[Feature] Run config tests with the latest release of KubeRay operator (ray-project#858)

Currently, we only run config tests with the nightly KubeRay operator. However, we also need to keep the compatibility of the current sample YAML files with the latest KubeRay operator release. If not, some bugs may happen (e.g. ray-project#856).

Author: kevin85421

.github/workflows/actions/configuration/action.yaml

@@ -48,12 +48,22 @@ runs:
       docker images ls -a
     shell: bash
 
-  - name: Run configuration tests for sample YAML files.
-    # Depends on the KubeRay operator image built in previous steps.
+  - name: Run tests for sample YAML files with the nightly operator.
+    # Depends on the KubeRay operator image built in previous steps
+    env:
+      GITHUB_ACTIONS: true
+      RAY_IMAGE: rayproject/ray:${{ inputs.ray_version }}
+      OPERATOR_IMAGE: kuberay/operator:${{ steps.vars.outputs.sha_short }}
+    run: |
+      python tests/test_sample_raycluster_yamls.py
+    shell: bash
+
+  - name: Run tests for sample YAML files with the latest KubeRay release.
+    # Depends on latest KubeRay release.
+    env:
+      GITHUB_ACTIONS: true
+      RAY_IMAGE: rayproject/ray:${{ inputs.ray_version }}
+      OPERATOR_IMAGE: kuberay/operator:v0.4.0 # The operator image in the latest KubeRay release.
     run: |
-      echo "Using Ray image ${{ inputs.ray_version }}"
-      GITHUB_ACTIONS=true \
-      RAY_IMAGE="rayproject/ray:${{ inputs.ray_version }}" \
-      OPERATOR_IMAGE="kuberay/operator:${{ steps.vars.outputs.sha_short }}" \
       python tests/test_sample_raycluster_yamls.py
     shell: bash

.github/workflows/test-job.yaml

@@ -378,3 +378,4 @@ jobs:
       - uses: ./.github/workflows/actions/configuration
         with:
           ray_version: 2.1.0
+

docs/guidance/volcano-integration.md

@@ -44,6 +44,7 @@ metadata:
 spec:
   rayVersion: '2.2.0'
   headGroupSpec:
+    serviceType: ClusterIP
     rayStartParams:
       block: 'true'
     replicas: 1
@@ -107,6 +108,7 @@ metadata:
 spec:
   rayVersion: '2.2.0'
   headGroupSpec:
+    serviceType: ClusterIP
     rayStartParams:
       block: 'true'
     replicas: 1
@@ -220,6 +222,7 @@ metadata:
 spec:
   rayVersion: '2.2.0'
   headGroupSpec:
+    serviceType: ClusterIP
     rayStartParams:
       block: 'true'
     replicas: 1

helm-chart/ray-cluster/values.yaml

@@ -200,5 +200,6 @@ additionalWorkerGroups:
     sidecarContainers: []
 
 # Configuration for Head's Kubernetes Service
-service: {}
-#   type: ClusterIP # The default type is ClusterIP.
+service:
+  # This is optional, and the default is ClusterIP.
+  type: ClusterIP

ray-operator/config/samples/ray-cluster.autoscaler.large.yaml

@@ -57,6 +57,7 @@ spec:
         memory: "512Mi"
   # Ray head pod template
   headGroupSpec:
+    serviceType: ClusterIP # optional
     # the following params are used to complete the ray start: ray start --head --block --port=6379 ...
     rayStartParams:
       # Flag "no-monitor" will be automatically set when autoscaling is enabled.

ray-operator/config/samples/ray-cluster.autoscaler.yaml

@@ -49,6 +49,7 @@ spec:
         memory: "512Mi"
   # Ray head pod template
   headGroupSpec:
+    serviceType: ClusterIP # optional
     # the following params are used to complete the ray start: ray start --head --block ...
     rayStartParams:
       dashboard-host: '0.0.0.0'

ray-operator/config/samples/ray-cluster.external-redis.yaml

@@ -79,6 +79,7 @@ metadata:
 spec:
   rayVersion: '2.2.0'
   headGroupSpec:
+    serviceType: ClusterIP # optional
     replicas: 1
     rayStartParams:
       dashboard-host: '0.0.0.0'

