Host Agents
The HEAT node agent is a small daemon that runs on each cluster host, outside the cluster itself. It reports the host’s health and hardware, and it delivers platform patches. This matters most on air-gapped sites, where there is no other channel to the host. Open Monitoring, then Host Agents (/admin/monitoring/host-agents).
The matrix
Summary counts show how many hosts exist, how many agents are reachable, and how many are missing. Re-probe all refreshes every host at once.
Each row is one host.
| Column | Meaning |
|---|---|
| Node | The host name |
| Agent | Whether the node agent is reachable |
| K3s | Whether the Kubernetes service is running on the host |
| Platform | The platform version the host reports |
| Disk free | Free space on the host |
| Actions | Details, and Stop K3s or Start K3s |
Details opens a side panel with Health and Spec tabs. The Spec tab lists the operating system, CPU, memory, disks, network and kernel of the host.
Starting and stopping K3s
The Stop K3s and Start K3s actions control the Kubernetes service on that host. Stopping K3s takes the host out of the cluster. On a single-node appliance that stops HEAT, so it asks you to confirm first.
Use these controls only when directed, for example during a planned host operation. On a single-node environment, stopping K3s is a full outage until you start it again.
When an agent is unreachable
An unreachable agent means the daemon is not running on that host. It does not by itself mean the cluster is down, but patches cannot be delivered and host health cannot be read until it is back. Restoring the agent is a host-level task for whoever administers the machine.