Nodes

A node is a machine — physical or virtual — that runs workloads. Nodes provide the CPU, memory, and storage that pods actually execute on. CasOS displays the status and capacity of every node in the cluster from the Nodes page.

Node health

The most important thing to know about a node is whether it is Ready. A node is Ready when the kubelet is running and can accept pods. NotReady means the node has reported a problem — often disk pressure, memory pressure, or a network issue. Unknown means the node controller has not heard from the node in a while (the default grace period is 40 seconds), which usually indicates the node is unreachable.

When a node stays NotReady or Unknown, the pod eviction timer starts. After a configurable grace period (default 5 minutes), Kubernetes moves the pods on that node to other nodes.

Node capacity and resource usage

The node list shows each node's allocatable CPU and memory — the amount available for pods after Kubernetes reserves capacity for system components. When metrics-server is installed, it also shows current CPU and memory utilization, which is the same data the Dashboard aggregates cluster-wide.

Adding worker nodes

CasOS runs the embedded control plane — the API server, scheduler, and controller manager. To actually schedule workloads you need at least one node with a kubelet. Install kubelet and containerd on the worker machine, then join it to the CasOS API server at https://<casos-host>:<apiserverPort> using a bootstrap token or a kubeconfig. See the upstream kubelet documentation for the join procedure.

In a single-machine setup, the machine running CasOS itself can also run a kubelet and act as a worker node.