[wip] feat: upgrade to Ray 2.50.1 to support top-level Resources and Labels field to KubeRay Autoscaling config

.buildkite/setup-env.sh

@@ -32,7 +32,7 @@ apt-get update
 apt-get install -y python3.11 python3-pip python3-venv python3-dev build-essential
 
 # Install requirements
-pip install --break-system-packages ray[default]==2.46.0
+pip install --break-system-packages ray[default]==2.51.0
 
 # Bypass Git's ownership check due to unconventional user IDs in Docker containers
 git config --global --add safe.directory /workdir

apiserver/test/cluster/cluster_external_redis.yaml

@@ -10,7 +10,7 @@ metadata:
     # ray.io/external-storage-namespace: "my-raycluster-storage"
   name: raycluster-external-redis
 spec:
-  rayVersion: '2.46.0'
+  rayVersion: '2.51.0'
   headGroupSpec:
     # The `rayStartParams` are used to configure the `ray start` command.
     # See https://github.com/ray-project/kuberay/blob/master/docs/guidance/rayStartParams.md for the default settings of `rayStartParams` in KubeRay.
@@ -26,7 +26,7 @@ spec:
       spec:
         containers:
           - name: ray-head
-            image: rayproject/ray:2.46.0
+            image: rayproject/ray:2.51.0
             resources:
               limits:
                 cpu: "1"
@@ -80,7 +80,7 @@ spec:
         spec:
           containers:
             - name: ray-worker
-              image: rayproject/ray:2.46.0
+              image: rayproject/ray:2.51.0
               volumeMounts:
                 - mountPath: /tmp/ray
                   name: ray-logs

apiserver/test/e2e/apiserversdk/constants.go

@@ -1,5 +1,5 @@
 package apiserversdk
 
 const (
-	RayImage = "rayproject/ray:2.46.0"
+	RayImage = "rayproject/ray:2.51.0"
 )

apiserver/test/e2e/defaults.go

@@ -1,6 +1,6 @@
 package e2e
 
 const (
-	RayVersion = "2.46.0"
-	RayImage   = "rayproject/ray:2.46.0"
+	RayVersion = "2.51.0"
+	RayImage   = "rayproject/ray:2.51.0"
 )

benchmark/perf-tests/100-raycluster/raycluster.yaml

@@ -5,15 +5,15 @@ metadata:
   labels:
     perf-test: ray-cluster
 spec:
-  rayVersion: '2.46.0'
+  rayVersion: '2.51.0'
   headGroupSpec:
     rayStartParams:
       dashboard-host: '0.0.0.0'
     template:
       spec:
         containers:
         - name: ray-head
-          image: rayproject/ray:2.46.0
+          image: rayproject/ray:2.51.0
           ports:
           - containerPort: 6379
             name: gcs
@@ -42,7 +42,7 @@ spec:
       spec:
         containers:
         - name: ray-worker
-          image: rayproject/ray:2.46.0
+          image: rayproject/ray:2.51.0
           volumeMounts:
             - mountPath: /tmp/ray
               name: ray-logs

benchmark/perf-tests/100-rayjob/config.yaml

@@ -39,7 +39,7 @@ steps:
     - basename: pytorch-mnist
       objectTemplatePath: pytorch-mnist-rayjob.yaml
       templateFillMap:
-        Image: "rayproject/ray:2.46.0" # replace with image built from images/ray-pytorch
+        Image: "rayproject/ray:2.51.0" # replace with image built from images/ray-pytorch
 - name: Creating RayJobs for Ray Data Image Resizing
   phases:
   - namespaceRange:
@@ -51,7 +51,7 @@ steps:
     - basename: ray-data-image-resize
       objectTemplatePath: ray-data-image-resize.yaml
       templateFillMap:
-        Image: "rayproject/ray:2.46.0" # replace with image built from images/ray-pytorch
+        Image: "rayproject/ray:2.51.0" # replace with image built from images/ray-pytorch
 - name: Wait for RayJobs complete
   measurements:
   - Identifier: WaitForRayJob

