Fix kernel-devel mismatch causing NVIDIA driver load failure - #32
Conversation
Pin kernel-devel and kernel-headers to the running kernel version instead of installing the latest available from repos. This prevents DKMS from building the NVIDIA module for a different kernel than the one currently running. Root cause: When the guest VM boots with kernel 5.14.0-570.12.1 but dnf installs kernel-devel-5.14.0-570.136.1 (latest from repos), DKMS builds the nvidia module for 570.136.1. When the guest tries to load the module, it fails because the running kernel is 570.12.1. Impact: nvidia-smi returns rc=9 (NVML_ERROR_GPU_IS_LOST) because the driver can't load, breaking all GPU validation tests. Fix: Detect the running kernel with uname -r and explicitly install kernel-devel-$(uname -r) to match. Tested: Confirmed kernel mismatch on MIG CI job guest VMs at 192.168.122.222 (running 570.12.1, DKMS built for 570.136.1). Related-Issue: MIG vGPU CI job failures since Aug 2026 Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Enterprise Run ID: 📒 Files selected for processing (1)
Included review availability: Your plan provides up to 12 included reviews per hour; 11 remain after this review. 📝 WalkthroughSummary by CodeRabbit
WalkthroughThe setup task captures the running kernel version and installs DKMS dependencies that match it. ChangesRHEL MIG kernel dependencies
Estimated code review effort: 2 (Simple) | ~5 minutes Merge Risk: ⚪ Minimal · up to The change pins kernel development packages to the running kernel to prevent NVIDIA module mismatches; no actionable merge-blocking risk remains. Suggested reviewers: 🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
Full details: Docstring CoverageExplanation No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 0 files. (1 skipped: 1 unsupported.) ✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
|
lgtm |
Pin kernel-devel and kernel-headers to the running kernel version instead of installing the latest available from repos. This prevents DKMS from building the NVIDIA module for a different kernel than the one currently running.
Root cause: When the guest VM boots with kernel 5.14.0-570.12.1 but dnf installs kernel-devel-5.14.0-570.136.1 (latest from repos), DKMS builds the nvidia module for 570.136.1. When the guest tries to load the module, it fails because the running kernel is 570.12.1.
Impact: nvidia-smi returns rc=9 (NVML_ERROR_GPU_IS_LOST) because the driver can't load, breaking all GPU validation tests.
Fix: Detect the running kernel with uname -r and explicitly install kernel-devel-$(uname -r) to match.
Tested: Confirmed kernel mismatch on MIG CI job guest VMs at 192.168.122.222 (running 570.12.1, DKMS built for 570.136.1).