Azure-Samples · marosset · Mar 30, 2026 · Mar 30, 2026
diff --git a/README.md b/README.md
@@ -10,7 +10,6 @@ Clone the repository to your local machine. Then make sure you have all the prer
 
 1. An Azure subscription. If you don't have an Azure subscription, you can create a free account [here](https://azure.microsoft.com/free/).
 1. The Azure CLI installed on your local machine. You can install the Azure CLI by following the instructions [here](https://docs.microsoft.com/cli/azure/install-azure-cli).
-1. The [Azure Kubernetes Service Preview extension](https://learn.microsoft.com/azure/aks/draft#install-the-aks-preview-azure-cli-extension) must be installed.
 1. [Helm](https://helm.sh/docs/intro/install/) must be installed.
 1. [Terraform client tools](https://developer.hashicorp.com/terraform/install) or [OpenTofu](https://opentofu.org/) must be installed. This guide makes use of Terrafrom, however the modules used should be compatible with OpenTofu.
 

diff --git a/aks-automatic/deploy.sh b/aks-automatic/deploy.sh
@@ -1,5 +1,8 @@
 #!/bin/bash
 
+# Ray version to use (can be overridden via environment variable)
+RAY_VERSION=${RAY_VERSION:-"2.54.1"}
+
 # Check if the user is logged into Azure CLI
 if ! az account show > /dev/null 2>&1; then
     echo "Please login to Azure CLI using 'az login' before running this script."
@@ -86,7 +89,7 @@ helm upgrade \
 --timeout 10m0s \
 --namespace "$kuberay_namespace" \
 --create-namespace kuberay-operator kuberay/kuberay-operator \
---version 1.1.1
+--version 1.6.0
 
 # Output the pods in the kuberay namespace
 kubectl get pods -n $kuberay_namespace
@@ -163,6 +166,10 @@ EOF
 # Download the PyTorch MNIST job YAML file
 curl -LO https://raw.githubusercontent.com/ray-project/kuberay/master/ray-operator/config/samples/pytorch-mnist/ray-job.pytorch-mnist.yaml
 
+# Update the Ray version in the downloaded job YAML
+sed -i "s/rayVersion: '.*'/rayVersion: '${RAY_VERSION}'/g" ray-job.pytorch-mnist.yaml
+sed -i "s|image: rayproject/ray:.*|image: rayproject/ray:${RAY_VERSION}|g" ray-job.pytorch-mnist.yaml
+
 # Train a PyTorch Model on Fashion MNIST
 kubectl apply -n $kuberay_namespace -f ray-job.pytorch-mnist.yaml
 

diff --git a/aks-automatic/main.tf b/aks-automatic/main.tf
@@ -19,7 +19,7 @@ resource "tls_private_key" "ssh_key" {
 }
 
 resource "azapi_resource" "aks_auto" {
-  type      = "Microsoft.ContainerService/managedClusters@2024-06-02-preview"
+  type      = "Microsoft.ContainerService/managedClusters@2026-01-01"
   name      = "aks-${var.project_prefix}-${random_string.suffix.result}"
   parent_id = azurerm_resource_group.rg.id
   location  = var.location
@@ -28,7 +28,7 @@ resource "azapi_resource" "aks_auto" {
   body = jsonencode({
 
     properties = {
-      kubernetesVersion = "1.31"
+      kubernetesVersion = "1.33"
       nodeResourceGroup = "MC-aks-${var.project_prefix}-${random_string.suffix.result}"
       agentPoolProfiles = [
         {
@@ -97,7 +97,7 @@ resource "null_resource" "wait_for_aks" {
 }
 
 resource "azapi_update_resource" "k8s-default-node-pool-systempool-taint" {
-  type        = "Microsoft.ContainerService/managedClusters@2024-06-02-preview"
+  type        = "Microsoft.ContainerService/managedClusters@2026-01-01"
   resource_id = azapi_resource.aks_auto.id
   body = jsonencode({
     properties = {

