Cluster and node status
Reference for Cast AI cluster lifecycle status values and node health indicators, including API status values for building integrations.
Cast AI exposes two sets of status values: cluster status, which tracks a cluster's lifecycle and connection state to Cast AI, and node status, which reflects the health and readiness of individual nodes.
Add a new cluster to Cast AI and choose your operating mode.
Remove a cluster from Cast AI and clean up associated resources.
Manage Cast AI components including the agent responsible for cluster connectivity.
API reference for listing and managing clusters programmatically.
Cluster status
A cluster's status reflects its connection state to Cast AI and determines whether automated optimization is active. The console label and the API status field value are not always the same string — the table below shows both.
| Console label | API status value | What it means |
|---|---|---|
| Connecting | connecting | Cluster is connecting to Cast AI in read-only mode, or transitioning from read-only to managed mode. |
| Read only | ready | Cluster is connected in read-only mode. Cost reporting and savings insights are available; automation is not. |
| Connected | ready | Cluster is connected in managed mode. All reporting features are available and automation can be configured. |
| Warning | warning | Cast AI detected an issue preventing normal operation. The autoscaler will not issue commands. This state is usually self-recovering. |
| Not responding (Read only) | — | Agent heartbeat lost on a read-only cluster. If not restored within 5 minutes, the status changes to Disconnected (Read only). Check the castai-agent pod in the castai-agent namespace. |
| Not responding | — | Agent heartbeat lost on a managed cluster. Autoscaling is not working. Check the castai-agent pod in the castai-agent namespace. |
| Failed | failed | Persistent errors prevented recovery. Cast AI will not retry automatically. Hover over the status in the console to view error details, then check Cast AI components in the castai-agent namespace. |
| Hibernating | hibernating | Cluster hibernation was requested and is in progress. |
| Hibernated | hibernated | Cluster is hibernated and not running any nodes. Use the resume action to bring it back to active. |
| Resuming | resuming | Resume from hibernation was requested and is in progress. |
| Disconnecting | deleting | Cluster disconnect was requested and cleanup is in progress. |
| Disconnected | deleted | Cluster has been disconnected from Cast AI. Hover over the status in the console to see the disconnection time. |
NoteThe API also returns
archivedfor clusters that have been fully removed from the console view. Archived clusters do not appear in list responses and are only observable when polling a specific cluster ID directly.
Not respondingstates are derived from the agent heartbeat and do not map to a distinct APIstatusvalue. For agent connectivity, use theagent_statusfield.
Node status
A node's status indicates its health and readiness to accept workloads.
| Status | What it means |
|---|---|
Cordoned | Scheduling new pods onto this node is disabled. A user, system process, or Cast AI may have cordoned the node — for example, Cast AI cordons nodes when pods cannot be evicted during rebalancing with graceful rebalancing enabled. Inspect the node to understand the reason. If cordoned during rebalancing, adjust the pod disruption budget and un-cordon the node. |
Creating | Cast AI is provisioning the node. |
Deleting | Cast AI is deleting the node. |
Deleted | The node was deleted. This is a short-lived terminal state. |
Detached | The node still exists in the cloud but is no longer part of the Kubernetes cluster. Delete it manually from the cloud provider. |
Draining | Kubernetes is gracefully evicting pods from the node in preparation for deletion. |
Interrupted | A spot instance interruption event was received, a cloud provider rebalancing recommendation was issued, or Cast AI predicted imminent interruption. Cast AI is managing the event and preparing replacement capacity. |
Lost | The node is no longer part of the Kubernetes cluster but Cast AI has not yet deleted it. If this state persists, contact Cast AI support. |
Not ready | The node cannot accept workloads. It may still be initializing as part of provisioning, or it may be experiencing issues such as network problems or insufficient resources that prevent it from communicating with the control plane. If this state persists, contact Cast AI support. |
Ready | The node is fully operational and accepting pods. |
Updated 14 days ago