Atlas: Node NotReady

What this usually means

The control plane has stopped receiving healthy heartbeats from the node, or the node reports a condition that makes it unsafe. Treat it as a partial partition until proven otherwise.

What to inspect first

Contain the impact. Then read conditions.

• Check conditions: Ready, MemoryPressure, DiskPressure, NetworkUnavailable.
• Check which workloads are on the node.

kubectl get nodes -o wide
kubectl describe node <node>
# Optional containment:
kubectl cordon <node>

Likely causes

NotReady can be network, kubelet, disk, or runtime instability.

• Network partition between node and API server.
• Disk pressure (node filesystem full; image garbage collection failing).
• CNI failure causing NetworkUnavailable.
• Container runtime issues; kubelet cannot manage pods.
• Host maintenance or kernel-level instability.

Resolution guidance

Prefer containment and controlled evacuation over heroic intervention.

• If the node is unhealthy, drain: `kubectl drain <node> --ignore-daemonsets --delete-emptydir-data` (validate flags for your posture).
• Fix disk pressure by freeing space; verify kubelet recovers conditions.
• If CNI is failing, restore the node networking plane before expecting pods to be reachable.
• Replace compromised nodes rather than patching indefinitely; treat nodes as cattle.