feat: add MigStrategy and EnableManagedGPU NBC variables for managed GPU experience #7695
Conversation
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
ai prompt for claude
This reverts commit 17b74b3.
should not enabled managed gpu in the original test
Reapply "Update work-prompt.md" This reverts commit 28f0bd0.
- Add mig_strategy field to gpu_config.proto with field number 8 - Add NVIDIA_MIG_STRATEGY environment variable mapping in parser.go - Add MigStrategy field to NodeBootstrappingConfiguration in types.go - Add GetMigStrategy template function in baker.go - Add NVIDIA_MIG_STRATEGY variable in cse_cmd.sh - Add comprehensive tests for Mixed, Single, and None strategy values
…_MIG_STRATEGY environment variable instead of hardcoded 'single' strategy.
- Add conditional logic to map RP values to device plugin flags:
* "Mixed" -> --mig-strategy mixed
* "Single"/"None"/empty -> --mig-strategy single (default)
- Add comprehensive comments explaining MIG strategy behavior
- Maintain backward compatibility with safe defaults
- Only use "mixed" when explicitly specified for safety
This reverts commit e9433c9.
ganeshkumarashok
requested review from
AbelHu,
Devinwong,
YaoC,
awesomenix,
cameronmeissner,
djsly,
ganeshkumarashok,
juan-lee,
junjiezhang1997,
lilypan26,
phealy,
r2k1,
timmy-wright and
zachary-bailey
as code owners
ganeshkumarashok
changed the title
ganeshkumarashok
approved these changes
Jan 26, 2026
…nshum/managed-gpu/mig-mode
lilypan26
approved these changes
Jan 27, 2026
awesomenix
approved these changes
Jan 27, 2026
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
What this PR does / why we need it:
Overview
This change adds a new MigStrategy variable to the NodeBootstrappingConfiguration (NBC) to support configurable MIG (Multi-Instance GPU) strategies for the nvidia-device-plugin.
Previously, when
GPUInstanceProfilewas set, the MIG strategy was hardcoded tosingle. This PR enables the Resource Provider (RP) to explicitly pass the desired MIG strategy via NBC, allowing users to control how MIG devices are exposed to Kubernetes.Supported MIG Strategies
None
MIG is not enabled.
Single (default)
All MIG devices are exposed as generic resources:
nvidia.com/gpuMixed
MIG devices are exposed with specific resource types, such as:
nvidia.com/mig-1g.5gbChanges
mig_strategyfield to the GPU config proto (scriptless NBC path)MigStrategyfield toNodeBootstrappingConfiguration(legacy NBC path)NVIDIA_MIG_STRATEGYenvironment variable to CSEstartNvidiaManagedExpServices()to configure thenvidia-device-pluginusing the new MIG strategy variableBackward Compatibility
Singleto preserve existing behaviorMixedis only used when explicitly specified via NBCWhich issue(s) this PR fixes:
Fixes #