Skip to navigation

Command Reference and Troubleshooting

View as Markdown

The runai-installer commands and flags are identical across same-cluster, multi-cluster, connected, and air-gapped installations - run each command against the relevant cluster with its own --config and --kubeconfig.

Global Flags

These flags apply to all runai-installer subcommands.

FlagDescriptionDefault
--logs-dirDirectory where installer log files are writtenCurrent directory
--step-timeoutMaximum time allowed for each installation step7m

install Command

The install command installs NVIDIA Run:ai from a configuration file.

./runai-installer install [flags]
FlagDescriptionDefault
--configPath to the installation configuration file (required)-
--dry-runPlan the installation without applying any changesfalse
--preflight-onlyRun preflight checks and exit without installingfalse
--skip-preflightSkip preflight checks and proceed with installationfalse
-y, --yesRun non-interactively without confirmation promptsfalse
--outputOutput format: text or jsontext on a TTY, json otherwise
--events-fileWrite NDJSON installation events to a file-
--status-fileWrite the final installation summary to a file-
--on-failureFailure policy: abort or skipabort
--kubeconfigPath to a kubeconfig file-
--ngc-token-pathPath to an NGC API key file (overrides registry in the configuration file)-
--execution-timeoutMaximum time allowed for the installation50m

Troubleshooting

Cluster Cannot Reach the Control Plane

If cluster registration fails or a cluster does not show as Connected, verify that the NVIDIA Run:ai cluster’s nodes can reach the control plane FQDN over HTTPS (port 443):

curl -v https://<control-plane-domain>

Installation Step Timed Out

Increase the per-step timeout for slow networks or large image pulls:

./runai-installer install \
--config config.yaml \
--step-timeout 15m

View Installer Logs

# Stream logs in real time
tail -f ./logs/runai-installer.log
# Show only errors
grep -i error ./logs/runai-installer.log
# Filter Helm operations
grep "\[Helm\]" ./logs/runai-installer.log
# Filter Kubernetes API calls
grep "\[K8s\]" ./logs/runai-installer.log

Registry Authentication Failed

Verify that the NGC API key file is valid and readable:

cat /path/to/ngc-token

The file must contain only the API key. The runai-installer CLI automatically trims leading and trailing whitespace.