benchmark/perf-tests/100-rayjob/pytorch-mnist-rayjob.yaml

@@ -13,7 +13,7 @@ spec:
       CPUS_PER_WORKER: "1"
       OMP_NUM_THREADS: "1" # Set OMP_NUM_THREADS to avoid KeyErorr race condition.
   rayClusterSpec:
-    rayVersion: '2.46.0'
+    rayVersion: '2.51.0'
     headGroupSpec:
       template:
         spec:

benchmark/perf-tests/100-rayjob/ray-data-image-resize.yaml

@@ -8,7 +8,7 @@ spec:
   shutdownAfterJobFinishes: true
   entrypoint: python ray_data_image_resize.py
   rayClusterSpec:
-    rayVersion: '2.46.0'
+    rayVersion: '2.51.0'
     headGroupSpec:
       template:
         spec:

benchmark/perf-tests/1000-raycluster/config.yaml

@@ -43,7 +43,7 @@ steps:
       objectTemplatePath: raycluster.yaml
       templateFillMap:
         Replicas: 3
-        Image: "rayproject/ray:2.46.0"
+        Image: "rayproject/ray:2.51.0"
 - name: Wait for RayClusters ready
   measurements:
   - Identifier: WaitForRayCluster

benchmark/perf-tests/1000-raycluster/raycluster.yaml

@@ -5,7 +5,7 @@ metadata:
   labels:
     perf-test: ray-cluster
 spec:
-  rayVersion: '2.46.0'
+  rayVersion: '2.51.0'
   headGroupSpec:
     serviceType: ClusterIP
     rayStartParams:

benchmark/perf-tests/1000-rayjob/config.yaml

@@ -39,7 +39,7 @@ steps:
     - basename: pytorch-mnist
       objectTemplatePath: pytorch-mnist-rayjob.yaml
       templateFillMap:
-        Image: "rayproject/ray:2.46.0"
+        Image: "rayproject/ray:2.51.0"
 - name: Creating RayJobs for Ray Data Image Resizing
   phases:
   - namespaceRange:
@@ -51,7 +51,7 @@ steps:
     - basename: ray-data-image-resize
       objectTemplatePath: ray-data-image-resize.yaml
       templateFillMap:
-        Image: "rayproject/ray:2.46.0"
+        Image: "rayproject/ray:2.51.0"
 - name: Wait for RayJobs complete
   measurements:
   - Identifier: WaitForRayJob

benchmark/perf-tests/1000-rayjob/pytorch-mnist-rayjob.yaml

@@ -26,7 +26,7 @@ spec:
             requests:
               cpu: "10m"
   rayClusterSpec:
-    rayVersion: '2.46.0'
+    rayVersion: '2.51.0'
     headGroupSpec:
       rayStartParams:
         disable-usage-stats: 'true'

benchmark/perf-tests/1000-rayjob/ray-data-image-resize.yaml

@@ -26,7 +26,7 @@ spec:
             requests:
               cpu: "10m"
   rayClusterSpec:
-    rayVersion: '2.46.0'
+    rayVersion: '2.51.0'
     headGroupSpec:
       rayStartParams:
         disable-usage-stats: 'true'

benchmark/perf-tests/10000-raycluster/config.yaml

@@ -43,7 +43,7 @@ steps:
       objectTemplatePath: raycluster.yaml
       templateFillMap:
         Replicas: 3
-        Image: "rayproject/ray:2.46.0"
+        Image: "rayproject/ray:2.51.0"
 - name: Wait for RayClusters ready
   measurements:
   - Identifier: WaitForRayCluster

benchmark/perf-tests/10000-raycluster/raycluster.yaml

@@ -5,7 +5,7 @@ metadata:
   labels:
     perf-test: ray-cluster
 spec:
-  rayVersion: '2.46.0'
+  rayVersion: '2.51.0'
   headGroupSpec:
     serviceType: ClusterIP
     rayStartParams:

