Feat/add addons templates

AGENTS.md

@@ -273,10 +273,16 @@ make build-installer IMG=<registry>/<project>:tag
 
 **Key points:**
 
-- The `dist/install.yaml` is generated from Kustomize manifests (CRDs, RBAC, Deployment)
+- The `dist/install.yaml` is generated from Kustomize manifests (CRDs, RBAC, Deployment, HelmRepository, GitRepository, addon ModuleTemplates)
 - Commit this file to your repository for easy distribution
 - Users only need `kubectl` to install (no additional tools required)
 
+**GitRepository version substitution:**
+
+- `config/modules/gitrepository.yaml` uses the placeholder `__MODULE_OPERATOR_VERSION__` for the Git ref tag
+- The `build-installer` target replaces it with the current version (`git describe --tags --always`) via `sed` before writing `dist/install.yaml`
+- Modules (CDI, KubeVirt, etc.) are thus sourced from the same version as the operator being built, enabling feature-branch testing without hardcoded tags
+
 **Example:** Users install with a single command:
 
 ```bash

Makefile

@@ -158,10 +158,22 @@ docker-buildx: ## Build and push docker image for the manager for cross-platform
 	rm Dockerfile.cross
 
 .PHONY: build-installer
-build-installer: manifests generate kustomize ## Generate a consolidated YAML with CRDs and deployment.
+build-installer: manifests generate kustomize ## Generate dist/install.yaml (CRDs+deploy+HelmRepository+ModuleTemplate+GitRepository+addon ModuleTemplates).
 	mkdir -p dist
 	cd config/manager && "$(KUSTOMIZE)" edit set image controller=${IMG}
-	"$(KUSTOMIZE)" build config/default > dist/install.yaml
+	{ "$(KUSTOMIZE)" build config/default; echo "---"; "$(KUSTOMIZE)" build config/modules; } | \
+		sed 's|__MODULE_OPERATOR_VERSION__|$(VERSION)|g' | \
+		sed "s|__CDI_VERSION__|$$(cat config/modules/cdi/VERSION | tr -d '\n')|g" | \
+		sed "s|__KUBEVIRT_VERSION__|$$(cat config/modules/kubevirt/VERSION | tr -d '\n')|g" | \
+		sed "s|__ROOK_CEPH_VERSION__|$$(cat config/modules/rook-ceph/VERSION | tr -d '\n')|g" | \
+		sed "s|__GATEWAY_API_VERSION__|$$(cat config/modules/gateway-api/VERSION | tr -d '\n')|g" | \
+		sed "s|__GPU_OPERATOR_VERSION__|$$(cat config/modules/gpu-operator/VERSION | tr -d '\n')|g" | \
+		sed "s|__HAMI_VERSION__|$$(cat config/modules/hami/VERSION | tr -d '\n')|g" | \
+		sed "s|__ISTIO_VERSION__|$$(cat config/modules/istio/VERSION | tr -d '\n')|g" | \
+		sed "s|__PROMETHEUS_VERSION__|$$(cat config/modules/prometheus/VERSION | tr -d '\n')|g" | \
+		sed "s|__LLM_D_INFRA_VERSION__|$$(cat config/modules/llm-d-infra/VERSION | tr -d '\n')|g" | \
+		sed "s|__GAIE_VERSION__|$$(cat config/modules/gateway-api-inference-extension/VERSION | tr -d '\n')|g" \
+		> dist/install.yaml
 
 ##@ Deployment
 

config/manager/kustomization.yaml

@@ -1,2 +1,8 @@
 resources:
 - manager.yaml
+apiVersion: kustomize.config.k8s.io/v1beta1
+kind: Kustomization
+images:
+- name: controller
+  newName: controller
+  newTag: latest

config/modules/cdi/VERSION

@@ -0,0 +1 @@
+1.65.0

config/modules/cdi/kustomization.yaml

@@ -0,0 +1,8 @@
+# CDI deployment — remote resources for FluxCD Kustomization
+# Source: https://github.com/kubevirt/containerized-data-importer
+apiVersion: kustomize.config.k8s.io/v1beta1
+kind: Kustomization
+
+resources:
+  - https://github.com/kubevirt/containerized-data-importer/releases/download/v__CDI_VERSION__/cdi-operator.yaml
+  - https://github.com/kubevirt/containerized-data-importer/releases/download/v__CDI_VERSION__/cdi-cr.yaml

config/modules/cdi/moduletemplate.yaml

@@ -0,0 +1,24 @@
+# ModuleTemplate — CDI (Containerized Data Importer) v__CDI_VERSION__
+# Uses FluxCD Kustomization (not Helm). Deploys from remote release YAMLs.
+#
+# Prerequisite: GitRepository (otterscale-module-operator) must exist in otterscale-system.
+# Apply: kubectl apply -k config/modules/cdi/
+apiVersion: module.otterscale.io/v1alpha1
+kind: ModuleTemplate
+metadata:
+  name: cdi
+spec:
+  namespace: cdi
+
+  description: >-
+    Containerized Data Importer v__CDI_VERSION__ — PVC data import for KubeVirt VM disk images.
+    Deploys via FluxCD Kustomization from upstream release YAMLs.
+
+  kustomization:
+    interval: 30m
+    prune: true
+    sourceRef:
+      kind: GitRepository
+      name: otterscale-module-operator
+      namespace: otterscale-system
+    path: config/modules/cdi

config/modules/gateway-api-inference-extension/VERSION

@@ -0,0 +1 @@
+1.3.1

config/modules/gateway-api-inference-extension/kustomization.yaml

@@ -0,0 +1,6 @@
+# Gateway API Inference Extension v__GAIE_VERSION__
+apiVersion: kustomize.config.k8s.io/v1beta1
+kind: Kustomization
+
+resources:
+  - https://github.com/kubernetes-sigs/gateway-api-inference-extension/releases/download/v__GAIE_VERSION__/manifests.yaml

config/modules/gateway-api-inference-extension/moduletemplate.yaml

@@ -0,0 +1,24 @@
+# ModuleTemplate — Gateway API Inference Extension v__GAIE_VERSION__
+# Uses FluxCD Kustomization. Requires Gateway API CRDs.
+apiVersion: module.otterscale.io/v1alpha1
+kind: ModuleTemplate
+metadata:
+  name: gateway-api-inference-extension
+spec:
+  namespace: default
+
+  description: >-
+    Gateway API Inference Extension v__GAIE_VERSION__ — CRDs and controllers
+    for AI/ML inference routing via Gateway API.
+
+  kustomization:
+    interval: 30m
+    prune: true
+    dependsOn:
+      - name: gateway-api
+        namespace: default
+    sourceRef:
+      kind: GitRepository
+      name: otterscale-module-operator
+      namespace: otterscale-system
+    path: config/modules/gateway-api-inference-extension

config/modules/gateway-api/VERSION

@@ -0,0 +1 @@
+1.5.1

config/modules/gateway-api/kustomization.yaml

@@ -0,0 +1,6 @@
+# Gateway API v__GATEWAY_API_VERSION__ — Standard channel CRDs
+apiVersion: kustomize.config.k8s.io/v1beta1
+kind: Kustomization
+
+resources:
+  - https://github.com/kubernetes-sigs/gateway-api/releases/download/v__GATEWAY_API_VERSION__/standard-install.yaml

config/modules/gateway-api/moduletemplate.yaml

@@ -0,0 +1,20 @@
+# ModuleTemplate — Kubernetes Gateway API v__GATEWAY_API_VERSION__ (Standard channel)
+# Uses FluxCD Kustomization. CRDs only (GatewayClass, Gateway, HTTPRoute, etc.)
