Lab 4.2 - 4.3 What does setting the resource limits really change?

vmayer

I have just completed chapter 4, but after doing the labs I am still confused about the actual impact on setting resource constraints. There is nothing shown about pods being evicted or node allocation or whatever this is actually for. If I set the limits or not, I will have the exact same impact when I deploy the hog application.

What is the actual impact of setting resource limits?

oleksazhel

@vmayer At 4.2. you set limit in 1CPU and requested load in 2CPU as a result you can see just 100% (not 200% what is equal to 2CPU) of CPU usage via top. It means that limits work.:

            limits:
              cpu: "1"
              memory: "4Gi"
            requests:
              cpu: "0.5"
              memory: "2500Mi"

            args:
            - -cpus
            - "2"
            - -mem-total
            - "950Mi"
            - -mem-alloc-size
            - "100Mi"
            - -mem-alloc-sleep
            - "1s"

In Lab 4.3. you set limits via deployment manifest and via namespace, but because "Per-deployment settings override the global namespace settings" there is no limitations and we eat 950MB. But in case you remove per-deployment limitations new pod will not be able to start. E.g., do following:

kubectl delete -n low-usage-limit deployments.apps hog
cp hog2.yaml hog3.yaml &&\
vim hog3.yaml

Remove following section and add {} after resources:

          limits:
            cpu: "1"
            memory: "4Gi"
          requests:
            cpu: "0.5"
            memory: "2500Mi"

Then:

kubectl create -f hog3.yaml

Monitor nodes load with top and check the status of new pod with kubectl get pod -A. It will be killed every time it exceeds 500Mi limit set on namespace.