test: revert e2e workflow

kryanbeane · kryanbeane · commit fa3dc463d04f · 2025-10-01T07:49:41.000+01:00
diff --git a/.github/workflows/e2e_tests.yaml b/.github/workflows/e2e_tests.yaml
@@ -5,22 +5,21 @@ on:
   pull_request:
     branches:
       - main
-      - "release-*"
+      - 'release-*'
       - ray-jobs-feature
-      - kueue-integration
     paths-ignore:
-      - "docs/**"
-      - "**.adoc"
-      - "**.md"
-      - "LICENSE"
+      - 'docs/**'
+      - '**.adoc'
+      - '**.md'
+      - 'LICENSE'
 
 concurrency:
   group: ${{ github.head_ref }}-${{ github.workflow }}
   cancel-in-progress: true
 
 env:
-  CODEFLARE_OPERATOR_IMG: "quay.io/project-codeflare/codeflare-operator:dev"
   KUEUE_VERSION: "v0.13.4"
+  KUBERAY_VERSION: "v1.4.0"
 
 jobs:
   kubernetes:
@@ -35,32 +34,15 @@ jobs:
       - name: Checkout common repo code
         uses: actions/checkout@v4
         with:
-          repository: "project-codeflare/codeflare-common"
-          ref: "main"
-          path: "common"
-
-      - name: Checkout CodeFlare operator repository
-        uses: actions/checkout@v4
-        with:
-          repository: project-codeflare/codeflare-operator
-          path: codeflare-operator
-
-      - name: Set Go
-        uses: actions/setup-go@v5
-        with:
-          go-version-file: "./codeflare-operator/go.mod"
-          cache-dependency-path: "./codeflare-operator/go.sum"
-
-      - name: Set up gotestfmt
-        uses: gotesttools/gotestfmt-action@v2
-        with:
-          token: ${{ secrets.GITHUB_TOKEN }}
+          repository: 'project-codeflare/codeflare-common'
+          ref: 'main'
+          path: 'common'
 
       - name: Set up specific Python version
         uses: actions/setup-python@v5
         with:
-          python-version: "3.12"
-          cache: "pip" # caching pip dependencies
+          python-version: '3.11'
+          cache: 'pip' # caching pip dependencies
 
       - name: Setup NVidia GPU environment for KinD
         uses: ./common/github-actions/nvidia-gpu-setup
@@ -73,48 +55,12 @@ jobs:
       - name: Install NVidia GPU operator for KinD
         uses: ./common/github-actions/nvidia-gpu-operator
 
-      - name: Wait for nodes to be ready
-        run: |
-          echo "Waiting for all nodes to be ready..."
-          kubectl wait --for=condition=Ready nodes --all --timeout=300s
-
-          echo "Checking node status..."
-          kubectl get nodes -o wide
-
-          echo "Checking for CNI readiness..."
-          for i in {1..30}; do
-            if kubectl get nodes -o jsonpath='{range .items[*]}{.metadata.name}{"\t"}{.status.conditions[?(@.type=="Ready")].status}{"\n"}{end}' | grep -v "True"; then
-              echo "Waiting for CNI to initialize (attempt $i/30)..."
-              sleep 10
-            else
-              echo "All nodes are ready!"
-              break
-            fi
-          done
-
-          # Final verification
-          kubectl describe nodes | grep -E "Ready|NetworkReady|RuntimeReady|PodCIDR"
-
-      - name: Deploy CodeFlare stack
+      - name: Deploy KubeRay and Kueue
         id: deploy
         run: |
-          cd codeflare-operator
-          echo Setting up CodeFlare stack
-          make setup-e2e
-          echo Deploying CodeFlare operator
-          make deploy -e IMG="${CODEFLARE_OPERATOR_IMG}" -e ENV="e2e"
-          kubectl wait --timeout=120s --for=condition=Available=true deployment -n openshift-operators codeflare-operator-manager
-          cd ..
-
-      - name: Verify CodeFlare deployment
-        run: |
-          # Wait for Kueue to be ready
-          echo "Waiting for Kueue controller to be ready..."
-          kubectl wait --for=condition=Available --timeout=300s deployment -n kueue-system kueue-controller-manager || {
-            echo "Kueue deployment status:"
-            kubectl get all -n kueue-system
-            exit 1
-          }
+          # Deploy KubeRay operator
+          echo "Deploying KubeRay ${KUBERAY_VERSION}..."
+          kubectl apply --server-side -k "github.com/ray-project/kuberay/ray-operator/config/default?ref=${KUBERAY_VERSION}&timeout=180s"
 
           # Wait for KubeRay to be ready
           echo "Waiting for KubeRay operator to be ready..."
@@ -124,44 +70,43 @@ jobs:
             exit 1
           }
 
