NVIDIA Run:ai Platform
Description and Purpose
The NVIDIA Run:ai Platform installs the Run:ai control plane with managed ingress, GPU, monitoring, storage, certificate, and CA-distribution components. Use it to provide centralized Run:ai scheduling and administration services for AI workloads.
| Attribute | Value |
|---|---|
| Application ID | 905abc6e-7cea-4a0a-bc84-ad215a3fc1fe |
| Helm repository | https://helm.ngc.nvidia.com/nvidia/runai |
| Helm chart | control-plane |
| Version | v2.25 |
As internal dependencies, OneKS also installs and fully manages the following components. They are not exposed as standalone applications, but appear inside the Kubernetes cluster among the resources created by this application:
- Longhorn: Persistent storage.
- trust-manager: Certificate trust distribution.
- NVIDIA GPU Operator: GPU management.
- Prometheus: Monitoring and metrics.
- HAProxy Kubernetes Ingress Controller: Ingress and service access.
Parameters
All application-specific parameters are mandatory:
| Parameter | Type | Sensitive | Description |
|---|---|---|---|
domain | String | No | Fully qualified domain name used to access the Run:ai control plane. |
adminUsername | String | No | Username for the initial Run:ai administrator account. |
registryEmail | String | No | Email address stored in the NVIDIA registry pull secret. |
adminPassword | String | Yes | Password for the initial Run:ai administrator account. |
ngcApiKey | String | Yes | NVIDIA NGC API key used to pull Run:ai images and charts. |
Sensitive values are used to create Kubernetes Secrets and are not displayed as plain text in normal application output.
Usage and Configuration
To install this application, follow the Managing Applications guide using the catalog ID shown above and provide the application-specific parameters described in the previous section.
From the OpenNebula Front-end, retrieve the kubeconfig for the target K8s Cluster:
oneks show cluster <CLUSTER_ID> --kubeconfig > kubeconfig
Connect to the Run:ai User Interface
After the application reaches ready, retrieve the ingress address. The following example assumes that the application was installed with runai as both the release name and target namespace:
KUBECONFIG=./kubeconfig kubectl get ingress -n runai runai-ingress \
-o json | jq -r '.status.loadBalancer.ingress[0].ip'
Map that address to the configured domain on the workstation from which you will access the user interface:
echo '<ingress_ip> <domain>' | sudo tee -a /etc/hosts
Open https://<domain> and sign in with adminUsername and adminPassword.
GIVE FEEDBACK
Was this resource helpful?
Glad to hear it
Sorry to hear that