Skip to main content

Resource Requests vs Limits

Hands-on example​

Assume:

  • A working Kubernetes cluster
  • kubectl is configured

Initial state​

No Pods exist.

kubectl get pods
No resources found in default namespace.

Step 1: Create a Pod with requests and limits​

kubectl apply -f - <<EOF
apiVersion: v1
kind: Pod
metadata:
  name: resource-demo
spec:
  containers:
  - name: app
    image: busybox
    command: ["sh", "-c", "sleep 3600"]
    resources:
      requests:
        memory: "64Mi"
        cpu: "100m"
      limits:
        memory: "128Mi"
        cpu: "200m"
EOF
pod/resource-demo created

Check the Pod.

kubectl get pod resource-demo
NAME            READY   STATUS    RESTARTS   AGE
resource-demo   1/1     Running   0          5s

Check allocated resources.

kubectl describe pod resource-demo | grep -A5 Limits
Limits:
  cpu:     200m
  memory:  128Mi
Requests:
  cpu:        100m
  memory:     64Mi

What changed:

  • Scheduler reserved requested resources on a node

What did not change:

  • Limits did not affect scheduling

Step 2: Exceed the memory limit​

Force the container to allocate memory.

kubectl exec resource-demo -- sh -c 'dd if=/dev/zero of=/tmp/mem bs=1M count=200'
Killed

Check Pod status.

kubectl get pod resource-demo
NAME            READY   STATUS      RESTARTS   AGE
resource-demo   0/1     OOMKilled   1          30s

What changed:

  • Container was killed due to memory limit

What did not change:

  • Pod was not rescheduled to another node

Key observation​

  • Requests affect placement
  • Limits affect runtime behavior