Skip to content

[Bug]: Missing H200 NVL (0x233B10DE) in default-mig-parted-config for all-1g.35gb #3272

Open

Description

@linlingo1222-del

Description

In default-mig-parted-config, the all-1g.35gb profile is missing the PCI device ID 0x233B10DE for NVIDIA H200 NVL (it currently only includes 0x233510DE for H200 SXM and 0x31C210DE for GB300).

Notably, 0x233B10DE is already listed alongside 0x233510DE in the all-balanced profile, and H200 NVL hardware supports four 1g.35gb MIG instances per GPU.

Because of this omission, applying nvidia.com/mig.config=all-1g.35gb on an H200 NVL node results in a silent no-op: nvidia-mig-manager skips all GPUs, exits with code 0, and falsely marks nvidia.com/mig.config.state=success while MIG remains disabled (or retains the previous MIG configuration).

Environment

  • GPU: 2x NVIDIA H200 NVL (PCI Device ID: 0x233B10DE)
  • NVIDIA Driver: 595.71.05 (CUDA 13.2)
  • GPU Operator Version: <v25.10.1>
  • Kubernetes Version: v1.33.5
  • Container Runtime: containerd://2.1.4
  • OS / Kernel: Ubuntu 24.04.3 LTS / 6.8.0-71-generic

Steps to Reproduce

  1. Deploy NVIDIA GPU Operator on a node equipped with NVIDIA H200 NVL (0x233B10DE).
  2. Label the node with nvidia.com/mig.config=all-1g.35gb:
    kubectl label nodes <node-name> [nvidia.com/mig.config=all-1g.35gb](https://nvidia.com/mig.config=all-1g.35gb) --overwrite
  3. Check the node labels and nvidia-smi output.

Actual Behavior

  • Node label reports nvidia.com/mig.config.state=success.
  • nvidia-smi shows MIG mode is still Disabled (no 1g.35gb instances created).
  • nvidia-mig-manager logs show mig-parted skipping the H200 NVL GPUs due to the device-filter mismatch:
    time="2026-11-19T23:10:14Z" level=debug msg="Applying MIG device configuration..."
    time="2026-11-19T23:10:14Z" level=debug msg="Walking MigConfig for (device-filter=[0x233510DE], devices=all)"
    time="2026-11-19T23:10:14Z" level=debug msg="Walking MigConfig for (device-filter=[0x31C210DE], devices=all)"
    time="2026-11-19T23:10:14Z" level=debug msg="Running apply-exit hook"
    MIG configuration applied successfully
    

Expected Behavior

all-1g.35gb should match NVIDIA H200 NVL (0x233B10DE) and partition each GPU into 4x 1g.35gb MIG instances.

Proposed Fix

Add "0x233B10DE" to the device-filter under all-1g.35gb in default-mig-parted-config:

  all-1g.35gb:
    # H200-141GB, H200 NVL
    - device-filter: ["0x233510DE", "0x233B10DE"]
      devices: all
      mig-enabled: true
      mig-devices:
        "1g.35gb": 4
    # GB300
    - device-filter: ["0x31C210DE"]
      devices: all
      mig-enabled: true
      mig-devices:
        "1g.35gb": 7

Activity

  1. added
    bugIssue/PR to expose/discuss/fix a bug
    needs-triageissue or PR has not been assigned a priority-px label
    on Nov 20, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    bugIssue/PR to expose/discuss/fix a bugneeds-triageissue or PR has not been assigned a priority-px label

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions