Bill of Materials

This document lists all Red Hat and community products/operators consumed by the Hybrid Mesh Platform pattern, along with their minimum versions and sources.

OpenShift Platform

Component Minimum Version Notes
OpenShift Container Platform 4.16+ Tested on 4.17-4.20
OpenShift GitOps 1.20+ Installed via OperatorHub

Red Hat Operators (redhat-operators)

Operator Channel Minimum Version Purpose
Advanced Cluster Management (ACM) release-2.16 2.16.x Multi-cluster management, GitOps fleet
Red Hat Advanced Cluster Security (ACS) stable 4.7+ Security scanning, policy enforcement
Red Hat OpenShift AI (RHOAI) stable-3.4 3.4+ ML/AI workloads, model serving (DSC API v2)
Red Hat Connectivity Link (RHCL) stable 1.4+ API management, Kuadrant gateway
Red Hat Service Interconnect (Skupper) stable-2 2.1+ Multi-cluster service mesh
Red Hat OpenShift Serverless stable 1.35+ Knative serving/eventing
Red Hat AMQ Streams (Kafka) stable 3.2+ Event streaming
AMQ Streams Console stable 3.2+ Kafka UI console
Red Hat OpenTelemetry stable 0.144+ Distributed tracing collection
Tempo Operator stable 0.20+ Trace storage backend
Red Hat Developer Hub (RHDH) fast-1.9 1.9+ Developer portal, Backstage
Red Hat Quay stable-3.14 3.14+ Container registry
OpenShift Virtualization (CNV) stable 4.20+ VM workloads
Cluster Observability Operator stable 1.0+ Unified monitoring
Kiali stable 2.22+ Service mesh observability
OpenShift External Secrets stable-v1 1.1+ Secrets management
Red Hat OpenShift Dev Spaces stable 3.19+ Cloud IDE

Community Operators (community-operators)

Operator Channel Minimum Version Purpose
Grafana Operator v5 5.24+ Dashboard provisioning
Camel K stable-v2 2.10+ Integration pipelines
GitLab Operator stable 8.9+ Hub SCM (webservice, gitaly, registry)
GitLab Runner Operator stable 1.17+ CI runners for Tekton/scaffolder pipelines

Helm Charts (External)

Chart Repository Version Purpose
clustergroup charts.validatedpatterns.io 0.9.* VP framework bootstrap
hashicorp-vault charts.validatedpatterns.io 0.1.* Secrets management
openshift-external-secrets charts.validatedpatterns.io 0.0.* ESO configuration
gitlab-operator OperatorHub (community) stable GitLab + Runner for scaffolder/Tekton
network-observer quay.io/skupper/helm 2.1.3 Skupper network visualization

Industrial Edge Components

Component Source Version Purpose
AMQ Broker redhat-operators 7.12.x Message broker for edge
Kubernetes Image Puller community-operators stable Pre-pull images on nodes

Validated Patterns Framework

Component Version Source
common (Makefile-common) main github.com/validatedpatterns/common
clustergroup-chart 0.9.* charts.validatedpatterns.io

Version Compatibility Matrix

OCP Version ACM ACS RHOAI Status
4.20 2.16 4.7 3.4 Tested
4.17-4.19 2.15+ 4.6+ 3.4+ Compatible
4.16 2.14+ 4.5+ Minimum (OpenShift AI 3.4 requires 4.17+)

Notes

  • All operators use Automatic install plan approval unless noted
  • Channel versions follow semantic versioning where stable-X.Y pins to major.minor
  • Community operators may have different support lifecycle than Red Hat operators
  • For production deployments, pin specific versions in subscription manifests

Cluster sizing

Sizing for v2.2 AI CV at the Edge (NeuroFace + OVMS ModelMesh + KServe YOLO on spokes). Industrial Edge optional and disabled by default.

Hub (CPU-only)

Tier Workers vCPU / worker Memory / worker Total alloc Use case
Recommended (workshop 30–50) 4 16 64 GiB 64 vCPU / 256 GiB Full platform stack
Minimum demo (1–5 users) 3 8 32 GiB 24 vCPU / 96 GiB Disable Kubecost/CNV under pressure

Verify: bash scripts/verify-node-capacity.sh (hub 30: ≥16 CPU / 64 GiB; hub 50: ≥20 CPU / 80 GiB).

Spokes — CPU-only (default)

Tier Workers vCPU / worker Memory / worker Total alloc Use case
Recommended (AI CV + DevSpaces) 3 8 32 GiB 24 vCPU / 96 GiB NeuroFace + OVMS + YOLO + DevSpaces
Minimum demo (AI CV only) 2 4 16 GiB 8 vCPU / 32 GiB NeuroFace + OVMS only

Verify: ROLE=spoke bash scripts/verify-node-capacity.sh (recommended: ≥24 CPU / 96 GiB; minimum: SPOKE_TIER=minimum ≥8 CPU / 32 GiB).

Spoke CPU workload budget (approximate):

Component CPU Memory
OVMS ModelMesh (face detection) 1 2 GiB
KServe YOLO PPE (HPA 1–4) 0.2–2 1–3 GiB
NeuroFace app 0.5 1 GiB
Kafka (cv-kafka) 1 2 GiB
Skupper connectors 0.2 0.5 GiB
ACS Secured Cluster 0.5 1.5 GiB
Istio ambient + ztunnel 0.3 0.5 GiB
DevSpaces (per workspace) 1 2 GiB
Optional IE stack +2 +4 GiB

Spokes — GPU-accelerated (optional)

Tier Workers vCPU / worker Memory / worker GPU / worker Use case
Recommended 3 8 32 GiB 1× T4 / A10G OVMS GPU, YOLO GPU, vLLM 7B
Production-like 3 16 64 GiB 1× A100 (24+ GB VRAM) Multi-model, LLM 14–70B

GPU VRAM budget: OVMS ~2 GiB, YOLO ~2 GiB, vLLM 7B ~16 GiB, vLLM 14–70B ~24–80 GiB.

GPU operators (optional)

Not installed by pattern default (CPU inference path). Required when GPU nodes are present.

Required (install order)

Operator Channel Catalog Purpose
Node Feature Discovery (NFD) stable redhat-operators Labels GPU nodes (feature.node.kubernetes.io/pci-10de.present)
NVIDIA GPU Operator stable certified-operators Drivers, device plugin, DCGM, container toolkit

Optional

Operator Channel Catalog When needed
NVIDIA Network Operator stable certified-operators GPUDirect RDMA, multi-node training
OpenShift Serverless stable redhat-operators Knative KServe (scale-to-zero on GPU)
NVIDIA NIM Operator stable certified-operators NVIDIA NIM optimized LLM containers
Cloud Instance GPU VRAM Use case
AWS g4dn.xlarge 1× T4 16 GB OVMS + YOLO demo
AWS g5.2xlarge 1× A10G 24 GB OVMS + YOLO + vLLM 7B
AWS p4d.24xlarge 8× A100 320 GB Large LLM 70B+
Azure Standard_NC4as_T4_v3 1× T4 16 GB Demo
Azure Standard_NC24ads_A100_v4 1× A100 80 GB Large LLM
GCP a2-highgpu-1g 1× A100 40 GB Production inference

Verify GPU: CHECK_GPU=1 ROLE=spoke bash scripts/verify-node-capacity.sh

Updating Versions

To update operator versions:

  1. Modify the channel field in charts/region/{hub,east,west}/values.yaml
  2. For external Helm charts, update chartVersion in the application definition
  3. Test in a non-production cluster before rolling out

References