sre-k8s-resource-quota

Run kubectl get resourcequota -n <namespace> -o json and compare status.used against status.hard for each resource. If used divided by hard exceeds 0.9 for cpu or memory, new pods requesting that resource will be rejected by the quota admission controller.

Use kubectl describe resourcequota -n <namespace> for a human-readable table of Used versus Hard values per resource. For programmatic analysis, kubectl get resourcequota -o json exposes status.hard and status.used fields for cpu, memory, pods, and object counts.

Without a LimitRange, containers in the namespace have no enforced default requests or limits, creating noisy-neighbor risk on shared nodes. Pods running without explicit resources.limits should be flagged, and a LimitRange with sensible defaults should be proposed for approval.

No, the skill is strictly read-only and never runs kubectl edit, patch, or delete on quota or pod objects. It escalates findings with used/hard values and top consumer pods so an on-call engineer can approve and apply any changes.

Namespace-level ResourceQuota can block scaling even when nodes have free capacity, because quota is enforced per namespace by the admission controller. Check whether remaining quota (hard minus used) is smaller than the new pod's resource requests.