diff --git a/aks-automatic/ray-job.pytorch-mnist-persist-logs.yaml b/aks-automatic/ray-job.pytorch-mnist-persist-logs.yaml
@@ -38,15 +38,15 @@ spec:
       CPUS_PER_WORKER: "2"
   # rayClusterSpec specifies the RayCluster instance to be created by the RayJob controller.
   rayClusterSpec:
-    rayVersion: '2.41.0'
+    rayVersion: '2.54.1'
     headGroupSpec:
       rayStartParams: {}
       # Pod template
       template:
         spec:
           containers:
             - name: ray-head
-              image: rayproject/ray:2.41.0
+              image: rayproject/ray:2.54.1
               ports:
                 - containerPort: 6379
                   name: gcs-server
@@ -104,7 +104,7 @@ spec:
           spec:
             containers:
               - name: ray-worker
-                image: rayproject/ray:2.41.0
+                image: rayproject/ray:2.54.1
                 resources:
                   limits:
                     cpu: "3"

diff --git a/aks-classic/deploy.sh b/aks-classic/deploy.sh
@@ -1,5 +1,8 @@
 #!/bin/bash
 
+# Ray version to use (can be overridden via environment variable)
+RAY_VERSION=${RAY_VERSION:-"2.54.1"}
+
 # Check if the user is logged into Azure CLI
 if ! az account show > /dev/null 2>&1; then
     echo "Please login to Azure CLI using 'az login' before running this script."
@@ -53,14 +56,18 @@ helm upgrade \
 --timeout 10m0s \
 --namespace "$kuberay_namespace" \
 --create-namespace kuberay-operator kuberay/kuberay-operator \
---version 1.1.1
+--version 1.6.0
 
 # Output the pods in the kuberay namespace
 kubectl get pods -n $kuberay_namespace
 
 # Download the PyTorch MNIST job YAML file
 curl -LO https://raw.githubusercontent.com/ray-project/kuberay/master/ray-operator/config/samples/pytorch-mnist/ray-job.pytorch-mnist.yaml
 
+# Update the Ray version in the downloaded job YAML
+sed -i "s/rayVersion: '.*'/rayVersion: '${RAY_VERSION}'/g" ray-job.pytorch-mnist.yaml
+sed -i "s|image: rayproject/ray:.*|image: rayproject/ray:${RAY_VERSION}|g" ray-job.pytorch-mnist.yaml
+
 # Train a PyTorch Model on Fashion MNIST
 kubectl apply -n $kuberay_namespace -f ray-job.pytorch-mnist.yaml
 

diff --git a/aks-classic/main.tf b/aks-classic/main.tf
@@ -27,10 +27,11 @@ resource "azurerm_kubernetes_cluster" "k8s" {
   }
 
   default_node_pool {
-    name       = "systempool"
-    vm_size    = var.system_node_pool_vm_size
-    node_count = var.system_node_pool_node_count
-    tags = { owner = var.resource_group_owner }
+    name                         = "systempool"
+    vm_size                      = var.system_node_pool_vm_size
+    node_count                   = var.system_node_pool_node_count
+    only_critical_addons_enabled = true
+    tags                         = { owner = var.resource_group_owner }
   }
 
   linux_profile {
@@ -71,7 +72,7 @@ resource "null_resource" "wait_for_aks" {
 }
 
 resource "azapi_update_resource" "k8s-default-node-pool-systempool-taint" {
-  type        = "Microsoft.ContainerService/managedClusters@2024-09-02-preview"
+  type        = "Microsoft.ContainerService/managedClusters@2026-01-01"
   resource_id = azurerm_kubernetes_cluster.k8s.id
   body = jsonencode({
     properties = {

diff --git a/docs/deploy-ray-aks.md b/docs/deploy-ray-aks.md
@@ -12,7 +12,6 @@ This article provides two methods to deploy the Ray cluster on AKS:
 * Review the [Ray cluster on AKS overview](./ray-on-aks.md) to understand the components and deployment process.
 * An Azure subscription. If you don't have an Azure subscription, you can create a free account [here](https://azure.microsoft.com/free/).
 * The Azure CLI installed on your local machine. You can install it using the instructions in [How to install the Azure CLI](https://docs.microsoft.com/cli/azure/install-azure-cli).
-* The [Azure Kubernetes Service Preview extension](https://learn.microsoft.com/azure/aks/draft#install-the-aks-preview-azure-cli-extension) installed.
 * [Helm](https://helm.sh/docs/intro/install/) installed.
 * [Terraform client tools](https://developer.hashicorp.com/terraform/install) or [OpenTofu](https://opentofu.org/) installed. This article uses Terrafrom, but the modules used should be compatible with OpenTofu.
 

