Install a Cluster

Customize Installation¶

Perform the cluster installation instructions explained here.
(Optional) make the following changes to the configuration file you have downloaded:

Key	Default	Description
`pspEnabled`	`false`	Set to `true` when using PodSecurityPolicy
`ingress-nginx.podSecurityPolicy.enabled`	Set to `true` when using PodSecurityPolicy
`runai-operator.config.project-controller.createNamespaces`	`true`	Set to `false` if unwilling to provide Run:AI the ability to create namespaces, or would want to create namespaces manually rather than use the Run:AI convention of `runai-<PROJECT-NAME>`. When set to `false`, will require an additional manual step when creating new Run:AI Projects.
`runai-operator.config.project-controller.createRoleBindings`	`true`	Automatically assign Users to Projects. Set to `false` if unwilling to provide Run:AI the ability to set RoleBinding. When set to `false`, will require an additional manual step when adding or removing users from Projects.
`runai-operator.config.project-controller.clusterWideSecret`	`true`	Set to `false` if unwilling to provide Run:AI the ability to create Kubernetes Secrets. When not enabled, automatic secret propagation will not be available
`runai-operator.config.mps-server.enabled`	`false`	Allow the use of NVIDIA MPS. MPS is useful with Inference workloads. Requires extra cluster permissions
`runai-operator.config.runai-container-toolkit.enabled`	`true`	Controls the usage of Fractions. Requires extra cluster permissions
`runai-operator.config.runaiBackend.password`	Default password already set	admin@run.ai password. Need to change only if you have changed the password here
`gpu-feature-discovery.enabled`	`true`	Install Node Feature Discovery. Set to `false` if already installed in cluster
`kube-prometheus-stack.enabled`	`true`	Install Prometheus. Set to `false` if Prometheus is already installed in cluster
`mpi-operator.enabled`	`true`	Set to `false` when using PodSecurityPolicy. MPI is the distributed-training operator from KubeFlow. Currently must run with root access

Feature Discovery¶

The Run:AI Cluster installation installs by default two pre-requisites: Kubernetes Node Feature Discovery (NFD) and NVIDIA GPU Feature Discovery (GFD).

If your Kubernetes cluster already has GFD installed, you will want to set gpu-feature-discovery.enabled to false.
NFD is a prerequisite of GFD. If GFD is not installed, but NFD is already installed, you can disable NFD installation by setting gpu-feature-discovery.nfd.deploy to false.

Prometheus¶

The Run:AI Cluster installation installs Prometheus by default. If your Kubernetes cluster already has Prometheus installed, set the flag kube-prometheus-stack.enabled to false.

When using an existing Prometheus installation, you will need to add additional rules to your Prometheus configuration. The rules can be found under deploy/runai-prometheus-rules.yaml.

NVIDIA Device Plugin¶

Run:AI has customized the NVIDIA device plugin for Kubernetes. If you have installed the NVIDIA GPU Operator or have previously installed this plug-in, run the following to disable the existing plug-in:

kubectl -n gpu-operator-resources patch daemonset nvidia-device-plugin-daemonset \
   -p '{"spec": {"template": {"spec": {"nodeSelector": {"non-existing": "true"}}}}}'

Install Cluster¶

Run:

Connected

helm repo add runai https://run-ai-charts.storage.googleapis.com
helm repo update

helm install runai-cluster runai/runai-cluster -n runai \
    -f runai-<cluster-name>.yaml --create-namespace

Info

To install a specific version, add --version <version> to the install command.

Airgapped

helm install runai-cluster -n runai  \ 
  runai-cluster-<version>.tgz -f runai-<cluster-name>.yaml --create-namespace

Tip

Use the --dry-run flag to gain an understanding of what is being installed before the actual installation. For more details see: Understanding cluster access foles.

Next Steps¶

Continue to configure web interfaces.

Last update: January 16, 2022