GPU admission

It is a scheduler extender for GPU admission. It provides the following features:

provides quota limitation according to GPU device type
avoids fragment allocation of node by working with gpu-manager

For more details, please refer to the documents in docs directory in this project

1. Build

$ make build

2. Run

2.1 Run gpu-admission.

$ bin/gpu-admission --address=127.0.0.1:3456 --v=4 --config=build/gpu-admission.config --master=127.0.0.1:8080

2.2 Configure kube-scheduler policy file, and run a kubernetes cluster.

Example for scheduler-policy-config.json:

{
"kind" : "Policy",
"apiVersion" : "v1",
"extenders" : [
  	{
          "urlPrefix": "http://127.0.0.1:3456/scheduler",
          "filterVerb": "predicates",
          "enableHttps": false,
          "nodeCacheCapable": false
  	}
    ],
"hardPodAffinitySymmetricWeight" : 10,
"alwaysCheckAllPredicates" : false
}

Do not forget to add config for scheduler: --policy-config-file=XXX --use-legacy-policy-config=true. Keep this extender as the last one of all scheduler extenders.

刘辉科 / gpu-admission

GPU admission

1. Build

2. Run

2.1 Run gpu-admission.

2.2 Configure kube-scheduler policy file, and run a kubernetes cluster.

简介

发行版

贡献者

近期动态

刘辉科 / gpu-admission .gitee-modal { width: 500px !important; }

GPU admission

1. Build

2. Run

2.1 Run gpu-admission.

2.2 Configure kube-scheduler policy file, and run a kubernetes cluster.

简介

发行版

贡献者

近期动态

搜索帮助

刘辉科 / gpu-admission