benchmark/perf-tests/10000-rayjob/config.yaml

@@ -38,7 +38,7 @@ steps:
     objectBundle:
     - basename: pytorch-mnist
       objectTemplatePath: pytorch-mnist-rayjob.yaml
-      Image: "rayproject/ray:2.46.0"
+      Image: "rayproject/ray:2.51.0"
 - name: Creating RayJobs for Ray Data Image Resizing
   phases:
   - namespaceRange:
@@ -49,7 +49,7 @@ steps:
     objectBundle:
     - basename: ray-data-image-resize
       objectTemplatePath: ray-data-image-resize.yaml
-      Image: "rayproject/ray:2.46.0"
+      Image: "rayproject/ray:2.51.0"
 - name: Wait for RayJobs complete
   measurements:
   - Identifier: WaitForRayJob

benchmark/perf-tests/10000-rayjob/pytorch-mnist-rayjob.yaml

@@ -26,7 +26,7 @@ spec:
             requests:
               cpu: "10m"
   rayClusterSpec:
-    rayVersion: '2.46.0'
+    rayVersion: '2.51.0'
     headGroupSpec:
       rayStartParams:
         disable-usage-stats: 'true'

benchmark/perf-tests/10000-rayjob/ray-data-image-resize.yaml

@@ -26,7 +26,7 @@ spec:
             requests:
               cpu: "10m"
   rayClusterSpec:
-    rayVersion: '2.46.0'
+    rayVersion: '2.51.0'
     headGroupSpec:
       rayStartParams:
         disable-usage-stats: 'true'

benchmark/perf-tests/5000-raycluster/config.yaml

@@ -43,7 +43,7 @@ steps:
       objectTemplatePath: raycluster.yaml
       templateFillMap:
         Replicas: 3
-        Image: "rayproject/ray:2.46.0"
+        Image: "rayproject/ray:2.51.0"
 - name: Wait for RayClusters ready
   measurements:
   - Identifier: WaitForRayCluster

benchmark/perf-tests/5000-raycluster/raycluster.yaml

@@ -5,7 +5,7 @@ metadata:
   labels:
     perf-test: ray-cluster
 spec:
-  rayVersion: '2.46.0'
+  rayVersion: '2.51.0'
   headGroupSpec:
     serviceType: ClusterIP
     rayStartParams:

benchmark/perf-tests/5000-rayjob/config.yaml

@@ -38,7 +38,7 @@ steps:
     objectBundle:
     - basename: pytorch-mnist
       objectTemplatePath: pytorch-mnist-rayjob.yaml
-      Image: "rayproject/ray:2.46.0"
+      Image: "rayproject/ray:2.51.0"
 - name: Creating RayJobs for Ray Data Image Resizing
   phases:
   - namespaceRange:
@@ -49,7 +49,7 @@ steps:
     objectBundle:
     - basename: ray-data-image-resize
       objectTemplatePath: ray-data-image-resize.yaml
-      Image: "rayproject/ray:2.46.0"
+      Image: "rayproject/ray:2.51.0"
 - name: Wait for RayJobs complete
   measurements:
   - Identifier: WaitForRayJob

benchmark/perf-tests/5000-rayjob/pytorch-mnist-rayjob.yaml

@@ -26,7 +26,7 @@ spec:
             requests:
               cpu: "10m"
   rayClusterSpec:
-    rayVersion: '2.46.0'
+    rayVersion: '2.51.0'
     headGroupSpec:
       rayStartParams:
         disable-usage-stats: 'true'

benchmark/perf-tests/5000-rayjob/ray-data-image-resize.yaml

@@ -26,7 +26,7 @@ spec:
             requests:
               cpu: "10m"
   rayClusterSpec:
-    rayVersion: '2.46.0'
+    rayVersion: '2.51.0'
     headGroupSpec:
       rayStartParams:
         disable-usage-stats: 'true'

helm-chart/ray-cluster/README.md

@@ -73,7 +73,7 @@ helm uninstall raycluster
 | Key | Type | Default | Description |
 |-----|------|---------|-------------|
 | image.repository | string | `"rayproject/ray"` | Image repository. |
-| image.tag | string | `"2.46.0"` | Image tag. |
+| image.tag | string | `"2.51.0"` | Image tag. |
 | image.pullPolicy | string | `"IfNotPresent"` | Image pull policy. |
 | nameOverride | string | `"kuberay"` | String to partially override release name. |
 | fullnameOverride | string | `""` | String to fully override release name. |

helm-chart/ray-cluster/ci/ci-values.yaml

@@ -1,3 +1,3 @@
 image:
   repository: rayproject/ray
-  tag: 2.46.0
+  tag: 2.51.0

helm-chart/ray-cluster/tests/raycluster_test.yaml

@@ -19,11 +19,11 @@ tests:
   - it: Should set Ray version when `head.rayVersion` is set
     set:
       head:
-        rayVersion: 2.46.0
+        rayVersion: 2.51.0
     asserts:
       - equal:
           path: spec.rayVersion
-          value: 2.46.0
+          value: 2.51.0
 
