mpi-container/job.yaml

taskGroups:
  - taskSpec:
      runnables:
        - container:
            imageUri: us-central1-docker.pkg.dev/<your project>/<your repo>/<your image name>
            entrypoint: bash
            commands: "/scripts/run-ssh.sh"
            volumes:
              - /etc/ssh:/etc/ssh
            options: "-td --network host"
          background: true
        - script:
            text: "until [ -f /etc/ssh/started ]; do sleep 1; done; echo sshd started on $(hostname)"
        - barrier:
            name: "vms-ready"
        - container:
            imageUri: us-central1-docker.pkg.dev/<your project>/<your repo>/<your image name>
            entrypoint: bash
            commands: "/scripts/run-mpi.sh"
            volumes:
              - /etc/cloudbatch-taskgroup-hosts:/etc/cloudbatch-taskgroup-hosts
              - /etc/ssh:/etc/ssh
            options: "--network host"
        - barrier:
            name: "finish-mpi"
    taskCount: 2
    taskCountPerNode: 1
    requireHostsFile: true
    permissiveSsh: true
allocationPolicy:
  instances:
    - policy:
        bootDisk:
          image: batch-debian
logsPolicy:
  destination: CLOUD_LOGGING