feat(dra): DRA driver installation

[runzhen]

What this PR does / why we need it:
Initial checkin for dra-driver-nvidia-gpu

Below items will be handled in another PR

1. add "enable DRA" filed in proto buffer.
2. support AzureLinux

unit_test.txt

Events:
  Type    Reason     Age   From               Message
  ----    ------     ----  ----               -------
  Normal  Scheduled  37s   default-scheduler  Successfully assigned default/pod to 3718-2026-06-16-ubuntu2404dradrivernvidiagpurunningdefaul000000
  Normal  Created    13s   kubelet            spec.containers{ctr0}: Created container: ctr0
  Normal  Started    13s   kubelet            spec.containers{ctr0}: Started container ctr0
  Normal  Pulled     13s   kubelet            spec.containers{ctr1}: Container image "mcr.microsoft.com/azuredocs/samples-tf-mnist-demo:gpu" already present on machine
  Normal  Created    13s   kubelet            spec.containers{ctr1}: Created container: ctr1
  Normal  Started    13s   kubelet            spec.containers{ctr1}: Started container ctr1

k logs -f pod --all-containers
GPU 0: NVIDIA A10-4Q (UUID: GPU-920b9300-69c2-11f1-9f17-62dd377f7a08)
GPU 0: NVIDIA A10-4Q (UUID: GPU-920b9300-69c2-11f1-9f17-62dd377f7a08)

Which issue(s) this PR fixes:

Fixes #

[Copilot]

Pull request overview

Copilot reviewed 13 out of 13 changed files in this pull request and generated 4 comments.

Comments suppressed due to low confidence (1)

parts/linux/cloud-init/artifacts/cse_helpers.sh:80

• ERR_DRA_DRIVER_START_FAIL is assigned exit code 124, but this file explicitly documents that 124 is the standard exit code returned by the timeout command when it times out (without --preserve-status). Reusing 124 will make failures ambiguous in logs/diagnostics and can break any tooling that interprets 124 as a timeout.

ERR_ENABLE_MANAGED_GPU_EXPERIENCE=123 # Error confguring managed GPU experience
ERR_DRA_DRIVER_START_FAIL=124 # dra-driver-nvidia-gpu could not be started by systemctl

# 123 is free for use

# Error code 124 is returned when a `timeout` command times out, and --preserve-status is not specified: https://man7.org/linux/man-pages/man1/timeout.1.html
ERR_VHD_BUILD_ERROR=125 # Reserved for VHD CI exit conditions

[Copilot]

Pull request overview

Copilot reviewed 13 out of 13 changed files in this pull request and generated 5 comments.

[Copilot]

Pull request overview

Copilot reviewed 13 out of 13 changed files in this pull request and generated 2 comments.

Comments suppressed due to low confidence (1)

parts/linux/cloud-init/artifacts/cse_helpers.sh:79

• ERR_DRA_DRIVER_START_FAIL is assigned exit code 124, but this file explicitly documents 124 as the default exit status from the timeout command. Reusing 124 for a CSE-specific failure makes it hard to distinguish an actual DRA failure from a generic timeout, and conflicts with the existing reservation comment.

ERR_ENABLE_MANAGED_GPU_EXPERIENCE=123 # Error confguring managed GPU experience
ERR_DRA_DRIVER_START_FAIL=124 # dra-driver-nvidia-gpu could not be started by systemctl

# 123 is free for use

# Error code 124 is returned when a `timeout` command times out, and --preserve-status is not specified: https://man7.org/linux/man-pages/man1/timeout.1.html

[Copilot]

Pull request overview

Copilot reviewed 13 out of 13 changed files in this pull request and generated 3 comments.

[Copilot]

Pull request overview

Copilot reviewed 13 out of 13 changed files in this pull request and generated 2 comments.

[Copilot]

Pull request overview

Copilot reviewed 13 out of 13 changed files in this pull request and generated 2 comments.