Fix NVIDIA container toolkit bug in all backends#2877
Merged
Conversation
Use cgroupfs as a Docker cgroup driver on all backends by default to work around an NVIDIA container toolkit bug where the container looses access to the GPU. The patch to `/etc/docker/daemon.json` is automatically applied in all VM-based backends if `/etc/docker/daemon.json` exists, has the NVIDIA runtime, does not explicitly set another cgroup driver, and `jq` is installed. This is not the case in Nebius. Lambda, and CUDO, so they still need custom code to apply the workaround - either installing `jq` or just writing a hardcoded `/etc/docker/daemon.json` that is known to work on this backend.
un-def
approved these changes
Jul 7, 2025
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Use cgroupfs as a Docker cgroup driver on all
backends by default to work around an NVIDIA
container toolkit bug where the container looses
access to the GPU.
The patch to
/etc/docker/daemon.jsonisautomatically applied in all VM-based backends if
/etc/docker/daemon.jsonexists, has the NVIDIAruntime, does not explicitly set another cgroup
driver, and
jqis installed. This is not thecase in Nebius. Lambda, and CUDO, so they still
need custom code to apply the workaround - either
installing
jqor just writing a hardcoded/etc/docker/daemon.jsonthat is known to work on this backend.#2860