NVIDIA / NVIDIA/nvidia-container-toolkit
[Question]: Intended Runtime Options Override Mechanism
Nobody has claimed this yet.
- Dominant language
- Go
- Stars
- 4.6k
- Forks
- 592
- Avg merge
- 1d 33m
- Merged PRs (30d)
- 61
Description
Question:
what is the intended mechanism today for setting containerd runtime overrides when using the tookit installer?
asking since there does still appear to be a --runtime-config-override flag, but seems to be an unintended artifact of https://github.com/NVIDIA/nvidia-container-toolkit/pull/722 which followed its removal in a previous PR https://github.com/NVIDIA/nvidia-container-toolkit/pull/643#discussion_r1780736399 with rationale that it was unnecessary complexity.
https://github.com/NVIDIA/nvidia-container-toolkit/blob/357b970814261fb17d9b4991b1d2636bce71d442/cmd/nvidia-ctk-installer/container/runtime/containerd/containerd.go#L78-L84
the "inheritance" mechanism which reads from the default or fallback runtime block works sufficiently well, but is tied closer to node infra/provisioning, which doesn't feel ideal for dynamic installations.
hoping to either re-introduce the --runtime-config-override feature or note down a recommended stance on how this should be done (and remove the dead code). maybe externally-managed containerd drop-in configs are the right way to go but then thats implicitly tied to the chosen NVIDIA_RUNTIME_NAME and whatnot.
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start with cmd/nvidia-ctk-installer/container/runtime/containerd/containerd.go around the referenced flag and pkg/config/engine/containerd/config.go around the default or fallback runtime logic. Compare the current installer behavior with the linked PR discussions and determine whether runtime overrides should be supported or documented through another mechanism. Done means the project has an agreed, documented approach and the dead code or missing behavior is resolved.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- go
- Domain
- devops, infrastructure
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Quiet
- Clarity
- Mostly clear
- Newbie friendliness
- 35/100