On the Backups tab of an instance's detail page, every completed backup has a Restore action. Confirming it submits the task immediately — there is no second confirmation step, such as typing the instance name. Before you confirm, know two things: restoring overwrites the current system disk with the backup, and this cannot be undone; and once restore completes, the platform automatically resizes the system disk to match the instance's current shape, regardless of the shape the backup was taken under.
The name "restore" suggests it only writes data back. In practice it triggers three actions together — stop, overwrite, and resize — and the resize step is unconditional:
| What happens | What does not happen |
|---|---|
| If the instance is running when restore starts: the platform runs a graceful shutdown step and then a force-stop step in sequence — the force stop always runs regardless of whether the graceful shutdown succeeded, and retries until the VM is confirmed off — before continuing | If the instance is already stopped when restore starts: no power action runs at all — the whole restore happens while stopped |
| The system disk is overwritten with the backup's content, becoming an exact copy of the disk at backup time | Other attached data disks, the instance shape itself, the region, the availability domain, the VPC, SSH keys, and other console resources are untouched |
| Once the overwrite finishes, the system disk is unconditionally resized to match the instance's current shape | Resize only grows the disk, never shrinks it: if the disk already meets or exceeds the target size, the resize step is skipped with no error |
| If the instance was running when restore started: it is powered back on after the resize, returning to Running | There is no option to restore data without the reboot — whether a stop/start cycle happens is decided entirely by the instance's state when you click Restore, and cannot be turned off |
The confirmation dialog reads:
Restoring will overwrite the instance's current system disk with this backup. This cannot be undone, and the instance will briefly restart during the process. Continue?
This text does not mention the disk resize, and "will briefly restart" only holds when the instance was running when you started the restore — a stopped instance is not restarted at all. Rely on the table above, not the dialog text alone, when assessing the impact.
- Backup status: only backups with a Completed status can be selected for restore. Pending, in-progress, and failed backups do not appear as eligible.
- Instance provisioned: an instance that has not finished its initial provisioning has no system disk to overwrite, so restore cannot be started.
- No conflicting action in progress: restore is rejected if the instance is security-locked, delinquent, has a conflicting draft invoice, or has another work request already running.
- Maintenance lock does not block restore: unlike the power actions (start, stop, restart), an instance under a maintenance lock can still start a restore.
- Go to Compute > Instances, open the target instance's detail page, and switch to the Backups tab.
- Select only a backup with a Completed status, and check its note, creation time, and size.
- Open the row's action menu and select Restore.
- Read the confirmation dialog's warning about the overwrite and the restart, then select Restore to confirm. The task runs immediately — there is no second step asking you to type the instance name.
- Track progress under Work requests on the instance detail page: entries for restoring the backup archive and calibrating the system disk appear in order, and — if the instance was running when you started — matching stop and start entries appear before and after them.
Do not select Restore again after submitting, and do not run a rebuild, upgrade, or delete on the instance before the task finishes.
The instance's state at the moment you start the restore determines which of the two flows it goes through:
- Running when restore starts: Running → Stopping → Stopped → restore and resize run while stopped → Starting → Running. Track the restore-archive and disk-resize steps individually under Work requests. The instance is fully offline for the entire cycle, not just briefly unreachable, and active connections drop.
- Stopped when restore starts: Stopped → restore and resize run while stopped → still Stopped once the task completes. The platform does not power the instance back on for you.
When restore completes, the platform calibrates the system disk to the disk size that corresponds to the instance's shape at the moment restore runs, not the shape it had when the backup was created. If you changed the instance's shape after taking the backup, the restored disk is sized for whatever shape the instance has now — not the backup's original shape, and not any disk size recorded inside the backup file itself.
Restoring after a downgrade does not shrink the disk: the resize step only runs when the current disk is smaller than the target size, and is skipped without error once the disk already meets it. There is no matching shrink path.
Before starting, the platform runs a set of checks. The scenarios below are grouped by the message shown.
This instance has been stopped and network-isolated. Compute, connection, and resource management actions are unavailable while the security review is active.
Contact Cypacky Support through a ticket for the status of the review. Restore stays unavailable until it finishes.
This instance is delinquent, so you cannot operate it. Pay the associated invoices or delete the instance. Contact Cypacky Networks Support if you have questions.
Settle the outstanding balance to restore access to the Restore action.
Selecting a backup that is still pending, in progress, or failed is rejected by the backend. Return to the backup list and confirm the target backup's status column shows Completed before retrying.
The instance has not completed its initial setup, so there is no system disk for restore to overwrite. Wait for provisioning to finish before retrying.
Restore is rejected while a create, rebuild, upgrade, migration, or delete task on the instance is pending or running. Open Work requests on the instance detail page and confirm the last task has finished before retrying.
Restore is rejected while the instance has an unresolved draft invoice. Complete or cancel the associated billing change first.
A maintenance lock does not appear in this list — an instance under a maintenance lock can still start a restore normally.
- Check that the work request shows Completed with no subsequent failure message.
- Confirm the instance reached the state you expected: an instance that was running when you started returns to Running; one that was already stopped stays Stopped, and you need to start it yourself.
- Log in and check files, mount points, boot services, network configuration, and the system clock.
- Run application-level health checks on databases, web services, and scheduled tasks.
- If the instance fails to boot or is unreachable after restore, keep the work request ID, the instance ID, and the backup ID, then contact support. Do not overwrite the system disk again without a new backup.
If the goal is a parallel new instance rather than overwriting the current one, do not restore over it — convert the backup into a custom image instead, then create a new instance from that image. Refer to Convert a backup to a custom image.