Skip to content

Add nvidia DRA driver GPU service - #4926

Merged
maherthomsi merged 4 commits into
bottlerocket-os:developfrom
maherthomsi:dra-changes
Sep 19, 2026
Merged

maherthomsi merged 4 commits into
bottlerocket-os:developfrom
maherthomsi:dra-changes

Conversation

@maherthomsi

@maherthomsi maherthomsi commented Sep 3, 2026 •

Copy link
Copy Markdown
Contributor

Dependent on bottlerocket-os/bottlerocket-core-kit#1035 and bottlerocket-os/bottlerocket-settings-sdk#148

Description of changes:
Integrates the NVIDIA GPU DRA driver into the NVIDIA Kubernetes variants and enables it via the new settings. Specifically:

  • Settings plugin: Wires the KubeletDraDriversV1 settings (from settings-sdk PR #148) into the aws-k8s-nvidia settings plugin, making settings.kubelet-dra-drivers.nvidia available on the NVIDIA variants.
  • Shared defaults: Adds the nvidia-dra-driver-gpu host service defaults. This adds in the rendered ExecStart, restart behavior, and the settings.kubelet-dra-drivers.nvidia.enabled opt-in toggle. We then symlink them into the aws-k8s-1.33-nvidia and aws-k8s-1.36-nvidia variants.
  • Variants: Includes the nvidia-dra-driver-gpu package (from core-kit PR #1035).
  • Migration: Adds an AddSettingsMigration for the new settings.kubelet-dra-drivers.nvidia.enabled setting so datastores can migrate cleanly.

Testing done:

  • Build 1.34 variant
  • Build 1.36 variant
  • Build 1.37 variant
  • FIPS variant
  • Verified the settings.kubelet-dra-drivers.nvidia.enabled toggle starts/stops the gpu-kubelet-plugin host service.
  • Confirmed the settings migration applies across the version boundary.

Terms of contribution:

By submitting this pull request, I agree that this contribution is dual-licensed under the terms of both the Apache License, version 2.0, and the MIT license.

@maherthomsi
maherthomsi force-pushed the dra-changes branch 3 times, most recently from 43267df to e08f576 Compare September 16, 2026 21:09
@maherthomsi
maherthomsi marked this pull request as ready for review September 16, 2026 21:28
@maherthomsi
maherthomsi force-pushed the dra-changes branch 3 times, most recently from eb6f413 to 3830dba Compare September 17, 2026 22:56

@ginglis13 ginglis13 left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

missing k8s-1.37-nvidia support

Comment thread variants/aws-k8s-1.34-nvidia/Cargo.toml
maherthomsi added a commit to maherthomsi/bottlerocket that referenced this pull request Sep 18, 2026
Comment thread sources/Cargo.toml Outdated
@maherthomsi
maherthomsi force-pushed the dra-changes branch 2 times, most recently from 881c967 to 50f74cb Compare September 19, 2026 00:44
@maherthomsi
maherthomsi force-pushed the dra-changes branch 2 times, most recently from 7b36e13 to 4978501 Compare September 19, 2026 01:57
Include the nvidia-dra-driver-gpu package in the aws-k8s-1.34-nvidia
through aws-k8s-1.37-nvidia variants and their -nvidia-fips
counterparts.

Signed-off-by: Maher Homsi <maherhom@amazon.com>
Add the nvidia-dra-driver-gpu host service defaults (rendered ExecStart,
restart behaviour, and the settings.kubelet-dra-drivers.nvidia.enabled
opt-in toggle) and symlink them into the aws-k8s-1.33-nvidia and
aws-k8s-1.36-nvidia variants.

Signed-off-by: Maher Homsi <maherhom@amazon.com>
Wire the KubeletDraDriversV1 settings into the aws-k8s-nvidia plugin so
the settings.kubelet-dra-drivers.nvidia settings are available on the
NVIDIA variants.

Signed-off-by: Maher Homsi <maherhom@amazon.com>
Add an AddSettingsMigration for the new
settings.kubelet-dra-drivers.nvidia.enabled setting so datastores can be
migrated across the 1.63.0 -> 1.64.0 boundary.

Signed-off-by: Maher Homsi <maherhom@amazon.com>
@maherthomsi
maherthomsi merged commit 576d093 into bottlerocket-os:develop Sep 19, 2026
3 checks passed
@maherthomsi
maherthomsi deleted the dra-changes branch September 19, 2026 02:39
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants