Notebook 생성

Notebook 생성

GPU에 Project 할당

MLXP 콘솔에서 GPU Resource 접근

GPU Zone을 클릭하여 GPU 정보를 조회

관리버튼을 눌러 할당하고자 하는 Project에 GPU 인스턴스 할당

Project가 GPU Zone 정보에서 조회 되는지 확인

Kubernetes에서 NodeSelector 주입

해당 단계 이전에 Notebook을 생성할 경우, 아래와 같은 로그와 함께 실패

Reissued from statefulset/test-with-gpu:
create Pod test-with-gpu-0 in StatefulSet test-with-gpu failed
error: admission webhook "default.pods.workload.mlx.navercorp.com" denied the request: 
zones.management.mlx.navercorp.com "test" not found

Context Switching을 통해 Project에 할당된 Kubernetes에 접속

• Project Home에는 Context Switching을 위한 Kuberconfig 파일 다운로드

• 다운로드 받은 매니페스트 파일의 내용을 클러스터에 저장

vi kubeconfig.yaml

apiVersion: v1
clusters:
- cluster:
    certificate-authority-data: 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
    server: https://a0896928-f343-4a13-ad0e-1bc00416cf77.kr.vnks.ntruss.com
  name: mlx-kpb4r
contexts:
- context:
    cluster: mlx-kpb4r
    namespace: p-yuhwan-test
    user: mlx-7fa96a70-f1d6-11f0-b997-246e96591a38
  name: 7fa96a70-f1d6-11f0-b997-246e96591a38@mlx-kpb4r/p-yuhwan-test
current-context: 7fa96a70-f1d6-11f0-b997-246e96591a38@mlx-kpb4r/p-yuhwan-test
kind: Config
preferences: {}
users:
- name: mlx-7fa96a70-f1d6-11f0-b997-246e96591a38
  user:
    token: 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.Aix8r7ZqIgFiPpYE7sFckzUJhaS6HBUB9y2BTr_Hj59fFyDpRioC4dRQNKHf-mQXWRybI_MgzwUlPcll2ij-_czvASBbwnFJmSyYqf-K5HR2062IUmCwyhvVaU9nmqPX7yUPaOF-Q7EhbmTyEBg7-vKRvBerxPq5CuCBrsEaOzoMIYzevY61HD-j12lDUr3OoX0rJxfwOUouTJBeDOfxevcwMUT21vqI1wzebU8d59MzGJ-mKe4mV4jJ-IwKZsh-G7vuGn1GZdpzKwBcmEJvX06EV__S3SLAElEAANLdAi2gqE0HhQuxJlLDaQFezkxSfQ9B9-Bx73-iHDPUObLOFQ

• 저장 후 kubeconfig를 환경변수로 적용

export KUBECONFIG=~/kubeconfig.yaml

kubectl config current-context
# 정상적으로 적용됐다면 다음과 같은 결과가 나옵니다.
# 7fa96a70-f1d6-11f0-b997-246e96591a38@mlx-kpb4r/p-yuhwan-test

Notebook에 GPU 할당을 위한 nodeSelector 적용

• Notebook 생성 시 Advanced Setting을 클릭

• 생성단계에서 Configuration 주입

Notebook 확인

• NodeSelector 적용 후, 웹 콘솔에서 Notebook 확인