Advanced Pod CPU and Memory Management

In the previous section, we created CPU and Memory constraints at the pod level. LimitRange are used to constraint compute, storage or enforce ratio between Request and Limit in a Namespace. In this section, we will separate the compute workloads by low-usage, high-usage and unrestricted-usage.

We will create three Namespaces:

mkdir ~/environment/resource-management

kubectl create namespace low-usage
kubectl create namespace high-usage
kubectl create namespace unrestricted-usage

Create Limit Ranges

Create LimitRange specification for low-usage and high-usage namespace level. The unrestricted-usage will not have any limits enforced.

cat <<EoF > ~/environment/resource-management/low-usage-limit-range.yml
apiVersion: v1
kind: LimitRange
  name: low-usage-range
  - max:
      cpu: 1
      memory: 300M 
      cpu: 0.5
      memory: 100M
    type: Container

kubectl apply -f ~/environment/resource-management/low-usage-limit-range.yml --namespace low-usage

cat <<EoF > ~/environment/resource-management/high-usage-limit-range.yml
apiVersion: v1
kind: LimitRange
  name: high-usage-range
  - max:
      cpu: 2
      memory: 2G 
      cpu: 1
      memory: 1G
    type: Container

kubectl apply -f ~/environment/resource-management/high-usage-limit-range.yml --namespace high-usage

Deploy Pods

Next we will deploy the pods to the nodes .

Failed Attempts

Creating a pod with values outside what is defined in the LimitRange in the namespace will cause an errors

# Error due to higher memory request than defined in low-usage namespace: wanted 1g memory above max of 300m
kubectl run --namespace low-usage --requests=memory=1G,cpu=0.5 --image  hande007/stress-ng basic-request-pod --restart=Never --  --vm-keep   --vm-bytes 2g --timeout 600s --vm 1 --oomable --verbose 

# Error due to lower cpu request than defined in high-usage namespace: wanted 0.5 below min of 1
kubectl run --namespace high-usage --requests=memory=1G,cpu=0.5 --image  hande007/stress-ng basic-request-pod --restart=Never --  --vm-keep   --vm-bytes 2g --timeout 600s --vm 1 --oomable --verbose 
Successful Attempts

Create pods without specifying Requests or Limits will inherit LimitRange values.

kubectl run --namespace low-usage --image  hande007/stress-ng low-usage-pod --restart=Never --  --vm-keep   --vm-bytes 200m --timeout 600s --vm 2 --oomable --verbose 
kubectl run --namespace high-usage  --image  hande007/stress-ng high-usage-pod --restart=Never --  --vm-keep   --vm-bytes 200m --timeout 600s --vm 2 --oomable --verbose 
kubectl run --namespace unrestricted-usage --image  hande007/stress-ng unrestricted-usage-pod --restart=Never --  --vm-keep   --vm-bytes 200m --timeout 600s --vm 2 --oomable --verbose 

Verify Pod Limits

Next we will verify that LimitRange values are being inherited by the pods in each namespace.

eval 'kubectl  -n='{low-usage,high-usage,unrestricted-usage}' get pod -o=custom-columns='Name:spec.containers[*].name','Namespace:metadata.namespace','Limits:spec.containers[*].resources.limits';'


Name            Namespace   Limits
low-usage-pod   low-usage   map[cpu:1 memory:300M]
Name             Namespace    Limits
high-usage-pod   high-usage   map[cpu:2 memory:2G]
Name                     Namespace            Limits
unrestricted-usage-pod   unrestricted-usage   <none>


Clean up before moving on to free up resources

kubectl delete namespace low-usage
kubectl delete namespace high-usage
kubectl delete namespace unrestricted-usage