cm-cluster-on-demand 11.33.0

NVIDIA Base Command Manager Cluster on Demand

Cluster on Demand (COD) is a command-line tool for provisioning and managing NVIDIA Base Command Manager (BCM) clusters on major cloud providers. BCM is a cluster management platform that handles node provisioning, workload scheduling, and software stack deployment across HPC and AI environments.

COD supports managing clusters in AWS, Azure, GCP, and OCI.

Installation

Install COD with the extra for your cloud provider:

pip install cm-cluster-on-demand[aws]
pip install cm-cluster-on-demand[azure]
pip install cm-cluster-on-demand[gcp]
pip install cm-cluster-on-demand[oci]

To install support for multiple providers at once:

pip install cm-cluster-on-demand[aws,azure,gcp,oci]

Usage

Each provider has its own CLI entry point. Use --help to explore all available options:

cm-cod-aws --help
cm-cod-aws cluster create --help

Each provider CLI supports the following commands:

Command	Description
cluster create	Create a new BCM cluster
cluster list	List existing clusters
cluster delete	Delete a cluster and its cloud resources
cluster start	Start head node instances of a stopped cluster
cluster stop	Stop all instances of a cluster
image list	List available BCM head node images

The following examples each create a cluster with 5 compute nodes and 1 head node.

AWS

cm-cod-aws cluster create \
    --on-error 'cleanup' \
    --aws-region 'us-east-1' \
    --wlm 'slurm' \
    --nodes '5' \
    --aws-access-key-id <AWS_ACCESS_KEY_ID> \
    --aws-secret-key <AWS_SECRET_KEY> \
    --cluster-password <CLUSTER_PASSWORD> \
    --license-product-key <LICENSE_PRODUCT_KEY> \
    --name mycluster

Azure

cm-cod-azure cluster create \
    --on-error 'cleanup' \
    --wlm 'slurm' \
    --nodes '5' \
    --cluster-password <CLUSTER_PASSWORD> \
    --license-product-key <LICENSE_PRODUCT_KEY> \
    --name mycluster

GCP

cm-cod-gcp cluster create \
    --on-error 'cleanup' \
    --head-node-zone 'europe-west4-c' \
    --wlm 'slurm' \
    --nodes '5' \
    --project-id <PROJECT_ID> \
    --cluster-password <CLUSTER_PASSWORD> \
    --license-product-key <LICENSE_PRODUCT_KEY> \
    --name mycluster

OCI

cm-cod-oci cluster create \
    --on-error 'cleanup' \
    --oci-region 'eu-amsterdam-1' \
    --wlm 'slurm' \
    --nodes '5' \
    --oci-tenancy <OCI_TENANCY> \
    --oci-user <OCI_USER> \
    --oci-fingerprint <OCI_FINGERPRINT> \
    --oci-key-file <OCI_KEY_FILE> \
    --cluster-password <CLUSTER_PASSWORD> \
    --license-product-key <LICENSE_PRODUCT_KEY> \
    --name mycluster

Documentation

For full setup and usage documentation, see the Cloudbursting Manual.