为什么``oom kill-disable`对` `kubectl create pod`无效，而对我自己系统中的` `docker run container `有效？

Question

我想禁止当内存被完全使用时k8s pod被oom杀死，首先我通过docker run尝试它，它工作；但是当我通过kubectl尝试它时，它失败了！

docker运行docker set oom-kill-disable success

kubectl version
Client Version: version.Info{Major:"1", Minor:"13", GitVersion:"v1.13.5", GitCommit:"2166946f41b36dea2c4626f90a77706f426cdea2", GitTreeState:"archive", BuildDate:"2019-07-30T03:39:26Z", GoVersion:"go1.11.5", Compiler:"gc", Platform:"linux/amd64"}
Server Version: version.Info{Major:"", Minor:"", GitVersion:"v0.0.0-master+2166946f41b36d", GitCommit:"2166946f41b36dea2c4626f90a77706f426cdea2", GitTreeState:"", BuildDate:"2019-08-21T08:42:24Z", GoVersion:"go1.11.5", Compiler:"gc", Platform:"linux/amd64"}

docker version = 19.03.7

oom.yaml

apiVersion: v1
kind: Pod
metadata:
  name: oom-test
spec:
  nodeSelector:
    kubernetes.io/hostname: "hostname"
  containers:
  - name: oom-test
    securityContext:
      capabilities:
        add: ["SYS_ADMIN", "SYS_RESOURCE", "IPC_LOCK"]
    image: myselfImage
    volumeMounts:
    - mountPath: /dev/shm
      name: cache-volume
    resources:
      limits:
        cpu: "4"
        memory: "10Gi"
      requests:
       cpu: "4"
       memory: "10Gi"
  volumes:
  - name: cache-volume
    emptyDir:
      medium: Memory
      sizeLimit: "5Gi"
  restartPolicy: Never

1.echo 1 > /sys/fs/cgroup/memory/kubepods/memory.oom_control
2.kubectl create -f oom.yaml
3.kubectl exec -ti oom-test bash
>>stress --vm 1 --vm-bytes 10G #pod is oom killed
NAME                     READY   STATUS      RESTARTS   AGE
devmachine-shm-test-02   0/1     OOMKilled   0          2m45s

为什么即使我设置了echo 1 > memory.oom_control，pod也会被杀死

Answer 1

当我为kubelet配置添加--cgroup-driver=cgroupfs时，它工作了！所以我认为这个问题可能是我自己的kubelet配置问题。