   - it: Should enable in-tree auto scaling when `head.enableInTreeAutoscaling` is `true`
     set:

helm-chart/ray-cluster/values.yaml

@@ -9,7 +9,7 @@ image:
   # -- Image repository.
   repository: rayproject/ray
   # -- Image tag.
-  tag: 2.46.0
+  tag: 2.51.0
   # -- Image pull policy.
   pullPolicy: IfNotPresent
 
@@ -59,7 +59,7 @@ common:
 head:
   # rayVersion determines the autoscaler's image version.
   # It should match the Ray version in the image of the containers.
-  # rayVersion: 2.46.0
+  # rayVersion: 2.51.0
 
   # If enableInTreeAutoscaling is true, the autoscaler sidecar will be added to the Ray head pod.
   # Ray autoscaler integration is supported only for Ray versions >= 1.11.0

kubectl-plugin/config/samples/create-cluster.complete.yaml

@@ -6,8 +6,8 @@ labels:
 annotations:
   dead: beef
 
-ray-version: 2.46.0
-image: rayproject/ray:2.46.0
+ray-version: 2.51.0
+image: rayproject/ray:2.51.0
 service-account: [email protected]
 
 head:

kubectl-plugin/pkg/cmd/create/create_cluster_test.go

@@ -54,8 +54,8 @@ func TestRayCreateClusterComplete(t *testing.T) {
 		"should set the image to the same version as the ray version when the image is the default and the ray version is not the default": {
 			args:          []string{"testRayClusterName"},
 			image:         defaultImageWithTag,
-			rayVersion:    "2.46.0",
-			expectedImage: fmt.Sprintf("%s:2.46.0", defaultImage),
+			rayVersion:    "2.51.0",
+			expectedImage: fmt.Sprintf("%s:2.51.0", defaultImage),
 		},
 	}
 
@@ -156,8 +156,8 @@ func TestSwitchesIncompatibleWithConfigFilePresent(t *testing.T) {
 		"should error when incompatible flags are used": {
 			args: []string{
 				"sample-cluster",
-				"--ray-version", "2.46.0",
-				"--image", "rayproject/ray:2.46.0",
+				"--ray-version", "2.51.0",
+				"--image", "rayproject/ray:2.51.0",
 				"--head-cpu", "1",
 				"--head-memory", "5Gi",
 				"--head-gpu", "1",
@@ -249,8 +249,8 @@ func TestNewCreateClusterCommand(t *testing.T) {
 		"should succeed when all flags are provided": {
 			args: []string{
 				"sample-cluster",
-				"--ray-version", "2.46.0",
-				"--image", "rayproject/ray:2.46.0",
+				"--ray-version", "2.51.0",
+				"--image", "rayproject/ray:2.51.0",
 				"--head-cpu", "1",
 				"--head-memory", "5Gi",
 				"--head-gpu", "1",
@@ -284,7 +284,7 @@ func TestNewCreateClusterCommand(t *testing.T) {
 			args: []string{
 				"sample-cluster",
 				"--file", "config.yaml",
-				"--ray-version", "2.46.0",
+				"--ray-version", "2.51.0",
 				"--dry-run",
 			},
 			expectError: "the following flags are incompatible with --file: [ray-version]",

kubectl-plugin/pkg/util/constant.go

@@ -1,7 +1,7 @@
 package util
 
 const (
-	RayVersion = "2.46.0"
+	RayVersion = "2.51.0"
 	RayImage   = "rayproject/ray:" + RayVersion
 
 	RayClusterLabelKey   = "ray.io/cluster"