ray-operator/config/samples/ray-cluster.getting-started.yaml

@@ -13,6 +13,7 @@ spec:
   rayVersion: '2.2.0' # should match the Ray version in the image of the containers
   # Ray head pod template
   headGroupSpec:
+    serviceType: ClusterIP # optional
     # the following params are used to complete the ray start: ray start --head --block ...
     rayStartParams:
       dashboard-host: '0.0.0.0'

ray-operator/config/samples/ray-cluster.heterogeneous.yaml

@@ -38,6 +38,7 @@ spec:
   ######################headGroupSpecs#################################
   # Ray head pod template
   headGroupSpec:
+    serviceType: ClusterIP # optional
     # the following params are used to complete the ray start: ray start --head --block ...
     rayStartParams:
       dashboard-host: '0.0.0.0'

ray-operator/config/samples/ray-cluster.mini.yaml

@@ -13,6 +13,7 @@ spec:
   rayVersion: '2.2.0' # should match the Ray version in the image of the containers
   # Ray head pod template
   headGroupSpec:
+    serviceType: ClusterIP # optional
     # the following params are used to complete the ray start: ray start --head --block --redis-port=6379 ...
     rayStartParams:
       dashboard-host: '0.0.0.0'

ray-operator/config/samples/ray_v1alpha1_rayjob.yaml

@@ -18,6 +18,7 @@ spec:
     rayVersion: '2.2.0' # should match the Ray version in the image of the containers
     # Ray head pod template
     headGroupSpec:
+      serviceType: ClusterIP # optional
       # the following params are used to complete the ray start: ray start --head --block --redis-port=6379 ...
       rayStartParams:
         dashboard-host: '0.0.0.0'

ray-operator/config/samples/ray_v1alpha1_rayservice.yaml

@@ -46,6 +46,7 @@ spec:
     ######################headGroupSpecs#################################
     # Ray head pod template.
     headGroupSpec:
+      serviceType: ClusterIP # optional
       # the following params are used to complete the ray start: ray start --head --block --redis-port=6379 ...
       rayStartParams:
         port: '6379' # should match container port named gcs-server

ray-operator/config/security/ray-cluster.pod-security.yaml

@@ -13,6 +13,7 @@ spec:
   rayVersion: '2.2.0'
   # Ray head pod configuration
   headGroupSpec:
+    serviceType: ClusterIP # optional
     # for the head group, replicas should always be 1.
     # headGroupSpec.replicas is deprecated in KubeRay >= 0.3.0.
     replicas: 1

tests/framework/utils.py

@@ -17,6 +17,7 @@ class CONST:
     # Docker images
     OPERATOR_IMAGE_KEY = "kuberay-operator-image"
     RAY_IMAGE_KEY = "ray-image"
+    KUBERAY_LATEST_RELEASE = "kuberay/operator:v0.4.0"
 
     # Kubernetes API clients
     K8S_CR_CLIENT_KEY = "k8s-cr-api-client"
@@ -109,12 +110,21 @@ def download_images():
 
     def __install_crd_and_operator(self):
         """Install both CRD and KubeRay operator by kuberay-operator chart"""
-        logger.info("Install both CRD and KubeRay operator by kuberay-operator chart")
         repo, tag = self.docker_image_dict[CONST.OPERATOR_IMAGE_KEY].split(':')
-        shell_subprocess_run(
-            f"helm install kuberay-operator {CONST.HELM_CHART_ROOT}/kuberay-operator/ "
-            f"--set image.repository={repo},image.tag={tag}"
-        )
+        if f"{repo}:{tag}" == CONST.KUBERAY_LATEST_RELEASE:
+            logger.info("Install both CRD and KubeRay operator with the latest release.")
+            shell_subprocess_run(
+                "helm repo add kuberay https://ray-project.github.io/kuberay-helm/"
+            )
+            shell_subprocess_run(
+                f"helm install kuberay-operator kuberay/kuberay-operator --version {tag[1:]}"
+            )
+        else:
+            logger.info("Install both nightly CRD and KubeRay operator by kuberay-operator chart")
+            shell_subprocess_run(
+                f"helm install kuberay-operator {CONST.HELM_CHART_ROOT}/kuberay-operator/ "
+                f"--set image.repository={repo},image.tag={tag}"
+            )
 
 def shell_subprocess_run(command, check = True):
     """