关于k8s驱逐节点上image和pod的问题

Question

查阅官网得知，通过设置驱逐条件的阈值，如果达到阈值，那么触发节点上进行pod的驱逐和image的驱逐。
问题: 我设置了驱逐条件如下，设置位于/etc/systemd/system/kubelet.service.d/10-kubeadm.conf
内容如下,主要添加了threshold of eviction

# Note: This dropin only works with kubeadm and kubelet v1.11+
[Service]
Environment="KUBELET_KUBECONFIG_ARGS=--bootstrap-kubeconfig=/etc/kubernetes/bootstrap-kubelet.conf --kubeconfig=/etc/kubernetes/kubelet.conf"
Environment="KUBELET_CONFIG_ARGS=--config=/var/lib/kubelet/config.yaml"
#threshold of eviction
Environment="KUBELET_EVICTION_ARGS=--eviction-hard=memory.available<100Mi,nodefs.available<1Gi,imagefs.available<1Gi"
# This is a file that "kubeadm init" and "kubeadm join" generates at runtime, populating the KUBELET_KUBEADM_ARGS variable dynamically
EnvironmentFile=-/var/lib/kubelet/kubeadm-flags.env
# This is a file that the user can use for overrides of the kubelet args as a last resort. Preferably, the user should use
# the .NodeRegistration.KubeletExtraArgs object in the configuration files instead. KUBELET_EXTRA_ARGS should be sourced from this file.
EnvironmentFile=-/etc/default/kubelet
ExecStart=
ExecStart=/usr/bin/kubelet $KUBELET_KUBECONFIG_ARGS $KUBELET_CONFIG_ARGS $KUBELET_KUBEADM_ARGS $KUBELET_EXTRA_ARGS $KUBERLET_EVICTION_ARGS

设置了区驱逐条件，内存小于100m，或者节点空间小于1g，或者images空间小于1g。我的节点的空间如下

Filesystem      Size  Used Avail Use% Mounted on
udev             63G     0   63G   0% /dev
tmpfs            13G   91M   13G   1% /run
/dev/sdc2       439G  384G   34G  93% /
tmpfs            63G   23G   41G  36% /dev/shm
tmpfs           5.0M  4.0K  5.0M   1% /run/lock
tmpfs            63G     0   63G   0% /sys/fs/cgroup
/dev/sdf        917G  389G  482G  45% /public
/dev/sdd        917G  828G   43G  96% /data
/dev/sde        917G  269G  603G  31% /weights
/dev/sdb        917G  9.0G  862G   2% /data2
/dev/sdc1       511M  3.7M  508M   1% /boot/efi
/dev/sda1       1.9T  484G  1.3T  28% /data1
tmpfs            13G   32K   13G   1% /run/user/108
tmpfs            13G     0   13G   0% /run/user/1009
tmpfs            13G     0   13G   0% /run/user/1005
tmpfs            13G     0   13G   0% /run/user/1008
tmpfs            13G     0   13G   0% /run/user/1002

根目录位于/dev/sdc/2 占用了93%,但是很明显绝对大于1G啊，为什么一直驱逐我的pod。
master节点kubectl get pod --all-namespace如下

kube-system            coredns-5644d7b6d9-d6xgf                     1/1     Running   0          20d
kube-system            coredns-5644d7b6d9-w76xz                     1/1     Running   0          25d
kube-system            etcd-vrlab                                   1/1     Running   0          25d
kube-system            kube-apiserver-vrlab                         1/1     Running   0          8d
kube-system            kube-controller-manager-vrlab                1/1     Running   0          8d
kube-system            kube-flannel-ds-amd64-9wzrb                  0/1     Evicted   0          7s
kube-system            kube-flannel-ds-amd64-c46qs                  1/1     Running   0          25d
kube-system            kube-flannel-ds-amd64-rvlzg                  1/1     Running   0          3d8h
kube-system            kube-proxy-7mz45                             0/1     Evicted   0          11m
kube-system            kube-proxy-csglg                             1/1     Running   0          24h
kube-system            kube-proxy-wpcl7                             1/1     Running   0          19d
kube-system            kube-scheduler-vrlab                         1/1     Running   4          11d
kube-system            metrics-server-6fd69cd864-2gqxb              1/1     Running   0          12d
kube-system            nvidia-device-plugin-daemonset-9x97f         0/1     Evicted   0          13m
kube-system            nvidia-device-plugin-daemonset-vlbqv         1/1     Running   0          3d8h
kubernetes-dashboard   dashboard-metrics-scraper-7cf5979dbb-27vf9   1/1     Running   0          30h
kubernetes-dashboard   kubernetes-dashboard-594549879-w7b2z         1/1     Running   0          30h

describe其中一个被驱逐的节点event信息如下

                 node.kubernetes.io/unschedulable:NoSchedule
Events:
  Type     Reason     Age    From                  Message
  ----     ------     ----   ----                  -------
  Warning  Evicted    4m36s  kubelet, amax-server  The node had condition: [DiskPressure].
  Normal   Scheduled  4m36s  default-scheduler     Successfully assigned kube-system/kube-flannel-ds-amd64-2svn9 to amax-server

可知确实是磁盘压力的问题。但是不知道问题在哪
可以看到flannel，kube-proxy一直再被驱逐？为什么？是哪里设置的不对吗？

Answer 1

已找到解决方案。目前重新配置的是/var/lib/kubelet/config.yaml，
以下四项是关于是否驱逐pod:

evictionHard:
  imagefs.available: 500Mi
  memory.available: 100Mi
  nodefs.available: 1Gi
  nodefs.inodesFree: 5%

上述是修改后的，虽说不会驱逐pod了，但是镜像文件仍然莫名的被删除，仔细阅读此yaml文件，发现为这两个字段

1.  `image-gc-high-threshold`，触发镜像垃圾回收的磁盘使用率百分比。默认值为 85%。
    
2.  `image-gc-low-threshold`，镜像垃圾回收试图释放资源后达到的磁盘使用率百分比。默认值为 80%。

一开始以为驱逐pod字段的imagefs为驱逐镜像的原因。然后自己配置所需百分比即可。