Getting started

Mon, 01 Jan 0001 00:00:00 +0000

This guide walks you through deploying the AI Computer Vision pattern from a clean environment. Install the east and west spoke clusters first, collect their API tokens, then install the hub last so RHACM can auto-import both spokes. Each section includes verification commands with expected output so you know whether each step succeeded.

Complete each cluster installation and verification before moving to the next. The hub imports spokes automatically when you provide spoke API tokens in ~/values-secret-ia-computer-vision.yaml. Cross-cluster features such as Skupper, multi-cluster observability, RHACS Secured Cluster registration, and the RHCL inference gateway require imported spokes.

Cluster sizing and version matrix

Mon, 01 Jan 0001 00:00:00 +0000

This page documents cluster sizing requirements, network prerequisites, and the tested product version matrix for the AI Computer Vision pattern.

Recommended AWS sizing

Cluster role Control plane Workers Notes

Hub

3 x m6a.2xlarge

3 x m6a.4xlarge

Hub runs 15+ operators and ~450 pods. m6a.2xlarge workers (8 vCPU) reach 98-99% CPU request allocation. Use m6a.4xlarge (16 vCPU, 64 GiB) for production. For sandbox/demo, m6a.2xlarge works if you remove the NoSchedule taint from master nodes.

East spoke

3 x m6a.2xlarge

Spoke workloads are lighter (~150 pods). m6a.2xlarge is sufficient for CPU-only inference.

West spoke

3 x m6a.2xlarge

Same as east.

Each m6a.2xlarge instance provides 8 vCPU and 32 GiB memory. Each m6a.4xlarge provides 16 vCPU and 64 GiB memory.

Architecture

Mon, 01 Jan 0001 00:00:00 +0000

This document describes the hub-spoke topology, component placement, traffic flow, security model, connectivity, and GitOps workflow for the AI Computer Vision pattern. It also explains why specific Red Hat products were chosen over alternatives.

Overview

The AI Computer Vision pattern uses a hub-spoke model with one central control plane cluster and two symmetric edge inference clusters. This architecture separates fleet management and developer platform services from latency-sensitive inference workloads.

You deploy the pattern on three independent OCP 4.20 or later clusters:

Scaffolding and secrets

Mon, 01 Jan 0001 00:00:00 +0000

This document explains how the AI Computer Vision software template orchestrates operators on the hub and spokes, and how HashiCorp Vault with the External Secrets Operator (ESO) delivers credentials to OpenShift without storing secret values in Git.

AI Computer Vision software template

What it is

The software template ai-computer-vision in Red Hat Developer Hub (RHDH) is a Backstage Template resource. It gives workshop users a self-service form to provision a personal NeuroFace stack on a chosen spoke (east or west) without opening tickets to platform teams.

Workshop mode

Mon, 01 Jan 0001 00:00:00 +0000

The AI Computer Vision pattern includes workshop mode for hands-on events. Workshop mode is enabled by default with 30 users (user1 through user30).

What gets provisioned

Platform Access method Default password

OCP Console

HTPasswd identity provider workshop-users

Welcome123!

Developer Hub

Keycloak realm backstage (OIDC)

Welcome123!

GitLab

User account + ws-userN group (owner)

Welcome123!

DevSpaces

Git credentials in userN-devspaces namespace

Welcome123!

Special accounts: platformadmin (platform engineering group) and admin (workshop administrator).

Ideas for customization

Mon, 01 Jan 0001 00:00:00 +0000

This document describes actionable ways to extend the AI Computer Vision pattern for production or proof-of-concept environments. Each scenario includes context, the exact files and values to change, and a validation command.

Per-user OIDC Gateway routing (ApplicationSet)

Context

Workshop users scaffold personal NeuroFace instances via Developer Hub. Each instance gets a unique HTTPRoute path (/user/userN/) on the shared spoke Gateway and an RHCL AuthPolicy with OIDC pointing to the user’s RHBK biometric realm.

CRD reference

Wed, 01 Jul 2026 00:00:00 +0000

Annotated YAML examples for custom resources used in the AI Computer Vision pattern.

Troubleshooting

Mon, 01 Jan 0001 00:00:00 +0000

This reference lists common problems during pattern installation and operation. Each entry includes symptoms, probable cause, diagnostic commands, and resolution steps.

Operator CSV not yet Succeeded — the application synced before the operator was ready
Duplicate OperatorGroup in the target namespace (TooManyOperatorGroups)
Stale subscription referencing a deleted or renamed package
Sync wave ordering — downstream app synced before its prerequisite
A custom resource requires a CRD that is not yet installed (for example, ObjectBucketClaim for Quay requires OpenShift Data Foundation)

AI Computer Vision

Getting started

Cluster sizing and version matrix

Recommended AWS sizing

Architecture

Overview

Scaffolding and secrets

AI Computer Vision software template

What it is

Workshop mode

What gets provisioned

Ideas for customization

Per-user OIDC Gateway routing (ApplicationSet)

Context

CRD reference

Troubleshooting

OpenShift GitOps application Degraded or Unknown

Symptom

Probable cause