-          # Verify webhook certificates
-          echo "Checking CodeFlare operator webhook certificates..."
-          kubectl get secret -n openshift-operators codeflare-operator-webhook-server-cert -o jsonpath='{.data.ca\.crt}' | base64 -d | openssl x509 -noout -text || {
-            echo "Warning: Webhook certificate might be missing or invalid"
+          # Deploy Kueue
+          echo "Deploying Kueue ${KUEUE_VERSION}..."
+          kubectl apply --server-side -f https://github.com/kubernetes-sigs/kueue/releases/download/${KUEUE_VERSION}/manifests.yaml
+
+          # Wait for Kueue to be ready
+          echo "Waiting for Kueue controller to be ready..."
+          kubectl wait --for=condition=Available --timeout=300s deployment -n kueue-system kueue-controller-manager || {
+            echo "Kueue deployment status:"
+            kubectl get all -n kueue-system
+            exit 1
           }
 
+          echo "✓ KubeRay and Kueue deployed successfully"
+
       - name: Add user to KinD
         uses: ./common/github-actions/kind-add-user
         with:
           user-name: sdk-user
 
       - name: Configure RBAC for sdk user with limited permissions
         run: |
-          # CRD permissions for discovering resource types
-          kubectl create clusterrole crd-reader --verb=get,list,watch --resource=customresourcedefinitions.apiextensions.k8s.io
-          kubectl create clusterrolebinding sdk-user-crd-reader --clusterrole=crd-reader --user=sdk-user
-
-          # AppWrapper permissions for CodeFlare workloads
-          kubectl create clusterrole appwrapper-creator --verb=get,list,watch,create,update,patch,delete --resource=appwrappers.workload.codeflare.dev
-          kubectl create clusterrolebinding sdk-user-appwrapper-creator --clusterrole=appwrapper-creator --user=sdk-user
-
-          # Existing permissions
           kubectl create clusterrole list-ingresses --verb=get,list --resource=ingresses
           kubectl create clusterrolebinding sdk-user-list-ingresses --clusterrole=list-ingresses --user=sdk-user
           kubectl create clusterrole namespace-creator --verb=get,list,create,delete,patch --resource=namespaces
           kubectl create clusterrolebinding sdk-user-namespace-creator --clusterrole=namespace-creator --user=sdk-user
           kubectl create clusterrole raycluster-creator --verb=get,list,create,delete,patch,watch --resource=rayclusters.ray.io
           kubectl create clusterrolebinding sdk-user-raycluster-creator --clusterrole=raycluster-creator --user=sdk-user
-          kubectl create clusterrole rayjob-creator --verb=get,list,create,delete,patch,watch,update --resource=rayjobs.ray.io,rayjobs/status
+          kubectl create clusterrole rayjob-creator --verb=get,list,create,delete,patch,watch,update --resource=rayjobs.ray.io,rayjobs.ray.io/status
           kubectl create clusterrolebinding sdk-user-rayjob-creator --clusterrole=rayjob-creator --user=sdk-user
           kubectl create clusterrole resourceflavor-creator --verb=get,list,create,delete --resource=resourceflavors.kueue.x-k8s.io
           kubectl create clusterrolebinding sdk-user-resourceflavor-creator --clusterrole=resourceflavor-creator --user=sdk-user
           kubectl create clusterrole clusterqueue-creator --verb=get,list,create,delete,patch --resource=clusterqueues.kueue.x-k8s.io
           kubectl create clusterrolebinding sdk-user-clusterqueue-creator --clusterrole=clusterqueue-creator --user=sdk-user
           kubectl create clusterrole localqueue-creator --verb=get,list,create,delete,patch --resource=localqueues.kueue.x-k8s.io
           kubectl create clusterrolebinding sdk-user-localqueue-creator --clusterrole=localqueue-creator --user=sdk-user
-          kubectl create clusterrole workload-creator --verb=get,list,watch --resource=workloads.kueue.x-k8s.io
-          kubectl create clusterrolebinding sdk-user-workload-creator --clusterrole=workload-creator --user=sdk-user
+          kubectl create clusterrole workload-reader --verb=get,list,watch --resource=workloads.kueue.x-k8s.io
+          kubectl create clusterrolebinding sdk-user-workload-reader --clusterrole=workload-reader --user=sdk-user
           kubectl create clusterrole list-secrets --verb=get,list --resource=secrets
           kubectl create clusterrolebinding sdk-user-list-secrets --clusterrole=list-secrets --user=sdk-user
           kubectl create clusterrole pod-creator --verb=get,list,watch --resource=pods
@@ -174,74 +119,28 @@ jobs:
           kubectl create clusterrolebinding sdk-user-node-reader --clusterrole=node-reader --user=sdk-user
           kubectl config use-context sdk-user
 
