The memory resources we are going to see on this blog are

- Memory request = Minimum or guaranteed memory for the container

- Memory limit = Maximum memory, should not exceed for the container

 

Let’s define the Pod configuration

networkandcode@cloudshell:~ (kubernetes-cka-224606)$ cat ex1.yml

---

apiVersion: v1

kind: Pod

metadata:

 name: pod1

 namespace: default

spec:

 containers:

 - name: ctr1

   image: polinux/stress

   resources:

     limits:

       memory: "150Mi"

     requests:

       memory: "100Mi"

   command: ["stress"]

   # --vm-bytes 120M lets the container try to allocates 150MiB of memory

   args: ["--vm", "1", "--vm-bytes", "120M", "--vm-hang", "1"]

…

 

Let’s create the Pod

networkandcode@cloudshell:~ (kubernetes-cka-224606)$ kubectl create -f ex1.yml

pod/pod1 created

 

The Pod should now be running
networkandcode@cloudshell:~ (kubernetes-cka-224606)$ kubectl get pods

NAME   READY STATUS    RESTARTS AGE

pod1   1/1 Running   0 24m

 

To view more details about the Pod
networkandcode@cloudshell:~ (kubernetes-cka-224606)$ kubectl get pod pod1 -o yaml

--TRUNCATED--

spec:

 containers:

 - args:

   - --vm

   - "1"

   - --vm-bytes

   - 120M

   - --vm-hang

   - "1"

   command:

   - stress

   image: polinux/stress

   imagePullPolicy: Always

   name: ctr1

   resources:

     limits:

       memory: 150Mi

     requests:

       cpu: 100m

       memory: 100Mi

  --TRUNCATED--

 

To get the metric details of the Pod
networkandcode@cloudshell:~ (kubernetes-cka-224606)$ kubectl top pod pod1

NAME   CPU(cores)   MEMORY(bytes)

pod1   80m      120Mi

 

So, it’s currently consuming 120Mi, which is 20Mi above the ‘memory request’ value(100Mi) we specified, however is  30Mi below the ‘memory limit’ configured(150Mi).

 

Let’s define another pod by modifying the memory resources as follows
networkandcode@cloudshell:~ (kubernetes-cka-224606)$ cat ex2.yml

---

apiVersion: v1

kind: Pod

metadata:

 name: pod2

 namespace: default

spec:

 containers:

 - name: ctr2

   image: polinux/stress

   resources:

     requests:

       memory: "50Mi"

     limits:

       memory: "100Mi"

   command: ["stress"]

   args: ["--vm", "1", "--vm-bytes", "120M", "--vm-hang", "1"]

...

 

Let’s create this Pod now
networkandcode@cloudshell:~ (kubernetes-cka-224606)$ kubectl create -f ex2.yml

pod/pod2 created


This Pod isn’t running due to memory constraint
networkandcode@cloudshell:~ (kubernetes-cka-224606)$ kubectl get pod pod2

NAME   READY STATUS             RESTARTS AGE

pod2   0/1 CrashLoopBackOff   3 67s

 

There is no metrics as it isn’t running

networkandcode@cloudshell:~ (kubernetes-cka-224606)$ kubectl top pod pod2

NAME   CPU(cores)   MEMORY(bytes)

pod2   0m      0Mi

 

Let’s check for more details
networkandcode@cloudshell:~ (kubernetes-cka-224606)$ kubectl get pod pod2 -o yaml
--TRUNCATED--

   containerStatuses:

 - containerID: docker://7a42ac697350621cbf53245a60a5e550b1314971b65eae204334cebef6aabd26

   image: polinux/stress:latest

   imageID: docker-pullable://polinux/stress@sha256:6d1825288ddb6b3cec8d3ac8a488c8ec2449334512ecb938483fc2b25cbbdb9a

   lastState:

     terminated:

       containerID: docker://5e34e57ffa71e06dfd9abf4550bf847320dcf14ea78f2e16d130f72fcce47362

       exitCode: 1

       finishedAt: 2019-03-23T18:25:55Z

       reason: OOMKilled

       startedAt: 2019-03-23T18:25:54Z

   name: ctr2

   ready: false

   restartCount: 4

   state:

     terminated:

       containerID: docker://7a42ac697350621cbf53245a60a5e550b1314971b65eae204334cebef6aabd26

       exitCode: 1

       finishedAt: 2019-03-23T18:26:45Z

       reason: OOMKilled

       startedAt: 2019-03-23T18:26:44Z

--TRUNCATED--

 

We can see this warning by checking the node details too
networkandcode@cloudshell:~ (kubernetes-cka-224606)$ kubectl describe nodes | grep Warning

 Warning  OOMKilling                 3m8s kernel-monitor, gke-cluster-default-pool-f22707e6-28mw   Memory cgroup out of memory: Kill process 11226 (stress) score 1958 or sacrifice child

 Warning  OOMKilling  3m7s kernel-monitor, gke-cluster-default-pool-f22707e6-28mw  Memory cgroup out of memory: Kill process 11274 (stress) score 1959 or sacrifice child

 Warning  OOMKilling  2m53s kernel-monitor, gke-cluster-default-pool-f22707e6-28mw  Memory cgroup out of memory: Kill process 11386 (stress) score 1958 or sacrifice child

 Warning  OOMKilling  2m28s kernel-monitor, gke-cluster-default-pool-f22707e6-28mw  Memory cgroup out of memory: Kill process 11482 (stress) score 1958 or sacrifice child

 Warning  OOMKilling  98s kernel-monitor, gke-cluster-default-pool-f22707e6-28mw  Memory cgroup out of memory: Kill process 11674 (stress) score 1959 or sacrifice child

 Warning  OOMKilling  8s kernel-monitor, gke-cluster-default-pool-f22707e6-28mw  Memory cgroup out of memory: Kill process 11985 (stress) score 1958 or sacrifice child

 

Let’s delete all the Pods we created - Pod1 and Pod2
networkandcode@cloudshell:~ (kubernetes-cka-224606)$ kubectl delete pods --all

pod "pod1" deleted

pod "pod2" deleted

 

Let’s define a new Pod which very high memory request that is more than the available memory on any one of the nodes in the Cluster

networkandcode@cloudshell:~ (kubernetes-cka-224606)$ cat ex3.yml

---

apiVersion: v1

kind: Pod

metadata:

 name: pod3

 namespace: default

spec:

 containers:

 - name: ctr3

   image: polinux/stress

   resources:

     requests:

       memory: "900Gi"

     limits:

       memory: "1000Gi"

   command: ["stress"]

   args: ["--vm", "1", "--vm-bytes", "120M", "--vm-hang", "1"]

…

 

It should remain in Pending state
networkandcode@cloudshell:~ (kubernetes-cka-224606)$ kubectl create -f ex3.yml

pod/pod3 created

networkandcode@cloudshell:~ (kubernetes-cka-224606)$ kubectl get pods pod3

NAME   READY STATUS    RESTARTS AGE

pod3   0/1 Pending   0 73s

 

Reference: https://kubernetes.io/docs/tasks/configure-pod-container/assign-memory-resource/

 

--end-of-post--