Key node checks
- Read Ready alongside pressure conditions rather than in isolation.
- Compare allocatable values with current workload placement.
- Review runtime, kubelet version and taints when scheduling looks unusual.
Operator actionIf a node is Ready but pressure signals are present, use events and workload placement before deciding whether the node needs escalation or simply closer watch.
Signals that point to host-local issues
| Signal | Interpretation |
|---|---|
| One node accumulates warnings | The issue may be localized to runtime, resources or placement. |
| Pods from several namespaces degrade on the same node | A host-level cause becomes more likely. |
| Healthy images but poor placement outcomes | Scheduling or capacity is more likely than registry failure. |