Skip to content

fix: correct GB300 network boot behavior#104

Merged
krish-nvidia merged 1 commit into
mainfrom
krish/gb300-boot-order-fix
Jul 15, 2026
Merged

fix: correct GB300 network boot behavior#104
krish-nvidia merged 1 commit into
mainfrom
krish/gb300-boot-order-fix

Conversation

@krish-nvidia

@krish-nvidia krish-nvidia commented Jul 15, 2026

Copy link
Copy Markdown

Addresses: NVIDIA/infra-controller#3574

  • Put the selected DPU first in the GB300 boot order.
  • Disable all other network boot options.
  • Set GB300 PXE retries to 0.
  • Update boot-order validation for the new configuration.

Tested on a real lenovo GB300.

Disable competing network boot options and remove PXE retries so the selected DPU remains the sole network boot target.
@krish-nvidia krish-nvidia self-assigned this Jul 15, 2026
@krish-nvidia krish-nvidia merged commit 6728148 into main Jul 15, 2026
2 checks passed

@martinraumann martinraumann left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This looks aligned with the GB300 boot-loop issue we saw.

Non-blocking recommendation: add regression coverage for the GB300-specific behavior here.

In particular:

  • DPU boot option already first, but competing Pxe/UefiHttp options still enabled
  • is_boot_order_setup returning false until non-target network boot options are disabled
  • Lenovo GB300 LEM0001 PXE retry count set to 0

The behavior is subtle because the previous early return looked correct unless the enabled boot options were also considered.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants