MaaS Installation Overview

Models-as-a-Service is compatible with the Open Data Hub project (ODH) and Red Hat OpenShift AI (RHOAI). MaaS is installed by enabling it in the DataScienceCluster resource:

• Install your platform (ODH or RHOAI operators and DSCInitialization).
• Install MaaS Components (Database, Gateways, DataScienceCluster).

Version Compatibility

MaaS Version	OCP	Kuadrant (ODH) / RHCL (RHOAI)	Gateway API
v0.0.2	4.19.9+	v1.3+ / v1.2+	v1.2+
v0.1.0+	4.19.9+	v1.4.2+ / v1.3	v1.2+

RHCL v1.4.0 — silent auth bypass

RHCL v1.4.0 contains a Wasm shim bug that silently bypasses gateway authentication. Management endpoints return AUTH_FAILURE; inference endpoints work but are unauthenticated. Upgrade to RHCL v1.4.1+. See Troubleshooting #16 for diagnosis.

Other Kubernetes flavors

Other Kubernetes flavors (e.g., upstream Kubernetes, other distributions) are currently being validated.

For the mapping between RHOAI product versions and MaaS releases, see RHOAI to MaaS Release Mapping.

Required Tools

The following tools are used across the installation guides:

• kubectl or oc — cluster access
• curl — used by Operator Setup (ODH/LWS)
• jq — used for validation and version parsing
• kustomize — used for Gateway AuthPolicy (MaaS Components)
• envsubst — used for policy templates (MaaS Components)

Requirements for Open Data Hub project

MaaS requires Open Data Hub version 3.0 or later, with the Model Serving component enabled (KServe) and properly configured for deploying models with LLMInferenceService resources.

Requirements for Red Hat OpenShift AI

MaaS requires Red Hat OpenShift AI (RHOAI) version 3.0 or later, with the Model Serving component enabled (KServe) and properly configured for deploying models with LLMInferenceService resources.

A specific requirement for MaaS v0.1.0+ is to set up RHOAI Model Serving with Red Hat Connectivity Link (RHCL) v1.3 or later.

Optional: Observability Prerequisites

If you plan to use MaaS dashboards, showback, or usage metrics, the ODH monitoring stack needs to be enabled in the Platform Operator.

To enable the ODH monitoring stack, you need to configure DSCI monitoring.metrics. For example:

kubectl apply -f - <<EOF
apiVersion: dscinitialization.opendatahub.io/v2
kind: DSCInitialization
metadata:
    name: default-dsci
spec:
    applicationsNamespace: opendatahub
    monitoring:
        managementState: Managed
        namespace: opendatahub
        metrics:
            storage:
                size: 90Gi
    trustedCABundle:
        managementState: Managed
EOF

Note that enabling the ODH monitoring stack also requires to install the Cluster Observability Operator and OpenTelemetry Operator.

See Managing observability (RHOAI 3.4).

Loki for Access logs

To enable storing access logs from the MaaS gateway for logs-based showback or auditing, you need to configure a LokiStack. First, install the Loki Operator, and then configure a LokiStack named usage in the monitoring namespace configured in DSCI spec.monitoring.namespace, for example:

apiVersion: loki.grafana.com/v1
kind: LokiStack
metadata:
  name: usage
  namespace: opendatahub
spec:
  limits:
    global:
      otlp:
        streamLabels:
          resourceAttributes:
            - name: service.name
            - name: subscription
            - name: model
            - name: response_type
            - name: kubernetes_namespace_name
  managementState: Managed
  size: 1x.demo
  storage:
    schemas:
      - effectiveDate: '2024-10-01'
        version: v13
    secret:
      credentialMode: static
      name: <storage-secret-name>
      type: s3
  tenants:
    mode: openshift-logging

GenAI Studio

To enable GenAI Studio in the RHOAI Dashboard, you need the LlamaStack Operator enabled in your DSC and a Dashboard feature flag. See OdhDashboardConfig Feature Flags for setup.