Merge pull request #533 from bjwswang/charts

feat: add kuberay operator and configure ray clusters in arcadia

config/samples/ray.io_v1_raycluster.yaml

@@ -0,0 +1,63 @@
+apiVersion: ray.io/v1
+kind: RayCluster
+metadata:
+  name: raycluster-kuberay
+  namespace: kuberay-system
+spec:
+  headGroupSpec:
+    rayStartParams:
+      dashboard-host: 0.0.0.0
+    template:
+      metadata:
+        labels:
+          app.kubernetes.io/instance: raycluster
+          app.kubernetes.io/name: kuberay
+      spec:
+        containers:
+        - image: kubeagi/ray-ml:2.9.0-py39-vllm
+          name: ray-head
+          resources:
+            limits:
+              cpu: "1"
+              memory: 2G
+              nvidia.com/gpu: 1
+            requests:
+              cpu: "1"
+              memory: 2G
+              nvidia.com/gpu: 1
+          volumeMounts:
+          - mountPath: /tmp/ray
+            name: log-volume
+        volumes:
+        - emptyDir: {}
+          name: log-volume
+  workerGroupSpecs:
+  - groupName: workergroup
+    replicas: 0
+    minReplicas: 0
+    maxReplicas: 5
+    rayStartParams: {}
+    template:
+      metadata:
+        labels:
+          app.kubernetes.io/instance: raycluster
+          app.kubernetes.io/name: kuberay
+      spec:
+        containers:
+        - image: kubeagi/ray-ml:2.9.0-py39-vllm
+          name: ray-worker
+          resources:
+            limits:
+              cpu: "1"
+              memory: 1G
+              nvidia.com/gpu: 1
+            requests:
+              cpu: "1"
+              memory: 1G
+              nvidia.com/gpu: 1
+          volumeMounts:
+          - mountPath: /tmp/ray
+            name: log-volume
+        volumes:
+        - emptyDir: {}
+          name: log-volume

deploy/charts/arcadia/Chart.yaml

@@ -2,7 +2,7 @@ apiVersion: v2
 name: arcadia
 description: A Helm chart(KubeBB Component) for KubeAGI Arcadia
 type: application
-version: 0.2.8
+version: 0.2.9
 appVersion: "0.1.0"
 
 keywords:

deploy/charts/arcadia/templates/config.yaml

@@ -7,6 +7,15 @@ data:
       kind: Datasource
       name: '{{ .Release.Name }}-minio'
       namespace: '{{ .Release.Namespace }}'
+{{- if gt (len .Values.ray.clusters) 0 }}
+    rayClusters:
+      {{- range .Values.ray.clusters }}
+      - name: {{ .name }}
+        headAddress: {{ .headAddress }}
+        pythonVersion: {{ .pythonVersion }}
+        dashboardHost: {{ .dashboardHost }}
+      {{- end }}
+{{- end }}
     gateway:
       apiServer: 'http://{{ .Release.Name }}-fastchat.{{ .Release.Namespace }}.svc.cluster.local:8000/v1'
       controller: 'http://{{ .Release.Name }}-fastchat.{{ .Release.Namespace }}.svc.cluster.local:21001'
@@ -18,6 +27,7 @@ data:
       kind: VectorStore
       name: '{{ .Release.Name }}-vectorstore'
       namespace: '{{ .Release.Namespace }}'
+
     #streamlit:
     #  image: 172.22.96.34/cluster_system/streamlit:v1.29.0
     #  ingressClassName: portal-ingress

deploy/charts/arcadia/values.yaml

@@ -8,7 +8,7 @@ global:
 # @param resources Resources to be used
 controller:
   loglevel: 3
-  image: kubeagi/arcadia:latest
+  image: kubeagi/arcadia:v0.1.0-20240110-0dd9a1f
   imagePullPolicy: IfNotPresent
   resources:
     limits:
@@ -21,7 +21,7 @@ controller:
 # @section graphql and bff server
 # related project: https://github.com/kubeagi/arcadia/tree/main/apiserver
 apiserver:
-  image: kubeagi/arcadia:latest
+  image: kubeagi/arcadia:v0.1.0-20240110-0dd9a1f
   enableplayground: false
   port: 8081
   ingress:
@@ -146,3 +146,13 @@ postgresql:
   primary:
     initdb:
       scriptsConfigMap: pg-init-data
+
+# @section ray is a unified framework for scaling AI and Python applications.In kubeagi,we use ray for distributed inference
+ray:
+  # clusters provided by ray
+  # For more information on cluster configurations,please refer to http://kubeagi.k8s.com.cn/docs/Configuration/DistributedInference/run-inference-using-ray
+  clusters:
+    - name: 3090-2-GPUs
+      headAddress: raycluster-kuberay-head-svc.kuberay-system.svc:6379
+      pythonVersion: 3.9.18
+      dashboardHost: raycluster-kuberay-head-svc.kuberay-system.svc:8265

