Skip to content

Keep data on a volume

Put an application's data on a volume, so you can replace the instance without losing it, then take a snapshot and prove that you can restore it.

You will use: an instance, a volume and a snapshot, in a project in Karachi.

Before you start

  • A running Linux instance, here db-1, that you can log in to with sudo.
  • A role of project admin or project member, and room in the project's volume quota.

1. Create and attach a volume

  1. Under Storage → Volumes, choose Create volume: name db-data, Empty, a Type, and Size 20. Create it.
  2. When it is available, choose Manage → Attach to instance and pick db-1.
Attach a volume to one of the project's instances
Attach a volume to one of the project's instances
curl -sS -X POST "$NUBESTACK_API/v1/orgs/$ORG/projects/$PROJECT/volumes" \
  -H "Authorization: Bearer $NUBESTACK_API_KEY" -H "Idempotency-Key: $(uuidgen)" \
  -H "Content-Type: application/json" \
  -d '{"location": "loc_01jbkhi", "display_name": "db-data", "size_gb": 20}'

# once its operation has succeeded:
curl -sS -X POST "$NUBESTACK_API/v1/orgs/$ORG/projects/$PROJECT/servers/srv_01jbdb1:attach-volume" \
  -H "Authorization: Bearer $NUBESTACK_API_KEY" -H "Idempotency-Key: $(uuidgen)" \
  -H "Content-Type: application/json" \
  -d '{"volume_id": "vol_01jbdbdata"}'

2. Format and mount it

Log in to the instance and find the new disk by its size. Its device name can differ between images and attachments; do not assume /dev/vdb.

lsblk -o NAME,SIZE,FSTYPE,MOUNTPOINTS

Format only a new, empty volume

Formatting erases the disk. A volume restored from a snapshot, or moved from another instance, already has a file system: mount it, do not format it.

sudo mkfs.ext4 -L db-data /dev/vdb        # the empty disk you found above
sudo mkdir -p /srv/db-data
echo 'LABEL=db-data /srv/db-data ext4 defaults,nofail 0 2' | sudo tee -a /etc/fstab
sudo mount /srv/db-data
df -h /srv/db-data

Mounting by label, with nofail, keeps the instance starting even if the volume is detached.

3. Take a snapshot

A snapshot taken by hand needs the volume detached, so nothing is half-written.

  1. Stop the application, then sudo umount /srv/db-data.
  2. Manage → Detach on the volume.
  3. Manage → Create snapshot, named db-data-before-upgrade.
  4. Attach the volume again and sudo mount /srv/db-data.

For snapshots on a timetable without detaching, use a backup schedule.

4. Restore it to a new volume

Restoring a volume from a snapshot The volume db-data is attached to the instance db-1. A snapshot of it, db-data-before-upgrade, records its contents at one moment, on the same storage as the volume. To restore, make a new volume from the snapshot and attach it to an instance; the original volume is not changed. db-1 Instance db-data Volume · 100 GiB db-data-before-upgrade Snapshot · same storage db-data-restored New volume attached snapshot restore A snapshot shares the volume's storage: it is a recovery point, not a backup on separate storage.
Restoring makes a new volume from the snapshot. The original volume and the snapshot are not changed.
  1. On the snapshot, choose Manage → Create volume from it, named db-data-restored.
  2. Attach the new volume to the instance, beside the original.

  3. Mount it read-only somewhere else and compare:

    sudo mkdir -p /mnt/restored
sudo mount -o ro LABEL=db-data /mnt/restored   # if both carry the label, use the device name from lsblk
ls /mnt/restored

  4. When you are satisfied, unmount and detach it, and delete it if it was only a test.

A restore you have tried is worth more than many you have not.

5. Grow it later

Detach the volume, choose Manage → Extend with the new size, attach it again, then grow the file system inside: sudo resize2fs /dev/vdb for ext4 on the whole disk.

Clean up

Unmount and detach the volume, delete its snapshots, then delete the volume. A volume cannot be deleted while it is attached or has snapshots.