Forge (Live Snapshots)
Overview
Forge provides live snapshots of running instances with full memory state capture. Create a checkpoint of a running instance, make changes, then choose to keep or revert everything -- a "try before you keep" workflow for risky operations.
Key capabilities:
- Live snapshots -- Snapshots are taken while the instance is running, with no downtime
- Memory state capture -- Full memory is saved, enabling restoration to the exact runtime state (running processes, open connections, etc.) when discarding changes
- Per-disk selection -- Choose which attached disks to include in the snapshot
- Atomic operation -- The snapshot either succeeds for all selected disks or fails entirely -- no partial snapshots
Use Forge before OS upgrades, kernel updates, configuration changes, or software installations to ensure you can always roll back.
Prerequisites
- Instance must be running
- Instance must not be suspended
- No other task may be running on the instance
- No existing active Forge session on the instance
- At least one attached disk must be selected
Enabling Forge
For Users
- Navigate to Instances and open the instance manage page
- Click the Forge tab
- Select the disks to include -- the primary disk is pre-selected by default
- Click Enable Forge
- Confirm the action
For Admins
- Navigate to Compute > Instances > Manage on the target instance
- Open the Forge tab
- Select the disks to include
- Click Enable Forge
- Confirm the action
What Happens
A snapshot is created that captures:
- The current state of all selected disks
- The full contents of the instance's memory
After activation, all writes to the selected disks are captured separately while the original disk data remains untouched. The instance remains running throughout -- there is a brief pause (usually under one second) while the snapshot is created.
Managing an Active Forge Session
Once Forge is active, the Forge tab displays the session details:
- Activated timestamp -- When the snapshot was taken
- Forged Disks -- Which disks are included
- Status badge -- Current state of the forge session
Two actions are available:
| Action | Button | Description |
|---|---|---|
| Commit Changes | Green "Commit Changes" button | Makes all changes since activation permanent |
| Discard Changes | Red "Discard Changes" button | Reverts disks to pre-snapshot state and restores the instance from the memory snapshot |
Both buttons are disabled while another task is running on the instance.
Committing Changes
Committing makes all changes since activation permanent by merging them back into the original disk images.
- Click Commit Changes
- Confirm the action -- this cannot be undone
The merge operation runs live while the instance continues operating. For instances with heavy write workloads and large disks, this may take several minutes. Progress updates are sent per-disk.
The commit operation may take time for large disks with significant write volume. The instance remains running during the commit.
Discarding Changes
Discarding reverts the instance to the exact state it was in when Forge was activated -- same disk contents, same memory, same running processes.
- Click Discard Changes
- Confirm the action -- all changes since activation will be lost
The instance is restored from the memory snapshot. If the memory restore fails for any reason, the system falls back to a cold start -- the disk state is still correctly reverted, but the instance boots fresh rather than resuming from memory.
If the memory restore is not possible, the instance is started normally. Disk data is always reverted regardless.
Status States
| Status | Badge Color | Description |
|---|---|---|
| Creating | Blue | Snapshot is being taken |
| Active | Yellow | Forge is active, changes are being captured separately |
| Committing | Blue | Changes are being merged into the original disks |
| Discarding | Blue | Changes are being reverted |
| Committed | Green | Changes were permanently merged (complete) |
| Discarded | Gray | Changes were reverted (complete) |
| Failed | Red | Operation failed |
During transitioning states (Creating, Committing, Discarding), the UI displays a spinner with a status message. The Commit and Discard buttons are hidden during these states.
Blocked Operations
While Forge is active, the following instance operations are blocked:
- Reinstall
- Resize (upgrade/downgrade)
- Migrate
- Suspend / unsuspend
- Add / remove / resize disks
- Create backups
- Mount / unmount ISOs
- Create images from instance
- Enable / disable public interface
- Destroy instance
These operations are incompatible with an active Forge session. You must complete the session (commit or discard) before performing any of these actions.
Power operations (stop, restart, kill) and VNC console access remain available.
Troubleshooting
Forge enable fails
- Verify the instance is running, not suspended, and has no active task
- Check that at least one disk is selected and attached
- Check the task log for error details
Commit takes a long time
The commit operation must merge all data written since activation back into the original disk images. For instances with heavy write workloads, this can take several minutes. The timeout is 1 hour. Monitor progress via the task system.
Discard fails to restore from memory
If the memory snapshot cannot be restored (e.g., the snapshot is too old or configuration changed), the system falls back to a normal start. The disk state is still correctly reverted -- only the in-memory state (running processes) is lost.
Instance stuck in a transitioning state
If a Forge operation fails without completing (e.g., due to a network issue), the session may remain in Creating, Committing, or Discarding status. Check the task status on the admin dashboard. If the task shows as failed, contact your administrator to resolve the stuck state.
Blocked operations while Forge is active
If you need to perform a blocked operation (resize, migrate, etc.), you must first commit or discard the active Forge session. There is no way to bypass this restriction.