# kubectl describe node k8s-n2
Name:               k8s-n2
Roles:              <none>
Labels:             beta.kubernetes.io/arch=amd64
                    beta.kubernetes.io/os=linux
                    gpu_type=k20m
                    kubernetes.io/hostname=k8s-n2
                    local_storage=ssd
                    role=gpu-node
                    storage=ceph
                    type=physics
Annotations:        node.alpha.kubernetes.io/ttl=0
                    volumes.kubernetes.io/controller-managed-attach-detach=true
CreationTimestamp:  Thu, 03 May 2018 06:38:34 +0000
Taints:             <none>
Unschedulable:      false
Addresses:
  InternalIP:  109.105.1.208
  Hostname:    k8s-n2
Capacity:
 cpu:                  64
 ephemeral-storage:    229351Mi
 hugepages-1Gi:        0
 hugepages-2Mi:        0
 memory:               165091016Ki
 nvidia.com/gpu-k20m:  3
 pods:                 110
Allocatable:
 cpu:                  64
 ephemeral-storage:    216442758401
 hugepages-1Gi:        0
 hugepages-2Mi:        0
 memory:               164988616Ki
 nvidia.com/gpu-k20m:  0
 pods:                 110

============================================================

1. 每个deploy一个gpu, 起了三个low都running,此时资源耗尽

第四个low起不来,第五个middle杀掉了第三个,第五个running

起第六个high,需要2个gpu,杀掉了第一个和第二个,这俩low的,第六个running

删掉第六个,第三,第四 running

删掉第五个,第一个起

 

在另一个ns里起第七个high,2个gpu,杀掉的一个第2或3,第七个没起来,始终杀不掉两个

 

最终,全都挂掉

0/10 nodes are available: 10 Insufficient nvidia.com/gpu-k20m, 3 Insufficient memory, 9 node(s) didn't match node selector.
Update plugin resources failed due to requested number of devices unavailable for nvidia.com/gpu-k20m. Requested: 1, Available: 0, which is unexpected.