+apiVersion: module.otterscale.io/v1alpha1
+kind: ModuleTemplate
+metadata:
+  name: gateway-api
+spec:
+  namespace: default
+
+  description: >-
+    Kubernetes Gateway API v__GATEWAY_API_VERSION__ — Standard channel CRDs (GatewayClass, Gateway, HTTPRoute, GRPCRoute, ReferenceGrant).
+
+  kustomization:
+    interval: 30m
+    prune: true
+    sourceRef:
+      kind: GitRepository
+      name: otterscale-module-operator
+      namespace: otterscale-system
+    path: config/modules/gateway-api

config/modules/gitrepository.yaml

@@ -0,0 +1,12 @@
+# Shared GitRepository for addon ModuleTemplates (CDI, KubeVirt, etc.)
+# All addon components deploy from this repo's config/modules/*/deploy paths.
+apiVersion: source.toolkit.fluxcd.io/v1
+kind: GitRepository
+metadata:
+  name: otterscale-module-operator
+  namespace: otterscale-system
+spec:
+  interval: 24h
+  url: https://github.com/otterscale/module-operator
+  ref:
+    tag: __MODULE_OPERATOR_VERSION__

config/modules/gpu-operator/VERSION

@@ -0,0 +1 @@
+25.10.1

config/modules/gpu-operator/helmrepository.yaml

@@ -0,0 +1,8 @@
+apiVersion: source.toolkit.fluxcd.io/v1
+kind: HelmRepository
+metadata:
+  name: nvidia-gpu-operator
+  namespace: otterscale-system
+spec:
+  interval: 24h
+  url: https://nvidia.github.io/gpu-operator

config/modules/gpu-operator/moduletemplate.yaml

@@ -0,0 +1,77 @@
+# ModuleTemplate — NVIDIA GPU Operator blueprint
+#
+# Deploys NVIDIA GPU Operator for driver, DCGM, DCGM Exporter, Node Status Exporter.
+# Device plugin and vGPU are disabled (use HAMI for GPU sharing).
+#
+# Apply:
+#   kubectl apply -f moduletemplate.yaml
+apiVersion: module.otterscale.io/v1alpha1
+kind: ModuleTemplate
+metadata:
+  name: gpu-operator
+spec:
+  namespace: gpu-operator
+
+  description: >-
+    Deploys NVIDIA GPU Operator (driver 580.x, DCGM, DCGM Exporter, Node Status Exporter).
+    Device plugin disabled for use with HAMI vGPU.
+
+  helmRelease:
+    interval: 30m
+
+    chart:
+      spec:
+        chart: gpu-operator
+        version: "__GPU_OPERATOR_VERSION__"
+        sourceRef:
+          kind: HelmRepository
+          name: nvidia-gpu-operator
+          namespace: otterscale-system
+        interval: 24h
+
+    install:
+      remediation:
+        retries: 3
+
+    upgrade:
+      remediation:
+        retries: 3
+      force: false
+
+    timeout: 30m
+
+    values:
+      cdi:
+        enabled: false
+
+      driver:
+        enabled: true
+        version: "580.105.08"
+        upgradePolicy:
+          autoUpgrade: false
+
+      # Offline version
+      # driver:
+      #   version: "580"
+      #   usePrecompiled: true
+
+      devicePlugin:
+        enabled: false
+
+      operator:
+        cleanupCRD: true
+        upgradeCRD: false
+
+      dcgm:
+        enabled: true
+
+      dcgmExporter:
+        enabled: true
+        serviceMonitor:
+          enabled: true
+
+      nodeStatusExporter:
+        enabled: true
+
+      vgpuDeviceManager:
+        enabled: false

config/modules/hami/VERSION

@@ -0,0 +1 @@
+2.8.0

config/modules/hami/helmrepository.yaml

@@ -0,0 +1,8 @@
+apiVersion: source.toolkit.fluxcd.io/v1
+kind: HelmRepository
+metadata:
+  name: hami
+  namespace: otterscale-system
+spec:
+  interval: 24h
+  url: https://project-hami.github.io/HAMi/

config/modules/hami/moduletemplate.yaml

@@ -0,0 +1,75 @@
+# ModuleTemplate — HAMI (Heterogeneous AI Computing Virtualization) blueprint
+#
+# Deploys HAMI for GPU sharing (vGPU) with NVIDIA GPU Operator.
