Curtin and preseed file

Accelerated Computing DGX User Forum
1

I want to create a packer image for auto-installing. I wonder if you have dgx_a100-curtin.yaml and a preseed.sh template.

Thanks.

2

Hi @pordonez ,

We have instructions for where to find the curtin file in the PXE section of the DGX OS 5 User Guide. You might also take a look at the DeepOps MaaS documentation, specifically the DGX section at the bottom.

Does that get you what you need?

ScottE

3

Scott,

You might also take a look at the DeepOps MaaS documentation, specifically the DGX section at the bottom?

Using this approach, will install Ubuntu 20.04 on the first SSD (nvme1n1 3.4TB), which is not our goal. Besides, it doesn’t install the OFED, NVIDIA and other software.

Our goal is to install the OS as raid 1 on nvme2n1 and nvme3n1 ssd’s. Below is the curtin storage script; however, I got an error (look below).

Any help is more than welcome.

Thanks.

Curtin script

storage:
  version: 1
  config:
    - id: nvme2n1
      type: disk
      ptable: gpt
      path: /dev/nvme2n1
      name: osdisk1
      wipe: superblock-recursive
    - id: nvme3n1
      type: disk
      ptable: gpt
      path: /dev/nvme3n1
      name: osdisk2
      wipe: superblock-recursive
    - id: nvme2n1-part1
      type: partition
      device: nvme2n1
      number: 1
      size: 512MB
      flag: boot
    - id: nvme3n1-part1
      type: partition
      device: nvme3n1
      number: 1
      size: 512MB
      flag: boot
    - id: nvme2n1-part2
      type: partition
      device: nvme2n1
      number: 2
      size: 1830900MB
      flag: raid
    - id: nvme3n1-part2
      type: partition
      device: nvme3n1
      number: 2
      size: 1830900MB
      flag: raid
    - id: md_device0
      type: raid
      name: md0
      raidlevel: 1
      devices:
        - nvme2n1-part2
        - nvme3n1-part2
    - id: nvme2n1-part1-fs1
      type: format
      fstype: fat32
      label: boot
      volume: nvme2n1-part1
    - id: md_root
      type: format
      fstype: ext4
      volume: md_device0
    - id: md_mount
      type: mount
      path: /
      device: md_root
      options: errors=remount-ro
      passno: 1
    - id: boot_mount
      type: mount
      path: /boot/efi
      device: nvme2n1-part1-fs1
      passno: 1

Error

Error rescanning devices, possibly known issue LP: #1489521
cmd: ['blockdev', '--rereadpt', '/dev/nvme2n1']
stdout:''
stderr:blockdev: ioctl error on BLKRRPART: Device or resource busy
exit_code:1