NPU Operator

Feature Introduction

Kubernetes provides access to special hardware resources (such as Ascend NPU) through device plugins. However, configuring and managing nodes with these hardware resources requires configuring multiple software components (such as drivers, container runtimes, or other libraries), which are complex, difficult to install, and error-prone. NPU Operator uses the Operator Framework in Kubernetes to automatically manage all software components required to configure Ascend devices. These components include Ascend drivers and firmware, enabling full-process cluster operation, supporting cluster job scheduling, operation and maintenance monitoring, fault recovery, and other functions of the MindCluster device plugin. By installing the corresponding components, NPU resource management, optimized scheduling of workloads, and containerization support for training and inference tasks can be achieved, enabling AI jobs to be deployed and run in container form on NPU devices.

Installation Deployment

For deployment methods, please refer to the NPU Operator User Guide.

Local Build

Image Builden

Build Parameters

• GOPRIVATE: Configure Go language private repository, equivalent to the GOPRIVATE environment variable.
• COMMIT: Hash value of the current git commit.
• VERSION: Component version.
• SOURCE_DATE_EPOCH: Timestamp of the image rootfs.

Build Commands

• Build and push to specified OCI repository.

  Using docker

  docker buildx build . -f <path/to/dockerfile> \
           -o type=image,name=<oci/repository>:<tag>,oci-mediatypes=true,rewrite-timestamp=true,push=true \
      --platform=linux/amd64,linux/arm64 \
      --provenance=false \
      --build-arg=GOPRIVATE=gopkg.openfuyao.cn \
      --build-arg=COMMIT=$(git rev-parse HEAD) \
      --build-arg=VERSION=0.0.0-latest \
      --build-arg=SOURCE_DATE_EPOCH=$(git log -1 --pretty=%ct)
  

  Using nerdctl

  nerdctl build . -f <path/to/dockerfile> \
      -o type=image,name=<oci/repository>:<tag>,oci-mediatypes=true,rewrite-timestamp=true,push=true \
      --platform=linux/amd64,linux/arm64 \
      --provenance=false \
      --build-arg=GOPRIVATE=gopkg.openfuyao.cn \
      --build-arg=COMMIT=$(git rev-parse HEAD) \
      --build-arg=VERSION=0.0.0-latest \
      --build-arg=SOURCE_DATE_EPOCH=$(git log -1 --pretty=%ct)
  

  Where <path/to/dockerfile> is the Dockerfile path, <oci/repository> is the image address, and <tag> is the image tag.

• Build and export OCI Layout to local tarball.

  Using docker

  docker buildx build . -f <path/to/dockerfile> \
      -o type=oci,name=<oci/repository>:<tag>,dest=<path/to/oci-layout.tar>,rewrite-timestamp=true \
      --platform=linux/amd64,linux/arm64 \
      --provenance=false \
      --build-arg=GOPRIVATE=gopkg.openfuyao.cn \
      --build-arg=COMMIT=$(git rev-parse HEAD) \
      --build-arg=VERSION=0.0.0-latest \
      --build-arg=SOURCE_DATE_EPOCH=$(git log -1 --pretty=%ct)
  

  Using nerdctl

  nerdctl build . -f <path/to/dockerfile> \
      -o type=oci,name=<oci/repository>:<tag>,dest=<path/to/oci-layout.tar>,rewrite-timestamp=true \
      --platform=linux/amd64,linux/arm64 \
      --provenance=false \
      --build-arg=GOPRIVATE=gopkg.openfuyao.cn \
      --build-arg=COMMIT=$(git rev-parse HEAD) \
      --build-arg=VERSION=0.0.0-latest \
      --build-arg=SOURCE_DATE_EPOCH=$(git log -1 --pretty=%ct)
  

  Where <path/to/dockerfile> is the Dockerfile path, <oci/repository> is the image address, <tag> is the image tag, and path/to/oci-layout.tar is the tar package path.

• Build and export image rootfs to local directory.

  Using docker

  docker buildx build . -f <path/to/dockerfile> \
      -o type=local,dest=<path/to/output>,platform-split=true \
      --platform=linux/amd64,linux/arm64 \
      --provenance=false \
      --build-arg=GOPRIVATE=gopkg.openfuyao.cn \
      --build-arg=COMMIT=$(git rev-parse HEAD) \
      --build-arg=VERSION=0.0.0-latest
  

  Using nerdctl

  nerdctl build . -f <path/to/dockerfile> \
      -o type=local,dest=<path/to/output>,platform-split=true \
      --platform=linux/amd64,linux/arm64 \
      --provenance=false \
      --build-arg=GOPRIVATE=gopkg.openfuyao.cn \
      --build-arg=COMMIT=$(git rev-parse HEAD) \
      --build-arg=VERSION=0.0.0-latest
  

  Where <path/to/dockerfile> is the Dockerfile path and path/to/output is the local directory path.

Helm Chart Build

• Package Helm Chart.

  helm package <path/to/chart> -u \
      --version=0.0.0-latest \
      --app-version=openFuyao-v25.09
  

  Where <path/to/chart> is the Chart folder path.

• Push Chart package to specified OCI repository.

  helm push <path/to/chart.tgz> oci://<oci/repository>:<tag>
  

  Where <path/to/chart.tgz> is the Chart package path, <oci/repository> is the Chart package push address, and <tag> is the Chart package tag.