+# Requires GPU Operator to be installed first (driver + nvidia-validations).
+#
+# Apply:
+#   kubectl apply -f moduletemplate.yaml
+apiVersion: module.otterscale.io/v1alpha1
+kind: ModuleTemplate
+metadata:
+  name: hami
+spec:
+  # HAMI scheduler and device plugin typically run in kube-system
+  namespace: kube-system
+
+  description: >-
+    Deploys HAMI for GPU sharing (vGPU) with device split, memory/core scaling.
+    Integrates with NVIDIA GPU Operator via gpuOperatorToolkitReady.
+
+  helmRelease:
+    interval: 30m
+
+    dependsOn:
+      - name: gpu-operator
+        namespace: gpu-operator
+
+    chart:
+      spec:
+        chart: hami
+        version: "__HAMI_VERSION__"
+        sourceRef:
+          kind: HelmRepository
+          name: hami
+          namespace: otterscale-system
+        interval: 24h
+
+    install:
+      remediation:
+        retries: 3
+
+    upgrade:
+      remediation:
+        retries: 3
+      force: false
+
+    timeout: 30m
+
+    values:
+      global:
+        imageTag: "v__HAMI_VERSION__"
+        annotations:
+          prometheus.io/scrape: "true"
+          prometheus.io/port: "31992"
+        # managedNodeSelectorEnable: false
+        # managedNodeSelector:
+        #   usage: "nvidia.com/gpu.workload.config"
+
+      scheduler:
+        service:
+          type: ClusterIP
+
+      devicePlugin:
+        enabled: true
+        gpuOperatorToolkitReady:
+          enabled: true
+          hostPath: "/run/nvidia/validations"
+        deviceSplitCount: 10
+        deviceMemoryScaling: 1
+        deviceCoreScaling: 1
+        service:
+          type: ClusterIP
+        runtimeClassName: "nvidia"
+        nvidiaNodeSelector:
+          gpu: null
+          nvidia.com/gpu.workload.config: "vgpu"

config/modules/istio/VERSION

@@ -0,0 +1 @@
+1.29.1

config/modules/istio/helmrepository.yaml

@@ -0,0 +1,8 @@
+apiVersion: source.toolkit.fluxcd.io/v1
+kind: HelmRepository
+metadata:
+  name: istio
+  namespace: otterscale-system
+spec:
+  interval: 24h
+  url: https://istio-release.storage.googleapis.com/charts

config/modules/istio/moduletemplate-base.yaml

@@ -0,0 +1,42 @@
+# ModuleTemplate — Istio base (CRDs)
+#
+# Installs Istio base chart (CRDs). Must be applied before istiod.
+# See: https://istio.io/latest/docs/setup/install/helm/
+#
+# Apply order: base → istiod
+apiVersion: module.otterscale.io/v1alpha1
+kind: ModuleTemplate
+metadata:
+  name: istio-base
+spec:
+  namespace: istio-system
+
+  description: >-
+    Installs Istio base chart (CRDs). Required before istiod.
+
+  helmRelease:
+    interval: 30m
+
+    chart:
+      spec:
+        chart: base
+        version: "__ISTIO_VERSION__"
+        sourceRef:
+          kind: HelmRepository
+          name: istio
+          namespace: otterscale-system
+        interval: 24h
+
+    install:
+      remediation:
+        retries: 3
+
+    upgrade:
+      remediation:
+        retries: 3
+      force: false
+
+    timeout: 10m
+
+    values:
+      defaultRevision: default