deploy/charts/kuberay-operator/.helmignore

@@ -0,0 +1,22 @@
+# Patterns to ignore when building packages.
+# This supports shell glob matching, relative path matching, and
+# negation (prefixed with !). Only one pattern per line.
+.DS_Store
+# Common VCS dirs
+.git/
+.gitignore
+.bzr/
+.bzrignore
+.hg/
+.hgignore
+.svn/
+# Common backup files
+*.swp
+*.bak
+*.tmp
+*~
+# Various IDEs
+.project
+.idea/
+*.tmproj
+.vscode/

deploy/charts/kuberay-operator/Chart.yaml

@@ -0,0 +1,6 @@
+apiVersion: v2
+description: A Helm chart for Kubernetes
+name: kuberay-operator
+version: 1.0.0
+icon: https://github.com/ray-project/ray/raw/master/doc/source/images/ray_header_logo.png
+type: application

deploy/charts/kuberay-operator/README.md

@@ -0,0 +1,117 @@
+# KubeRay Operator
+
+This document provides instructions to install both CRDs (RayCluster, RayJob, RayService) and KubeRay operator with a Helm chart.
+
+## Helm
+
+Make sure the version of Helm is v3+. Currently, [existing CI tests](https://github.com/ray-project/kuberay/blob/master/.github/workflows/helm-lint.yaml) are based on Helm v3.4.1 and v3.9.4.
+
+```sh
+helm version
+```
+
+## Install CRDs and KubeRay operator
+
+* Install a stable version via Helm repository (only supports KubeRay v0.4.0+)
+  ```sh
+  helm repo add kuberay https://ray-project.github.io/kuberay-helm/
+
+  # Install both CRDs and KubeRay operator v1.0.0.
+  helm install kuberay-operator kuberay/kuberay-operator --version 1.0.0
+
+  # Check the KubeRay operator Pod in `default` namespace
+  kubectl get pods
+  # NAME                                READY   STATUS    RESTARTS   AGE
+  # kuberay-operator-6fcbb94f64-mbfnr   1/1     Running   0          17s
+  ```
+
+* Install the nightly version
+  ```sh
+  # Step1: Clone KubeRay repository
+
+  # Step2: Move to `helm-chart/kuberay-operator`
+
+  # Step3: Install KubeRay operator
+  helm install kuberay-operator .
+  ```
+
+* Install KubeRay operator without installing CRDs
+  * In some cases, the installation of the CRDs and the installation of the operator may require different levels of admin permissions, so these two installations could be handled as different steps by different roles.
+  * Use Helm's built-in `--skip-crds` flag to install the operator only. See [this document](https://helm.sh/docs/chart_best_practices/custom_resource_definitions/) for more details.
+  ```sh
+  # Step 1: Install CRDs only (for cluster admin)
+  kubectl create -k "github.com/ray-project/kuberay/manifests/cluster-scope-resources?ref=v1.0.0&timeout=90s"
+
+  # Step 2: Install KubeRay operator only. (for developer)
+  helm install kuberay-operator kuberay/kuberay-operator --version 1.0.0 --skip-crds
+  ``` 
+
+## List the chart
+
+To list the `my-release` deployment:
+
+```sh
+helm ls
+# NAME                    NAMESPACE       REVISION        UPDATED                                 STATUS          CHART                           APP VERSION
+# kuberay-operator        default         1               2023-09-22 02:57:17.306616331 +0000 UTC deployed        kuberay-operator-1.0.0
+```
+
+## Uninstall the Chart
+
+```sh
+# Uninstall the `kuberay-operator` release
+helm uninstall kuberay-operator
+
+# The operator Pod should be removed.
+kubectl get pods
+# No resources found in default namespace.
+```
+
+## Working with Argo CD
+
+If you are using [Argo CD](https://argoproj.github.io) to manage the operator, you will encounter the issue which complains the CRDs too long. Same with [this issue](https://github.com/prometheus-operator/prometheus-operator/issues/4439).
+The recommended solution is to split the operator into two Argo apps, such as:
+
+* The first app just for installing the CRDs with `Replace=true` directly, snippet:
+
+```yaml
+apiVersion: argoproj.io/v1alpha1
+kind: Application
+metadata:
+  name: ray-operator-crds
+spec:
+  project: default
+  source:
+    repoURL: https://github.com/ray-project/kuberay
+    targetRevision: v1.0.0-rc.0
+    path: helm-chart/kuberay-operator/crds
+  destination:
+    server: https://kubernetes.default.svc
+  syncPolicy:
+    syncOptions:
+    - Replace=true
+...
+```
+
+* The second app that installs the Helm chart with `skipCrds=true` (new feature in Argo CD 2.3.0), snippet:
+
+```yaml
+apiVersion: argoproj.io/v1alpha1
+kind: Application
+metadata:
+  name: ray-operator
+spec:
+  source:
+    repoURL: https://github.com/ray-project/kuberay
+    targetRevision: v1.0.0-rc.0
+    path: helm-chart/kuberay-operator
+    helm:
+      skipCrds: true
+  destination:
+    server: https://kubernetes.default.svc
+    namespace: ray-operator
+  syncPolicy:
+    syncOptions:
+    - CreateNamespace=true
+...
+```