The BestEffort pod has the lowest priority in the Resource QoS classes. Therefore, in the case of a resource shortage, this BestEffort pod will be terminated by the Kubernetes scheduler, then will yield CPU and memory resources to other, higher priority pods.

In order to configure a pod as a BestEffort, you need to set the resource limit as 0 (explicit), or specify no resource limit (implicit).

1. Prepare a pod configuration that explicitly sets the spec.containers.resources.limits as 0:

$ cat besteffort-explicit.yml
apiVersion: v1
kind: Pod
metadata:
  name: besteffort
spec:
  containers:
  - name: nginx
    image: nginx
    resources:
      limits:
        cpu: 0
        memory: 0

2. Create the pod on both the chap8-qos and chap8-no-qos namespaces:

$ kubectl create -f besteffort-explicit.yml --namespace=chap8-qos
pod "besteffort" created


$ kubectl create -f besteffort-explicit.yml --namespace=chap8-no-qos
pod "besteffort" created

3. Check the QoS class; both pods have the BestEffort class:

$ kubectl describe pods besteffort --namespace=chap8-qos | grep QoS
QoS Class:       BestEffort

$ kubectl describe pods besteffort --namespace=chap8-no-qos | grep QoS
QoS Class:       BestEffort

There is a pitfall :  if you don't set any resource settings in the pod configuration, the pod takes a value from the namespace's default settings. Therefore, if you create a pod with no resource settings, the result will be different between chap8-qos and chap8-no-qos. The following example demonstrates how the namespace settings affect the result:

1. Delete the preceding pods from the chap8-qos and chap8-no-qos namespaces:

$ kubectl delete pod --all --namespace=chap8-qos
pod "besteffort" deleted

$ kubectl delete pod --all --namespace=chap8-no-qos
pod "besteffort" deleted

2. Prepare a pod configuration that doesn't have resource settings:

$ cat besteffort-implicit.yml
apiVersion: v1
kind: Pod
metadata:
  name: besteffort
spec:
  containers:
  - name: nginx
    image: nginx

3. Create the pod on both namespaces:

$ kubectl create -f besteffort-implicit.yml --namespace=chap8-qos
pod "besteffort" created

$ kubectl create -f besteffort-implicit.yml --namespace=chap8-no-qos
pod "besteffort" created

4. The result of the QoS class is different:

$ kubectl describe pods besteffort --namespace=chap8-no-qos |grep QoS
QoS Class:       BestEffort

$ kubectl describe pods besteffort --namespace=chap8-qos |grep QoS
QoS Class:       Burstable

Because the chap8-qos namespace has the default setting request.cpu: 0.1, it causes the pod to configure with the Burstable class. Therefore, we will use the chap8-no-qos namespace, which avoids this unexpected result.