Argo Workflow 프레임 추출

frame_extract.yaml

apiVersion: argoproj.io/v1alpha1
kind: Workflow
metadata:
  generateName: media-works-
spec:
  entrypoint: media-works-ffmpeg
  volumes:
  - name: workdir
    persistentVolumeClaim:
      claimName: azure-managed-disk
  templates:
  - name: media-works-ffmpeg
    container:
      image: briankim/ffmpeg:v1
      command: [ffmpeg]
      args: ["-i", "/mnt/vol/big_file.mp4","-r", "1/1", "/mnt/vol/$frame%03d.jpg"]
      volumeMounts:
      - name: workdir
        mountPath: /mnt/vol

PVC 접근을 위한 더미 Pod 설정 파일

• data_access.yaml

apiVersion: v1
kind: Pod
metadata:
    name: dataaccess
spec:
    containers:
    - name: alpine
      image: alpine:latest
      command: ['sleep', 'infinity']
      volumeMounts:
      - name: workdir
        mountPath: /mnt/vol
    volumes:
    - name: workdir
      persistentVolumeClaim:
        claimName: azure-managed-disk

저 파일을 kubectl apply -f data_access.yaml 로 띄운 다음,

$ kubectl cp small_file.mp4 dataaccess:/mnt/vol/small_file.mp4

식으로 Persistent Volumne에 파일 복사

프레임 추출 Job 실행을 위한 컨테이너 Dockerfile (briankim/ffmpeg:v1 이미지, Docker Hub에 올려놨어요.)

• Dockerfile

FROM ubuntu:20.04
LABEL maintainer="Brian Kim"
RUN apt update && apt install -y ffmpeg

• 병렬 처리를 위한 Dag 설정

apiVersion: argoproj.io/v1alpha1
kind: Workflow
metadata:
  generateName: dag-diamond-
spec:
  entrypoint: diamond
  volumes:
  - name: workdir
    persistentVolumeClaim:
      claimName: azure-managed-disk
  templates:
  - name: ffmpeg
    container:
      image: briankim/ffmpeg:v1
      command: [ffmpeg]
      args: ["-i", "/mnt/vol/small_file.mp4","-r", "1/1", "/mnt/vol/$frame%03d.jpg"]
      volumeMounts:
      - name: workdir
        mountPath: /mnt/vol
  - name: diamond
    dag:
      tasks:
      - name: A
        template: ffmpeg
      - name: B
        template: ffmpeg
      - name: C
        template: ffmpeg
      - name: D
        template: ffmpeg
      - name: E
        template: ffmpeg