diff --git a/sample-tuning-setup/direct-blob-access/deploy.sh b/sample-tuning-setup/direct-blob-access/deploy.sh
@@ -255,7 +255,7 @@ fi
 
 echo "Deploying KubeRay Operator in ${KUBERAY_NAMESPACE} namespace"
 helm repo add kuberay https://ray-project.github.io/kuberay-helm/
-helm install kuberay-operator kuberay/kuberay-operator --version 1.4.2 --namespace ${KUBERAY_NAMESPACE} --create-namespace
+helm install kuberay-operator kuberay/kuberay-operator --version 1.6.0 --namespace ${KUBERAY_NAMESPACE} --create-namespace
 
 # Output the pods in the kuberay namespace
 kubectl get pods -n ${KUBERAY_NAMESPACE}

diff --git a/sample-tuning-setup/direct-blob-access/rayjob_direct_access.tpl b/sample-tuning-setup/direct-blob-access/rayjob_direct_access.tpl
@@ -13,7 +13,7 @@ spec:
     # Uncomment the next line to experiment with autoscaling.
     # enableInTreeAutoscaling: true
     # The version of Ray you are using. Make sure all Ray containers are running this version of Ray.
-    rayVersion: '2.43.0'
+    rayVersion: '2.54.1'
     headGroupSpec:
       # Kubernetes Service Type, valid values are 'ClusterIP', 'NodePort' and 'LoadBalancer'
       serviceType: ClusterIP

diff --git a/sample-tuning-setup/direct-blob-access/variables.tf b/sample-tuning-setup/direct-blob-access/variables.tf
@@ -22,7 +22,7 @@ variable "project_prefix" {
 
 variable "azure_kubernetes_version" {
   description = "Version of the azure kubernetes"
-  default     = "1.33.0"
+  default     = "1.33"
   type        = string
 }
 
@@ -96,7 +96,7 @@ variable "azure_storage_profile" {
 variable "kuberay_version" {
   description                    = "Kuberay version that needs to be installed"
   type                           = string
-  default                        = "1.4.2"
+  default                        = "1.6.0"
 }
 
 variable "kuberay_scrape_config_path" {

diff --git a/sample-tuning-setup/rayjob.yaml b/sample-tuning-setup/rayjob.yaml
@@ -12,7 +12,7 @@ spec:
     # Uncomment the next line to experiment with autoscaling.
     # enableInTreeAutoscaling: true
     # The version of Ray you are using. Make sure all Ray containers are running this version of Ray.
-    rayVersion: '2.43.0'
+    rayVersion: '2.54.1'
     headGroupSpec:
       # Kubernetes Service Type, valid values are 'ClusterIP', 'NodePort' and 'LoadBalancer'
       serviceType: ClusterIP
@@ -24,7 +24,7 @@ spec:
           containers:
           # The Ray head container
           - name: ray-head
-            image: rayproject/ray-ml:2.43.0.84f276-py310-cpu
+            image: rayproject/ray:2.54.1
             imagePullPolicy: Always
             # Optimal resource allocation will depend on your Kubernetes infrastructure and might
             # require some experimentation.
@@ -67,7 +67,7 @@ spec:
         spec:
           containers:
           - name: machine-learning # must consist of lower case alphanumeric characters or '-', and must start and end with an alphanumeric character (e.g. 'my-name',  or '123-abc')
-            image: rayproject/ray-ml:2.43.0.84f276-py310-cpu
+            image: rayproject/ray:2.54.1
             # Optimal resource allocation will depend on your Kubernetes infrastructure and might
             # require some experimentation.
             # Setting requests=limits is recommended with Ray. K8s limits are used for Ray-internal

diff --git a/sample-tuning-setup/terraform/variables.tf b/sample-tuning-setup/terraform/variables.tf
@@ -22,7 +22,7 @@ variable "project_prefix" {
 
 variable "azure_kubernetes_version" {
   description = "Version of the azure kubernetes"
-  default = "1.32"
+  default = "1.33"
   type        = string
 }
 
@@ -96,7 +96,7 @@ variable "azure_storage_profile" {
 variable "kuberay_version" {
   description                    = "Kuberay version that needs to be installed"
   type                           = string
-  default                        = "1.4.2"
+  default                        = "1.6.0"
 }
 
 variable "kuberay_persistent_volume_claim_name" {