Skip to content

Installing and Using Metax GPU Components

This section provides guidance on installing the Metax components—gpu-extensions, gpu-operator, etc.—as well as instructions for using Metax GPUs in both full-card and vGPU modes.

Prerequisites

  1. Download and install the required tar package from the Metax Software Center. This document uses metax-gpu-k8s-package.0.7.10.tar.gz as an example.
  2. Prepare a basic Kubernetes environment.

Component Overview

Metax provides two Helm chart packages: metax-extensions and gpu-operator. Choose which to install based on your usage scenario.

  1. metax-extensions: Contains the gpu-device and gpu-label components. When using this approach, application container images must be built based on the MXMACA® base image. This solution only supports full GPU usage.
  2. gpu-operator: Contains components such as gpu-device, gpu-label, driver-manager, container-runtime, and operator-controller. With this approach, application images do not need to include the MXMACA® SDK, and it supports both full GPU and vGPU modes.

Installation Steps

  1. Extract the contents from the tar file:

    From /home/metax/metax-docs/k8s/metax-gpu-k8s-package.0.7.10.tar.gz, extract:

    • deploy-gpu-extensions.yaml – deployment YAML file
    • metax-gpu-extensions-0.7.10.tgz, metax-operator-0.7.10.tgz – Helm chart packages
    • metax-k8s-images.0.7.10.run – offline image archive

  2. Check if the driver is already installed:

    $ lsmod | grep metax 
metax 1605632 0 
ttm 86016 3 drm_vram_helper,metax,drm_ttm_helper 
drm 618496 7 drm_kms_helper,drm_vram_helper,ast,metax,drm_ttm_helper,ttm
    • If no output appears, the driver is not installed.
    • If output appears, the driver is already installed.
    • When using gpu-operator, it is not recommended to pre-install the MXMACA kernel driver on worker nodes. If already installed, you do not need to uninstall it.

  3. Install the driver

Installing gpu-extensions

  1. Push images:

    tar -xf metax-gpu-k8s-package.0.7.10.tar.gz
./metax-k8s-images.0.7.10.run push {registry}/metax

  2. Push Helm charts:

    helm plugin install https://github.com/chartmuseum/helm-push
helm repo add --username rootuser --password rootpass123 metax http://172.16.16.5:8081
helm cm-push metax-operator-0.7.10.tgz metax
helm cm-push metax-gpu-extensions-0.7.10.tgz metax

  3. Install metax-gpu-extensions on the DCE 5.0 platform.

    After successful deployment, you can verify the resources on the nodes:

    Resource View

  4. Once configured, nodes will be labeled with Metax GPU:

    Metax Node Label

Installing gpu-operator

Known issues when installing gpu-operator:

  1. The images for metax-operator, gpu-label, gpu-device, and container-runtime must include the amd64 suffix.
  2. The metax-maca component image is not included in the metax-k8s-images.0.7.13.run package. You must manually download images like maca-mxc500-2.23.0.23-ubuntu20.04-x86_64.tar.xz, load them, and update the metax-maca component image accordingly.

  3. The metax-driver image needs to be downloaded from https://pub-docstore.metax-tech.com:7001 as k8s-driver-image.2.23.0.25.run. Then run:

    ./k8s-driver-image.2.23.0.25.run push {registry}/metax

    to push it to your image registry, and update the metax-driver image reference accordingly.

Using the GPU

After installation, you can use the Metax GPU in workloads. When enabling GPU support, be sure to select Metax GPU as the GPU type.

Using GPU

To check GPU usage, enter the container and run mx-smi.

GPU Usage

Comments