Why is my Kubernetes Pod OOMKilled?

2 min readJan 14, 2021

Find OOMKilled Pods

An OOMKilled Kubernetes Pod has a status of OOMKilled. The kubectl get pod and kubectl describe pod commands will both display the OOMKilled status.

$ kubectl -n k8salliance get pod nginx
NAME            READY     STATUS      RESTARTS   AGE
nginx           0/1       OOMKilled   1          28s

The OOMKilled status means that Kubernetes stopped the Pod because the Pod exceeded its memory limits.

How Does Kubernetes Determine Memory and CPU Limits?

Kubernetes allocates memory and CPU resources to a Pod according to the Pod’s resource settings.

A Pod’s resource settings define the CPU and memory available to containers in the Pod.

The request value specifies the initial amount allocated to the Pod. The limit value specifies the maximum amount available to the containers in the Pod. Each container in the Pod launches with the request amount and may consume up to the limitamount.

View Namespace Memory and CPU Usage Limits

If a Pod does not specify resource settings, it inherits the resource settings of the namespace.

The kubectl describe command displays the resource limits of a namespace.

$ kubectl describe namespace k8salliance
Name:         k8salliance
Labels:       <none>
Annotations:  <none>
Status:       ActiveResource Quotas
 Name:                       gke-resource-quotas
 Resource                    Used  Hard
 --------                    ---   ---
 count/ingresses.extensions  0     100
 count/jobs.batch            0     5k
 pods                        0     1500
 services                    0     500No resource limits.

Modify Namespace Memory and CPU Usage Limits

To set the default resource limits of a namespace, create a LimitRange object in the namespace.

Any Pod created in the namespace without resource usages defined will inherit the namespace’s default resource request and limit settings.

kubectl -n k8salliance create -f - <<EOF
apiVersion: v1
kind: LimitRange
metadata:
  name: default-resources
spec:
  limits:
  - default:
      memory: 128Mi
    defaultRequest:
      memory: 64Mi
    type: Container
EOF$ kubectl describe namespace k8salliance        
Name:         k8salliance
Labels:       <none>
Annotations:  <none>
Status:       Active
Resource Quotas
 Name:                       gke-resource-quotas
 Resource                    Used  Hard
 --------                    ---   ---
 count/ingresses.extensions  0     100
 count/jobs.batch            0     5k
 pods                        0     1500
 services                    0     500
Resource Limits
 Type       Resource  Min  Max  Default Request  Default Limit  Max Limit/Request Ratio
 ----       --------  ---  ---  ---------------  -------------  -----------------------
 Container  memory    -    -    64Mi             128Mi          -

Set Pod Resource Limits

To prevent your Kubernetes Pod from being OOMKilled, you need to define the memory and CPU resources required by the Pod.

apiVersion: v1
kind: Pod
metadata:
  name: default-mem-demo
spec:
  containers:
  - name: default-mem-demo-ctr
    image: nginx