Clusters
Cloud OS nodes can form a cluster: one primary and several workers, connected by an encrypted WireGuard mesh and kept consistent with the Raft consensus protocol. The Control Panel shows every cluster of your workspace and draws its topology.
Cluster List
Each cluster shows:
- its name and, in MSP workspaces, the client it belongs to,
- the primary node,
- the number of nodes,
- the current Raft term (it increases on every leader election, so a fast-growing term hints at an unstable cluster),
- its health:
| Health | Meaning |
|---|---|
| ok (shown as healthy) | All nodes are online and every link from the primary is up |
| degraded | At least one node is degraded or offline, or a link is slow (above 100 ms) or down |
| down | The primary is offline, or no node of the cluster is reachable |
Topology
Select a cluster to see its map: the primary in the middle and a link to every worker. Each node shows its status, role, WireGuard address and the round-trip time from the primary. Links are drawn as up, degraded or down. Below the map you see the node count, Raft term, number of apps placed on the cluster and the mean mesh latency.
The map is built from what each node reports on its heartbeat (every 30 seconds), so it reflects the cluster as the nodes see it. Members that are part of the OS cluster but not registered in the Control Panel still appear, drawn with a dashed outline (not managed). Click a managed node on the map to open it.
Cluster Page
Open a cluster (the arrow on its row, or Open on the map) to see its page: health, subnet, primary and client at the top, then the average CPU, memory and disk of its nodes, the mesh latency, the topology map and the Raft facts (leader, commit index). The Nodes tab lists every member with its WireGuard address, round-trip time, version and hardware; remove a worker from the cluster with the × on its row. Join token creates a token to add another Cloud OS node, and Delete removes the cluster (its nodes stay in your fleet).
Clusters → Autoscale adds managed nodes on Hetzner or DigitalOcean with a step-by-step wizard and lists them so you can decommission them (Team and Enterprise plans).
Next Steps
- High Availability — Control Panel HA and rolling upgrades
- Fleet Management — node health and regions