Skip to content

Sacred Systems

Kubelet and the Discipline of Obedience

The kubelet is where the platform’s abstract intent becomes real processes. It obeys—but it also refuses when the node is dying.

Text

Authored as doctrine; evaluated as operations.

Doctrine

The kubelet is a delegated authority. It enforces pod specs, reports node status, and makes local decisions under pressure.

Kubblai doctrine: nodes are not cattle when you care about latency, storage, or GPU. They are institutions with constraints.

Node pressure and eviction truth

Evictions are not random. They are the node protecting itself: memory pressure, disk pressure, PID exhaustion. Your platform must be designed for this truth.

If evictions surprise you, you are missing telemetry.

Image pulls, registries, and cold starts

Registry latency and image size become outages during scale events. Pre-pulling, caching, and sensible image hygiene are operational requirements.

Kubblai doctrine: artifact size is availability.

Operator practice

Know how to read node conditions and correlate with pod behavior.

kubectl

shell

kubectl describe node <node>
kubectl get pods -A -o wide --field-selector spec.nodeName=<node>