-      - name: Verify cluster readiness before tests
-        run: |
-          echo "=== Pre-test cluster verification ==="
-          echo "Current context:"
-          kubectl config current-context
-
-          echo -e "\nNode status:"
-          kubectl get nodes -o wide
-
-          echo -e "\nSystem pods status:"
-          kubectl get pods -A | grep -E "(kube-system|ray-system|kueue-system|openshift-operators)" || true
-
-          echo -e "\nChecking for any pods in error state:"
-          kubectl get pods -A --field-selector=status.phase!=Running,status.phase!=Succeeded | grep -v "NAMESPACE" || echo "No pods in error state"
-
-          echo -e "\nKueue resources:"
-          kubectl get resourceflavors,clusterqueues,localqueues -A || true
-
-          echo -e "\nRay CRDs:"
-          kubectl get crd | grep ray || true
-
       - name: Run e2e tests
         run: |
-          export CODEFLARE_TEST_OUTPUT_DIR="${{ runner.temp }}/test-logs"
-          mkdir -p ${CODEFLARE_TEST_OUTPUT_DIR}
+          export CODEFLARE_TEST_OUTPUT_DIR=${{ env.TEMP_DIR }}
           echo "CODEFLARE_TEST_OUTPUT_DIR=${CODEFLARE_TEST_OUTPUT_DIR}" >> $GITHUB_ENV
 
           set -euo pipefail
           pip install poetry
           poetry install --with test,docs
           echo "Running e2e tests..."
-          poetry run pytest -v -s ./tests/e2e -m 'kind and nvidia_gpu' > ${CODEFLARE_TEST_OUTPUT_DIR}/e2e-pytest_output.log 2>&1
-        env:
-          GRPC_DNS_RESOLVER: "native"
-
-      - name: Run RayJob e2e tests
-        run: |
-          set -euo pipefail
-          echo "Running RayJob e2e tests..."
-          # Set environment variable to prevent default queue assignment for non-Kueue tests
-          export DISABLE_DEFAULT_KUEUE_QUEUE=true
-
-          # Run only the tests that are designed for Kueue integration
-          poetry run pytest -v -s ./tests/e2e/rayjob/ -x > ${CODEFLARE_TEST_OUTPUT_DIR}/rayjob_pytest_output.log 2>&1
+          poetry run pytest -v -s ./tests/e2e -m 'kind and nvidia_gpu' > ${CODEFLARE_TEST_OUTPUT_DIR}/pytest_output.log 2>&1
         env:
           GRPC_DNS_RESOLVER: "native"
 
       - name: Switch to kind-cluster context to print logs
         if: always() && steps.deploy.outcome == 'success'
         run: kubectl config use-context kind-cluster
 
-      - name: Print RayJob E2E Pytest output log
-        if: always() && steps.deploy.outcome == 'success'
-        run: |
-          echo "Printing RayJob Pytest output logs"
-          cat ${CODEFLARE_TEST_OUTPUT_DIR}/rayjob_pytest_output.log || echo "No RayJob test output found"
-
-      - name: Print E2E Pytest output log
-        if: always() && steps.deploy.outcome == 'success'
-        run: |
-          echo "Printing E2E Pytest output logs"
-          cat ${CODEFLARE_TEST_OUTPUT_DIR}/e2e-pytest_output.log || echo "No E2E test output found"
-
-      - name: Print CodeFlare operator logs
+      - name: Print Pytest output log
         if: always() && steps.deploy.outcome == 'success'
         run: |
-          echo "Printing CodeFlare operator logs"
-          kubectl logs -n openshift-operators --tail -1 -l app.kubernetes.io/name=codeflare-operator | tee ${CODEFLARE_TEST_OUTPUT_DIR}/codeflare-operator.log
+          echo "Printing Pytest output logs"
+          cat ${CODEFLARE_TEST_OUTPUT_DIR}/pytest_output.log
 
       - name: Print KubeRay operator logs
         if: always() && steps.deploy.outcome == 'success'
@@ -253,7 +152,7 @@ jobs:
         uses: ./common/github-actions/kind-export-logs
         if: always() && steps.deploy.outcome == 'success'
         with:
-          output-directory: ${{ env.CODEFLARE_TEST_OUTPUT_DIR }}
+          output-directory: ${CODEFLARE_TEST_OUTPUT_DIR}
 
       - name: Upload logs
         uses: actions/upload-artifact@v4
@@ -263,4 +162,3 @@ jobs:
           retention-days: 10
           path: |
             ${{ env.CODEFLARE_TEST_OUTPUT_DIR }}/**/*.log
-          if-no-files-found: warn
diff --git a/src/codeflare_sdk/ray/cluster/cluster.py b/src/codeflare_sdk/ray/cluster/cluster.py
@@ -395,6 +395,13 @@ def is_dashboard_ready(self) -> bool:
         if dashboard_uri is None:
             return False
 
