CLI Reference

nvcrectl Overview

View as Markdown

nvcrectl is the command-line interface for the Cluster Readiness Engine. It handles installation, certification lifecycle, workload execution, and reporting — without requiring direct kubectl access for most operations.

Install

While this repository is internal, release assets require authenticated API downloads, so fetch the installer with the gh CLI (gh auth login first):

$export NVCRECTL_VERSION=v0.1.0-rc.9
$gh release download "${NVCRECTL_VERSION}" --repo NVIDIA/cluster-readiness-engine \
>  --pattern installer --output - | bash -s -- -v "${NVCRECTL_VERSION}"
$nvcrectl --version

The installer also creates a kubectl-nvcre symlink so the CLI is available as a kubectl plugin (kubectl nvcre ...).

Command groups

Command groupPurpose
nvcrectl setupInstall and uninstall the controller and its dependencies
nvcrectl certificationRun, render, report, and list-categories for Certification resources
nvcrectl workloadrunRun, render, report, status, and cancel WorkloadRun resources
nvcrectl clusterInspect GPU nodes, platform, and network topology
nvcrectl workflowRender Workflow manifests offline with overrides applied

Global flags

These flags are available on subcommands that connect to a cluster:

FlagDescription
--kubeconfigPath to kubeconfig (defaults to ~/.kube/config)
--contextKubernetes context to use
--namespace / -nNamespace override (not available on all subcommands)

Shell completion

$nvcrectl completion bash >> ~/.bashrc
$nvcrectl completion zsh >> ~/.zshrc