Managing Clusters¶
Everything below is available from a cluster's detail page in the console, and via the REST API.
Inspect¶
GET /v1/clusters— list your clusters.GET /v1/clusters/{id}— one cluster, with its spec and state.GET /v1/clusters/{id}/nodes— the cluster's nodes and their states.
The detail page shows the cluster's state, its node groups, access information, exposed ports, and recent operations.
Power control¶
You can power-cycle a cluster or an individual node without deleting it:
POST /v1/clusters/{id}/power— fan areboot/stop/startaction across all nodes.POST /v1/clusters/{id}/nodes/{nodeId}/power— one node.GET /v1/clusters/{id}/power— per-node live power state (and, where the hardware reports it, draw and temperature).
Stopping a cluster parks its nodes powered off; starting brings them back. This is useful to pause an idle cluster without losing it.
Resize¶
Change a cluster's size by adjusting its node groups. Growing a group reserves more machines of that profile (bounded by availability); shrinking returns machines to the pool. Resize runs as an operation, like create.
Note
Some interconnect-bound profiles resize in whole-domain units (for example a full NVLink domain at a time) rather than single nodes. The console only offers valid resize targets for the profile.
SSH keys¶
Manage the SSH keys authorized on a cluster after creation:
PUT /v1/clusters/{id}/ssh-keys— replace the authorized key set.
Replacing the set is atomic: the keys you send become the complete authorized list. Providing keys takes precedence over any password-based access. See Access & Networking.
Delete¶
DELETE /v1/clusters/{id}— tear the cluster down.
Deletion releases the nodes back to the pool (they're sanitized and
re-validated before they become available again) and stops metering. The
cluster record is retained in DELETED state for history and cost reporting.
flowchart LR
R[READY] -->|DELETE| D[DELETING]
D -->|nodes sanitized + returned| X[DELETED]
X -.->|nodes re-validated| P[(Pool: AVAILABLE)]
Deletion is final for the cluster's data
Anything on cluster-local storage is lost when the cluster is deleted. Copy out results, or use durable shared storage, before deleting.
Next: Access & Networking.