+        # For ClusterIP services in local/test environments, we can't make HTTP requests
+        # Instead, check if the Ray pods are running as a proxy for dashboard readiness
+        if ".svc.cluster.local" in dashboard_uri:
+            # This is a ClusterIP service - check if Ray cluster is ready instead
+            status, ready = self.status(print_to_console=False)
+            return ready
+
         try:
             response = requests.get(
                 dashboard_uri,
diff --git a/tests/e2e/cluster_apply_kind_test.py b/tests/e2e/cluster_apply_kind_test.py
@@ -5,6 +5,7 @@
 from codeflare_sdk.common.utils import constants
 
 from support import (
+    get_ray_image,
     initialize_kubernetes_client,
     create_namespace,
     delete_namespace,
@@ -40,7 +41,7 @@ def test_cluster_apply(self):
             worker_cpu_limits="1",
             worker_memory_requests="1Gi",
             worker_memory_limits="2Gi",
-            image=f"rayproject/ray:{constants.RAY_VERSION}",
+            image=get_ray_image(),
             write_to_file=True,
             verify_tls=False,
         )
@@ -50,7 +51,7 @@ def test_cluster_apply(self):
         cluster.apply()
 
         # Wait for the cluster to be ready
-        cluster.wait_ready(dashboard_check=False)
+        cluster.wait_ready()
         status, ready = cluster.status()
         assert ready, f"Cluster {cluster_name} is not ready: {status}"
 
@@ -74,7 +75,7 @@ def test_cluster_apply(self):
             worker_cpu_limits="1",
             worker_memory_requests="1Gi",
             worker_memory_limits="2Gi",
-            image=f"rayproject/ray:{constants.RAY_VERSION}",
+            image=get_ray_image(),
             write_to_file=True,
             verify_tls=False,
         )
diff --git a/tests/e2e/heterogeneous_clusters_kind_test.py b/tests/e2e/heterogeneous_clusters_kind_test.py
@@ -48,16 +48,17 @@ def run_heterogeneous_clusters(
                     namespace=self.namespace,
                     num_workers=1,
                     head_cpu_requests="500m",
-                    head_cpu_limits="500m",
-                    head_memory_requests=2,
-                    head_memory_limits=2,
+                    head_cpu_limits="1",
+                    head_memory_requests="1Gi",
+                    head_memory_limits="2Gi",
                     worker_cpu_requests="500m",
                     worker_cpu_limits=1,
-                    worker_memory_requests=1,
-                    worker_memory_limits=4,
+                    worker_memory_requests="1Gi",
+                    worker_memory_limits="2Gi",
                     worker_extended_resource_requests={
                         gpu_resource_name: number_of_gpus
                     },
+                    image=get_ray_image(),
                     write_to_file=True,
                     verify_tls=False,
                     local_queue=queue_name,
diff --git a/tests/e2e/local_interactive_sdk_kind_test.py b/tests/e2e/local_interactive_sdk_kind_test.py
@@ -55,12 +55,15 @@ def run_local_interactives(
                 namespace=self.namespace,
                 num_workers=1,
                 head_cpu_requests="500m",
-                head_cpu_limits="500m",
+                head_cpu_limits="1",
+                head_memory_requests="1Gi",
+                head_memory_limits="2Gi",
                 worker_cpu_requests="500m",
                 worker_cpu_limits=1,
-                worker_memory_requests=1,
-                worker_memory_limits=4,
+                worker_memory_requests="1Gi",
+                worker_memory_limits="2Gi",
                 worker_extended_resource_requests={gpu_resource_name: number_of_gpus},
+                image=get_ray_image(),
                 verify_tls=False,
             )
         )
diff --git a/tests/e2e/mnist_raycluster_sdk_aw_kind_test.py b/tests/e2e/mnist_raycluster_sdk_aw_kind_test.py
@@ -43,12 +43,15 @@ def run_mnist_raycluster_sdk_kind(
                 namespace=self.namespace,
                 num_workers=1,
                 head_cpu_requests="500m",
-                head_cpu_limits="500m",
+                head_cpu_limits="1",
+                head_memory_requests="1Gi",
+                head_memory_limits="2Gi",
                 worker_cpu_requests="500m",
                 worker_cpu_limits=1,
-                worker_memory_requests=1,
-                worker_memory_limits=4,
+                worker_memory_requests="1Gi",
+                worker_memory_limits="2Gi",
                 worker_extended_resource_requests={gpu_resource_name: number_of_gpus},
+                image=get_ray_image(),
                 write_to_file=True,
                 verify_tls=False,
                 appwrapper=True,
diff --git a/tests/e2e/mnist_raycluster_sdk_kind_test.py b/tests/e2e/mnist_raycluster_sdk_kind_test.py
diff --git a/tests/e2e/support.py b/tests/e2e/support.py
