Reconfiguring the Cluster
Some cluster changes can be made against a running cluster from the reconfigure screen, without a redeployment. Reconfigure is reached from System Configuration on the setup screen.
What Can Be Changed Live​
| Change | Impact |
|---|---|
| Add a storage engine, or add drives to a new engine | The cluster rebuilds data onto the added drives or ranks while staying online |
| Remove a storage engine | Data is drained off the engine first. Requires the cluster to have redundancy factor ≥ 2. |
| Change a node's fabric address | Runs as a planned full stop — every engine stops while the address moves, storage I/O pauses cluster-wide, and clients must remount afterwards |
| Swap a node's fabric NIC for another of the same type | Same planned full stop as an address change |
Engine restarts and drain/rebuild may briefly stall I/O on the affected nodes. Once started, reconfiguration runs to completion — cancelling mid-flight is not supported.
What Cannot Be Changed Live​
- Drives on an existing engine cannot be changed while the cluster is running. To expand capacity, add an engine instead.
- Layouts that would require a data migration, or that the cluster classes as unsupported, are refused at preview with an explanation — those need a redeployment.
- A shrink that would drop the cluster below three ranks while volume groups exist is refused.
Running a Reconfigure​
- Open the reconfigure screen and edit the layout — engines are chipped as Current, Adding, or Removing
- The preview refuses unsafe layouts before any change reaches the cluster
- Review the change plan — each row shows the kind, node, engine number, and detail
- Tick I understand the impact and want to proceed and click Apply
- Track progress on the reconfigure status page
The wizard survives a browser refresh mid-apply; it resumes rather than bouncing back to step 1.
Actions that stop or restart storage services are blocked while a volume group is still being created, with a message naming the group. Wait for creation to finish before reconfiguring.
Safe Mode​
Enabling Safe Mode blocks writes cluster-wide and may disrupt ongoing operations. A warning dialog explains this and requires explicit confirmation before it takes effect.