Skip to content

[Issue]: Missing amdgpu package repo for CoreOS 9.8 (OCP v4.22) #550

Description

@LaVLaS

Problem Description

Deploy amd-gpu-operator >= v1.5.z on any OCP 4.22 cluster with a DeviceConfig that builds the rpm from amdgpu package repository under the default https://repo.radeon.com/amdgpu/.

The kernel driver builds fail while attempt to pull from the amdgpu radeon repo url due to the new DockerfileTemplate.rpm.coreos attempt to fall back to the non-existent OS '*/el/9/main/x86_64/` package repo

[1/2] STEP 8/9: RUN dnf clean all && cat /etc/yum.repos.d/amdgpu.repo && dnf install amdgpu-dkms -y && depmod ${KERNEL_VERSION} && find /lib/modules/${KERNEL_VERSION} -name "*.ko.xz" -exec xz -d {} \; && depmod ${KERNEL_VERSION}
(microdnf:8): librhsm-WARNING **: 18:52:44.002: Found 0 entitlement certificates
(microdnf:8): librhsm-WARNING **: 18:52:44.004: Found 0 entitlement certificates
Complete.
[amdgpu]
name=amdgpu
baseurl=https://repo.radeon.com/amdgpu/30.20.1/el/9/main/x86_64/
enabled=1
priority=50
gpgcheck=1
gpgkey=https://repo.radeon.com/rocm/rocm.gpg.key
...
error: cannot update repo 'amdgpu': Cannot download repomd.xml: Cannot download repodata/repomd.xml: All mirrors were tried; Last error: Status code: 404 for https://repo.radeon.com/amdgpu/30.20.1/el/9/main/x86_64/repodata/repomd.xml (IP: 23.212.251.199)
error: build error: building at STEP "RUN dnf clean all && cat /etc/yum.repos.d/amdgpu.repo && dnf install amdgpu-dkms -y && depmod ${KERNEL_VERSION} && find /lib/modules/${KERNEL_VERSION} -name "*.ko.xz" -exec xz -d {} \; && depmod ${KERNEL_VERSION}": while running runtime: exit status 1

Operating System

RHEL CoreOS 9.8 (Plow) / OCP v4.22

CPU

N/A

GPU

N/A

ROCm Version

All amdgpu versions

ROCm Component

No response

Steps to Reproduce

  1. Deploy release candidate for OCP v4.22
  2. Create DeviceConfig that attempt to build the kernel driver from the amdgpu-dkms RPM in the amdgpu package repo
  3. Build fails

(Optional for Linux users) Output of /opt/rocm/bin/rocminfo --support

No response

Additional Information

The workaround to use useSourceImage: true in DeviceConfig fails for a similar reason since https://hub.docker.com/r/rocm/amdgpu-driver/tags has no tags for coreos-9.8-*

Metadata

Metadata

Assignees

Labels

No labels
No labels

Type

No type
No fields configured for issues without a type.

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions