nvidia-ctk can't set nvidia-container-runtime.runtimes

[work-chausat]

[root@51-cv-gpu-node-01 ~]# nvidia-ctk config --set nvidia-container-runtime.runtimes=docker-runc, runc
ERRO[0000] invalid --set option nvidia-container-runtime.runtimes=docker-runc,: unsupported type for [nvidia-container-runtime.runtimes docker-runc,] ([docker-runc runc])

[root@51-cv-gpu-node-01 ~]# nvidia-ctk config --set nvidia-container-runtime.runtimes="docker-runc, runc"
ERRO[0000] invalid --set option nvidia-container-runtime.runtimes=docker-runc, runc: unsupported type for [nvidia-container-runtime.runtimes docker-runc, runc] ([docker-runc runc])

[elezar]

Thanks for the issue. The original implementation was quite brittle and I have reworked it as https://gitlab.com/nvidia/container-toolkit/container-toolkit/-/merge_requests/500

I will finalize the MR and include this fix in an upcoming release.

[work-chausat]

Thanks for the issue. The original implementation was quite brittle and I have reworked it as https://gitlab.com/nvidia/container-toolkit/container-toolkit/-/merge_requests/500

I will finalize the MR and include this fix in an upcoming release.

wonderful job!

[elezar]

Hi @work-chausat. We have released NVIDIA Container Toolkit v1.15.0-rc.1 to our experimental repos. If you're able to, please check that this resolves the issue that you're seeing here.

(We aslo have https://gitlab.com/nvidia/cloud-native/cnt-docs/-/merge_requests/354 open to update our documentation to include instructions to enable the experimental repos).

[work-chausat]

@elezar Thanks for your fix. v1.15.0-rc.1 works.
nvidia-ctk config -i --set nvidia-container-runtime.runtimes="runc.sh,docker-runc,runc" got:
runtimes = ["runc.sh", "docker-runc", "runc"]

nvidia-ctk config -i --set nvidia-container-runtime.runtimes="runc.sh, docker-runc, runc" got:
runtimes = ["runc.sh", " docker-runc", " runc"]

Do we need trim?

[github-actions]

This issue is stale because it has been open 90 days with no activity. This issue will be closed in 30 days unless new comments are made or the stale label is removed. To skip these checks, apply the "lifecycle/frozen" label.

[ArangoGutierrez]

I am closing this issue. If any problems persist using the latest version of the NVIDIA Container Toolkit, please open a new issue.