Skip to main content
In this guide, we demonstrate how Kubernetes enforces CPU and memory constraints using two example pods, rabbit and elephant. You will learn how to:
  • Identify resource requests and limits
  • Diagnose pod failures caused by insufficient resources
  • Update pod specifications to prevent OOM kills

​
Pod “rabbit”: Identifying CPU Requests

First, confirm that the rabbit pod is running and examine its resource settings: 
kubectl get pod rabbit
kubectl describe pod rabbit
Excerpt from the description shows: 
Containers:
  cpu-stress:
    Limits:
      cpu: 2
    Requests:
      cpu: 1
Here, the rabbit pod requests 1 CPU and caps at 2 CPUs. To clean up: 
kubectl delete pod rabbit
# pod "rabbit" deleted
The image shows a terminal window displaying Kubernetes pod details and events, alongside a task prompt to delete the "rabbit" pod.

​
Pod “elephant”: Diagnosing Memory Issues

When you deploy the elephant pod, it enters a CrashLoopBackOff: 
kubectl get pods
NAME       READY   STATUS             RESTARTS   AGE
elephant   0/1     CrashLoopBackOff   1          8s
Inspect its status and events: 
kubectl describe pod elephant
Key output sections: 
Containers:
  mem-stress:
    State: Waiting
    Reason: CrashLoopBackOff
    Last State:
      Terminated:
        Reason: OOMKilled
        Exit Code: 1
    Limits:
      memory: 10Mi
    Requests:
      memory: 5Mi
...
Events:
  Type     Reason          Age   From               Message
  Normal   Scheduled       23s   default-scheduler  Successfully assigned default/elephant to controlplane
  Normal   Pulled          20s   kubelet            Successfully pulled image "polinux/stress"
  Normal   Started         18s   kubelet            Started container mem-stress
  Normal   Back-off        18s   kubelet            Back-off restarting failed container
The pod is OOMKilled because it exceeded its 10Mi memory limit while the stress workload uses around 15Mi.
The CrashLoopBackOff status indicates the container repeatedly failed due to out-of-memory errors. Always ensure your limits.memory exceed the actual usage of your application.
The image shows a Kubernetes pod named "elephant" with a memory issue, requiring an increase in its memory limit from 15Mi to 20Mi. The terminal displays details about the pod's status, including a "CrashLoopBackOff" error due to being "OOMKilled."

​
Updating the Memory Limit

You cannot patch resource limits on a running pod. Instead, edit the pod manifest and recreate it: 
kubectl edit pod elephant
The YAML editor opens. Find the resources section and change: 
spec:
  containers:
  - name: mem-stress
    resources:
      limits:
        memory: 20Mi
      requests:
        memory: 5Mi
Save and exit. You’ll see an error because live edits to limits are not allowed: 
error: pods "elephant" is invalid
A copy of your changes has been stored to "/tmp/kubectl-edit-*.yaml"
error: Edit cancelled, no valid changes were saved.
Apply the updated manifest by deleting and recreating the pod: 
kubectl replace --force -f /tmp/kubectl-edit-*.yaml
Verify the new memory configuration: 
kubectl describe pod elephant | grep -A3 "Limits:"
Expected output: 
Limits:
  memory: 20Mi
Requests:
  memory: 5Mi
...
Ready: True
The image shows a code editor with a YAML configuration file for a Kubernetes pod named "elephant," alongside instructions to increase its memory limit to 20Mi.

​
Resource Summary

PodCPU RequestCPU LimitMemory RequestMemory Limit
rabbit1 CPU2 CPU––
elephant––5Mi20Mi

​
Clean Up

Remove the elephant pod to finish the exercise: 
kubectl delete pod elephant
# pod "elephant" deleted

​
Further Reading

Watch Video