed
The Job (and its worker pod) are cleaned up and deleted
Script Arguments
--source
Name of the root disk volume being cloned
--destination
Desired name of new volume
Region will always be appended as -region, e.g new-vol-lga1
--region
Name of destination region
Storage class is automatically derived as block-nvme-region
Last modified 1yr ago