Pod Resource Management

Original Prepium workshop · Tested on Kubernetes v1.35.0 on 1 October 2026 in a kind cluster. Run this independent example in a disposable cluster. It uses its own quota and numbers, not the simulator's task.

When a ResourceQuota silently stops your Pods

A ResourceQuota caps the total resources a namespace may claim. Once a quota limits requests.cpu, limits.memory or similar, every new Pod in that namespace must declare those values, or the API server refuses to create it. The refusal does not appear on the Deployment you applied. It appears as an event on the ReplicaSet, which is why this failure looks silent.

1. Reproduce the symptom

Save this as quota-budget.yaml. It contains a quota and a Deployment whose container has no resources.

apiVersion: v1
kind: ResourceQuota
metadata:
  name: team-budget
  namespace: prepium-tutorial
spec:
  hard:
    requests.cpu: "1"
    requests.memory: 1Gi
    limits.cpu: "2"
    limits.memory: 2Gi
---
apiVersion: apps/v1
kind: Deployment
metadata:
  name: api
  namespace: prepium-tutorial
spec:
  replicas: 2
  selector:
    matchLabels:
      app: api
  template:
    metadata:
      labels:
        app: api
    spec:
      containers:
        - name: web
          image: nginx:1.27-alpine

kubectl create namespace prepium-tutorial
kubectl apply -f quota-budget.yaml
kubectl -n prepium-tutorial get deploy,rs,pods

The Deployment exists and shows 0/2 ready, the ReplicaSet wants 2 Pods, and there are no Pods at all. Not Pending, not crashing: never created. Follow the ownership chain down to where the error is recorded:

kubectl -n prepium-tutorial get events --field-selector reason=FailedCreate
kubectl -n prepium-tutorial describe rs

The message is precise: failed quota: team-budget: must specify limits.cpu for: web; limits.memory for: web; requests.cpu for: web; requests.memory for: web. It names the quota and the container, and lists every missing field.

2. Read the budget, then size the Pods

kubectl -n prepium-tutorial describe quota team-budget

The Hard column is the namespace total; Used is what admitted Pods already claim. Decide per-Pod values that fit the replica count you need, then set them on the Pod template:

kubectl -n prepium-tutorial set resources deployment api \
  --requests=cpu=200m,memory=256Mi --limits=cpu=400m,memory=512Mi
kubectl -n prepium-tutorial rollout status deployment/api
kubectl -n prepium-tutorial describe quota team-budget

Both Pods start. The quota now shows requests.cpu 400m / 1 and requests.memory 512Mi / 1Gi: two Pods' worth of what you declared.

3. Hit the ceiling on purpose

kubectl -n prepium-tutorial scale deployment api --replicas=6
kubectl -n prepium-tutorial get deploy api
kubectl -n prepium-tutorial get events --field-selector reason=FailedCreate | tail -1

Only 4 of 6 replicas appear. CPU had room for 5 Pods, but memory ran out at 4 (4 × 256Mi requests = 1Gi, and 4 × 512Mi limits = 2Gi). The event says exactly which dimension is exhausted: exceeded quota: team-budget, requested: limits.memory=512Mi,requests.memory=256Mi, used: limits.memory=2Gi,requests.memory=1Gi. The quota is enforced on whichever resource runs out first, so check every row, not just CPU.

4. Give containers defaults with a LimitRange

A namespace admin can avoid the "must specify" failure for everyone by adding defaults. Save this as limitrange-defaults.yaml:

apiVersion: v1
kind: LimitRange
metadata:
  name: container-defaults
  namespace: prepium-tutorial
spec:
  limits:
    - type: Container
      defaultRequest:
        cpu: 100m
        memory: 128Mi
      default:
        cpu: 200m
        memory: 256Mi

kubectl -n prepium-tutorial scale deployment api --replicas=1
kubectl apply -f limitrange-defaults.yaml
kubectl -n prepium-tutorial run bare --image=nginx:1.27-alpine
kubectl -n prepium-tutorial get pod bare -o jsonpath='{.spec.containers[0].resources}{"\n"}'

The Pod was created without any resources in its manifest, yet it shows requests: cpu 100m, memory 128Mi and limits: cpu 200m, memory 256Mi. The LimitRange admission step filled them in before the quota was checked. Defaults apply only to Pods created after the LimitRange; existing Pods are unchanged.

Cleanup: kubectl delete namespace prepium-tutorial (the quota and LimitRange are namespaced and go with it). Reference: Resource Quotas and Limit Ranges. Next, try autoscaling on CPU requests.