Fix removal of GPUs from nvlink partitions#337
Merged
tmcroberts97 merged 2 commits intoNVIDIA:mainfrom Feb 23, 2026
Merged
Conversation
Refactor the nvlink partition monitor so that it iterates through the nvlink_info of every instance, instead of the instance gpu config. The gpu config array will not be populated when a user attempts to detach GPUs, which meant the monitor had nothing to iterate over, so didn't remove the GPUs from the partition (or take any action at all in fact). Iterating over the nvlink_info array of GPUs fixes this, since they should always be present. Signed-off-by: Thomas McRoberts <tmcroberts@nvidia.com>
tmcroberts97
force-pushed
the
fix/partition-removal
branch
from
2b760e0 to
7b1ebbd
Compare
kensimon
reviewed
Feb 19, 2026
Contributor
kensimon
left a comment
kensimon
left a comment
There was a problem hiding this comment.
Just some readability nits and some potential simplification, otherwise lgtm.
rtamma-nv
reviewed
Feb 20, 2026
| gpu_config.logical_partition_id; | ||
| if db_logical_partition_id.is_none() { | ||
| // How can this happen? | ||
| tracing::error!( |
Contributor
There was a problem hiding this comment.
In this case the GPU is in an inconsistent state, and if tenant applies nvlink_config, we will never be able to sync it.
Should we take a corrective action of removing the GPU from the partition?
Contributor
Author
There was a problem hiding this comment.
Probably. I can make another MR with this change.
rtamma-nv
reviewed
Feb 20, 2026
Signed-off-by: Thomas McRoberts <tmcroberts@nvidia.com>
Contributor
Author
|
@kensimon thanks for the review! All changes made. |
rtamma-nv
approved these changes
Feb 20, 2026
kensimon
approved these changes
Feb 20, 2026
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
3 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.
Description
Refactor the nvlink partition monitor so that it iterates through the nvlink_info of every instance, instead of the instance gpu config. The gpu config array will not be populated when a user attempts to detach GPUs, which meant the monitor had nothing to iterate over, so didn't remove the GPUs from the partition (or take any action at all in fact).
Iterating over the nvlink_info array of GPUs fixes this, since they should always be present.
Type of Change
Related Issues (Optional)
Breaking Changes
Testing
Additional Notes