deploy-existing-openshift

Deploying a private AI model serving stack on OpenShift uses Helm to install llm-d, vLLM, and gateway components. The workflow verifies cluster access, manages namespaces, and applies cluster-specific overrides for ARO or shared Kuadrant environments.

Deploying llm-d and vLLM on OpenShift requires active cluster access, namespace management permissions, and Helm for release orchestration. The deployment process validates gateway prerequisites and namespace ownership before proceeding with component installation.

You can enable optional guardrails and observability workflows when deploying vLLM on OpenShift. The Helm-based deployment supports configuring these production features alongside the core model serving stack components during installation or redeployment.

Redeploying an AI model serving stack on OpenShift without external provider exposure involves using Helm to manage llm-d and vLLM releases. This approach handles existing-cluster adoption and applies ARO-specific overrides to maintain a fully private environment.

Safely tearing down a private LLM deployment from OpenShift involves following specific teardown guidance to cleanly remove Helm releases. This ensures llm-d, vLLM, and gateway components are properly uninstalled from their managed namespaces without leaving residual resources.

Deploying llm-d on OpenShift supports ARO-specific configurations through cluster-specific overrides. The Helm-based deployment workflow applies these overrides to accommodate ARO environments and shared Kuadrant setups during the installation process.