Building Red Hat 8.4 Offline Yum Source¶
The AI platform comes with pre-installed CentOS v7.9 and GPU Operator offline packages with kernel v3.10.0-1160. For other OS types or nodes with different kernels, users need to manually build the offline yum source.
This guide explains how to build an offline yum source package for Red Hat 8.4 based on any node in the Global cluster. It also demonstrates how to use it during the installation of the GPU Operator by specifying the RepoConfig.ConfigMapName parameter.
Prerequisites¶
- The user has already installed the addon offline package v0.12.0 or higher on the platform.
- The OS of the cluster nodes where the GPU Operator will be deployed must be Red Hat v8.4, and the kernel version must be identical.
- Prepare a file server that can communicate with the cluster network where the GPU Operator will be deployed, such as Nginx or MinIO.
- Prepare a node that can access the internet, the cluster where the GPU Operator will be deployed, and the file server. Ensure that Docker is already installed on this node.
- The nodes in the Global cluster must be Red Hat 8.4 4.18.0-305.el8.x86_64.
Procedure¶
This guide uses a node with Red Hat 8.4 4.18.0-305.el8.x86_64 as an example to demonstrate how to build an offline yum source package for Red Hat 8.4 based on any node in the Global cluster. It also explains how to use it during the installation of the GPU Operator by specifying the RepoConfig.ConfigMapName parameter.
Step 1: Download the Yum Source from the Bootstrap Node¶
Perform the following steps on the master node of the Global cluster.
-
Use SSH or any other method to access any node in the Global cluster and run the following command:
The expected output should resemble the following:
[extension-0] baseurl = http://10.5.14.200:9000/kubean/redhat/$releasever/os/$basearch gpgcheck = 0 name = kubean extension 0 [extension-1] baseurl = http://10.5.14.200:9000/kubean/redhat-iso/$releasever/os/$basearch/AppStream gpgcheck = 0 name = kubean extension 1 [extension-2] baseurl = http://10.5.14.200:9000/kubean/redhat-iso/$releasever/os/$basearch/BaseOS gpgcheck = 0 name = kubean extension 2 -
Create a folder named redhat-base-repo under the root directory:
-
Download the RPM packages from the yum source to your local machine:
Download the RPM packages from extension-1 :
Download the RPM packages from extension-2 :
Step 2: Download the elfutils-libelf-devel-0.187-4.el8.x86_64.rpm Package¶
Perform the following steps on a node with internet access. Before proceeding, ensure that there is network connectivity between the node with internet access and the master node of the Global cluster.
-
Run the following command on the node with internet access to download the elfutils-libelf-devel-0.187-4.el8.x86_64.rpm package:
-
Transfer the elfutils-libelf-devel-0.187-4.el8.x86_64.rpm package from the current directory to the node mentioned in step 1:
For example:
Step 3: Generate the Local Yum Repository¶
Perform the following steps on the master node of the Global cluster mentioned in Step 1.
-
Enter the yum repository directories:
-
Generate the repository index for the directories:
You have now generated the offline yum source named redhat-base-repo for kernel version 4.18.0-305.el8.x86_64 .
Step 4: Upload the Local Yum Repository to the File Server¶
In this example, we will use Minio, which is built-in as the file server in the bootstrap node. However, you can choose any file server that suits your needs. Here are the details for Minio:
- Access URL:
http://10.5.14.200:9000(usually the {bootstrap-node-IP} + {port-9000}) - Login username: rootuser
-
Login password: rootpass123
-
On the current node, establish a connection between the local mc command-line tool and the Minio server by running the following command:
For example:
The expected output should be similar to:
The mc command-line tool is provided by the Minio file server as a client command-line tool. For more details, refer to the MinIO Client documentation.
-
Create a bucket named redhat-base in the current location:
The expected output should be similar to:
-
Set the access policy of the redhat-base bucket to allow public downloads so that it can be accessed during the installation of the GPU Operator:
The expected output should be similar to:
-
Copy the offline yum repository files ( redhat-base-repo ) from the current location to the Minio server's minio/redhat-base bucket:
Step 5: Create a ConfigMap to Store Yum Repository Information in the Cluster¶
Perform the following steps on the control node of the cluster where you will deploy the GPU Operator.
-
Run the following command to create a file named redhat.repo , which specifies the configuration information for the yum repository storage:
# The file name must be redhat.repo, otherwise it won't be recognized when installing gpu-operator cat > redhat.repo << EOF [extension-0] baseurl = http://10.5.14.200:9000/redhat-base/redhat-base-repo/Packages # The file server address where the yum source is stored in Step 1 gpgcheck = 0 name = kubean extension 0 [extension-1] baseurl = http://10.5.14.200:9000/redhat-base/redhat-base-repo/Packages # The file server address where the yum source is stored in Step 1 gpgcheck = 0 name = kubean extension 1 EOF -
Based on the created redhat.repo file, create a configmap named local-repo-config in the gpu-operator namespace:
The expected output should be similar to:
The local-repo-config configuration file is used to provide the value for the RepoConfig.ConfigMapName parameter during the installation of the GPU Operator. You can choose a different name for the configuration file.
-
View the contents of the local-repo-config configuration file:
You have successfully created the offline yum source configuration file for the cluster where the GPU Operator will be deployed. You can use it by specifying the RepoConfig.ConfigMapName parameter during the offline installation of the GPU Operator.