Skip to content

[Bug]: Published 580.126.20-rhel10.0 image runs real dnf in DTK mode - predates the dnf stub present on main #985

Description

@aqeelat

Describe the bug

main (rhel10/ocp_dtk_entrypoint L182) stubs dnftrue before the vendored nvidia-driver build, making DTK builds repo/entitlement-free. The published 580.126.20-rhel10.0 image does not contain this stub: DTK-mode builds still run real dnf install kernel-headers/-devel/kernel-$(uname -r), which fails without RHEL entitlements or a custom repoConfig repo.

To Reproduce

DTK-mode deployment with digest sha256:869385f84684eb2b6f24c3fb48fbd468107c34bf589e2809ec9e9e3af7a68b64:

+ dnf -q -y --releasever=10 --setopt=install_weak_deps=False install kernel-6.12.0-233.el10.x86_64
Error: Unable to find a match: kernel-6.12.0-233.el10.x86_64

Real repo-resolution output confirms the stub is absent; newer tags' script layers contain it.

Expected behavior

Published images built from current scripts; ideally clarity on the GitHub-mirror ↔ image-build-source relationship so script verification against main is meaningful. Also: the 580.126.20-rhel10.0 tag has been removed from nvcr.io (digest pins still pull) — consumers of that line have no path forward.

Environment

  • gpu-driver-container source: digest sha256:869385f84684eb2b6f24c3fb48fbd468107c34bf589e2809ec9e9e3af7a68b64 (580.126.20-rhel10.0); comparison tag 580.178.04-rhel10.0
  • NVIDIA Driver Version: 580.126.20
  • Host OS: CentOS Stream CoreOS 10 (SCOS)
  • Kernel Version: 6.12.0-233.el10.x86_64
  • Container Runtime: CRI-O 1.34
  • CPU Architecture: x86_64
  • GPU Model: NVIDIA L4
  • Kubernetes Distro: OKD 4.22.0-okd-scos.6
  • GPU Operator version: v26.3.3

Metadata

Metadata

Assignees

Labels

No labels
No labels

Type

No type

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions