Skip to content

GitLab

Unverified Commit e5f7ee9f authored by Dillon Cullinan's avatar Dillon Cullinan Committed by GitHub
Browse files

ci: Disable deploy tests (#4397) 


Signed-off-by: default avatarDillon Cullinan <dcullinan@nvidia.com>
parent 54addb54
Hide whitespace changes
Inline Side-by-side
Showing with 322 additions and 322 deletions
.github/workflows/container-validation-backends.yml
View file @ e5f7ee9f
...@@ -517,325 +517,325 @@ jobs: ...@@ -517,325 +517,325 @@ jobs:
# Upload complete workflow metrics including container metrics # Upload complete workflow metrics including container metrics
python3 .github/workflows/upload_complete_workflow_metrics.py python3 .github/workflows/upload_complete_workflow_metrics.py
deploy-operator: # deploy-operator:
runs-on: cpu-amd-m5-2xlarge # runs-on: cpu-amd-m5-2xlarge
if: needs.changed-files.outputs.has_code_changes == 'true' # if: needs.changed-files.outputs.has_code_changes == 'true'
needs: [changed-files, operator, vllm, sglang, trtllm] # needs: [changed-files, operator, vllm, sglang, trtllm]
env: # env:
DYNAMO_INGRESS_SUFFIX: dev.aire.nvidia.com # DYNAMO_INGRESS_SUFFIX: dev.aire.nvidia.com
outputs: # outputs:
NAMESPACE: ${{ steps.deploy-operator-step.outputs.namespace }} # NAMESPACE: ${{ steps.deploy-operator-step.outputs.namespace }}
steps: # steps:
- name: Output Node Name # - name: Output Node Name
shell: bash # shell: bash
run: | # run: |
echo ${K8S_NODE_NAME} # echo ${K8S_NODE_NAME}
- uses: actions/checkout@v4 # - uses: actions/checkout@v4
- name: Deploy Operator # - name: Deploy Operator
id: deploy-operator-step # id: deploy-operator-step
env: # env:
BRANCH: ${{ github.ref_name }} # BRANCH: ${{ github.ref_name }}
run: | # run: |
set -x # set -x
# Set namespace # # Set namespace
# Invalid patterns: https://kubernetes.io/docs/concepts/overview/working-with-objects/names/ # # Invalid patterns: https://kubernetes.io/docs/concepts/overview/working-with-objects/names/
BRANCH_SANITIZED="${BRANCH//\//-}" # BRANCH_SANITIZED="${BRANCH//\//-}"
BRANCH_SANITIZED="${BRANCH_SANITIZED/pull-request/pr}" # BRANCH_SANITIZED="${BRANCH_SANITIZED/pull-request/pr}"
BRANCH_SANITIZED="${BRANCH_SANITIZED//./-}" # BRANCH_SANITIZED="${BRANCH_SANITIZED//./-}"
NAMESPACE="gh-id-${{ github.run_id }}-${BRANCH_SANITIZED}-dt" # NAMESPACE="gh-id-${{ github.run_id }}-${BRANCH_SANITIZED}-dt"
echo "namespace=${NAMESPACE}" >> "$GITHUB_OUTPUT" # echo "namespace=${NAMESPACE}" >> "$GITHUB_OUTPUT"
# Setup kubeconfig # # Setup kubeconfig
echo "${{ secrets.AZURE_AKS_CI_KUBECONFIG_B64 }}" | base64 -d > .kubeconfig # echo "${{ secrets.AZURE_AKS_CI_KUBECONFIG_B64 }}" | base64 -d > .kubeconfig
chmod 600 .kubeconfig # chmod 600 .kubeconfig
export KUBECONFIG=$(pwd)/.kubeconfig # export KUBECONFIG=$(pwd)/.kubeconfig
kubectl config set-context --current --namespace=$NAMESPACE --kubeconfig "${KUBECONFIG}" # kubectl config set-context --current --namespace=$NAMESPACE --kubeconfig "${KUBECONFIG}"
kubectl config current-context # kubectl config current-context
# Create a namespace for this job # # Create a namespace for this job
echo "Creating an ephemeral namespace..." # echo "Creating an ephemeral namespace..."
kubectl create namespace $NAMESPACE # kubectl create namespace $NAMESPACE
echo "Attaching the labels for secrets and cleanup" # echo "Attaching the labels for secrets and cleanup"
kubectl label namespaces ${NAMESPACE} nscleanup/enabled=true nscleanup/ttl=7200 gitlab-imagepull=enabled ngc-api=enabled nvcr-imagepull=enabled --overwrite=true # kubectl label namespaces ${NAMESPACE} nscleanup/enabled=true nscleanup/ttl=7200 gitlab-imagepull=enabled ngc-api=enabled nvcr-imagepull=enabled --overwrite=true
# Set the namespace as default # # Set the namespace as default
kubectl config set-context --current --namespace=$NAMESPACE # kubectl config set-context --current --namespace=$NAMESPACE
# Check if Istio is installed # # Check if Istio is installed
kubectl get pods -n istio-system # kubectl get pods -n istio-system
# Check if default storage class exists # # Check if default storage class exists
kubectl get storageclass # kubectl get storageclass
# Install Helm chart # # Install Helm chart
export VIRTUAL_ENV=/opt/dynamo/venv # export VIRTUAL_ENV=/opt/dynamo/venv
export KUBE_NS=$NAMESPACE # export KUBE_NS=$NAMESPACE
export ISTIO_ENABLED=true # export ISTIO_ENABLED=true
export ISTIO_GATEWAY=istio-system/ingress-alb # export ISTIO_GATEWAY=istio-system/ingress-alb
export VIRTUAL_SERVICE_SUPPORTS_HTTPS=true # export VIRTUAL_SERVICE_SUPPORTS_HTTPS=true
export DYNAMO_CLOUD=https://${NAMESPACE}.${DYNAMO_INGRESS_SUFFIX} # export DYNAMO_CLOUD=https://${NAMESPACE}.${DYNAMO_INGRESS_SUFFIX}
# Install dynamo env secrets # # Install dynamo env secrets
kubectl create secret generic hf-token-secret --from-literal=HF_TOKEN=${{ secrets.HF_TOKEN }} -n $KUBE_NS || true # kubectl create secret generic hf-token-secret --from-literal=HF_TOKEN=${{ secrets.HF_TOKEN }} -n $KUBE_NS || true
# Create docker pull secret for operator image # # Create docker pull secret for operator image
kubectl create secret docker-registry docker-imagepullsecret --docker-server=${{ secrets.AZURE_ACR_HOSTNAME }} --docker-username=${{ secrets.AZURE_ACR_USER }} --docker-password=${{ secrets.AZURE_ACR_PASSWORD }} --namespace=${NAMESPACE} # kubectl create secret docker-registry docker-imagepullsecret --docker-server=${{ secrets.AZURE_ACR_HOSTNAME }} --docker-username=${{ secrets.AZURE_ACR_USER }} --docker-password=${{ secrets.AZURE_ACR_PASSWORD }} --namespace=${NAMESPACE}
# Install helm dependencies # # Install helm dependencies
helm repo add bitnami https://charts.bitnami.com/bitnami # helm repo add bitnami https://charts.bitnami.com/bitnami
cd deploy/cloud/helm/platform/ # cd deploy/cloud/helm/platform/
helm dep build . # helm dep build .
# Install platform with namespace restriction for single profile testing # # Install platform with namespace restriction for single profile testing
helm upgrade --install dynamo-platform . --namespace ${NAMESPACE} \ # helm upgrade --install dynamo-platform . --namespace ${NAMESPACE} \
--set dynamo-operator.namespaceRestriction.enabled=true \ # --set dynamo-operator.namespaceRestriction.enabled=true \
--set dynamo-operator.namespaceRestriction.allowedNamespaces[0]=${NAMESPACE} \ # --set dynamo-operator.namespaceRestriction.allowedNamespaces[0]=${NAMESPACE} \
--set dynamo-operator.controllerManager.manager.image.repository=${{ secrets.AZURE_ACR_HOSTNAME }}/ai-dynamo/dynamo \ # --set dynamo-operator.controllerManager.manager.image.repository=${{ secrets.AZURE_ACR_HOSTNAME }}/ai-dynamo/dynamo \
--set dynamo-operator.controllerManager.manager.image.tag=${{ github.sha }}-operator-amd64 \ # --set dynamo-operator.controllerManager.manager.image.tag=${{ github.sha }}-operator-amd64 \
--set dynamo-operator.imagePullSecrets[0].name=docker-imagepullsecret # --set dynamo-operator.imagePullSecrets[0].name=docker-imagepullsecret
# Wait for all deployments to be ready # # Wait for all deployments to be ready
timeout 300s kubectl rollout status deployment -n $NAMESPACE --watch # timeout 300s kubectl rollout status deployment -n $NAMESPACE --watch
deploy-test-vllm: # deploy-test-vllm:
runs-on: cpu-amd-m5-2xlarge # runs-on: cpu-amd-m5-2xlarge
if: needs.changed-files.outputs.has_code_changes == 'true' # if: needs.changed-files.outputs.has_code_changes == 'true'
needs: [changed-files, deploy-operator, vllm] # needs: [changed-files, deploy-operator, vllm]
permissions: # permissions:
contents: read # contents: read
strategy: # strategy:
fail-fast: false # fail-fast: false
max-parallel: 1 # max-parallel: 1
matrix: # matrix:
profile: # profile:
- agg # - agg
- agg_router # - agg_router
- disagg # - disagg
- disagg_router # - disagg_router
name: deploy-test-vllm (${{ matrix.profile }}) # name: deploy-test-vllm (${{ matrix.profile }})
env: # env:
FRAMEWORK: vllm # FRAMEWORK: vllm
DYNAMO_INGRESS_SUFFIX: dev.aire.nvidia.com # DYNAMO_INGRESS_SUFFIX: dev.aire.nvidia.com
DEPLOYMENT_FILE: "deploy/${{ matrix.profile }}.yaml" # DEPLOYMENT_FILE: "deploy/${{ matrix.profile }}.yaml"
MODEL_NAME: "Qwen/Qwen3-0.6B" # MODEL_NAME: "Qwen/Qwen3-0.6B"
steps: &deploy-test-steps # steps: &deploy-test-steps
- name: Output Node Name # - name: Output Node Name
shell: bash # shell: bash
run: | # run: |
echo ${K8S_NODE_NAME} # echo ${K8S_NODE_NAME}
- uses: actions/checkout@v4 # - uses: actions/checkout@v4
- name: Setup Kubeconfig # - name: Setup Kubeconfig
env: # env:
NAMESPACE: ${{ needs.deploy-operator.outputs.NAMESPACE }} # NAMESPACE: ${{ needs.deploy-operator.outputs.NAMESPACE }}
run: | # run: |
set -x # set -x
# Setup kubeconfig # # Setup kubeconfig
echo "${{ secrets.AZURE_AKS_CI_KUBECONFIG_B64 }}" | base64 -d > .kubeconfig # echo "${{ secrets.AZURE_AKS_CI_KUBECONFIG_B64 }}" | base64 -d > .kubeconfig
chmod 600 .kubeconfig # chmod 600 .kubeconfig
export KUBECONFIG=$(pwd)/.kubeconfig # export KUBECONFIG=$(pwd)/.kubeconfig
kubectl config set-context --current --namespace=$NAMESPACE --kubeconfig "${KUBECONFIG}" # kubectl config set-context --current --namespace=$NAMESPACE --kubeconfig "${KUBECONFIG}"
kubectl config get-contexts # kubectl config get-contexts
- name: Run Tests # - name: Run Tests
env: # env:
NAMESPACE: ${{ needs.deploy-operator.outputs.NAMESPACE }} # NAMESPACE: ${{ needs.deploy-operator.outputs.NAMESPACE }}
run: | # run: |
set -x # set -x
export KUBECONFIG=$(pwd)/.kubeconfig # export KUBECONFIG=$(pwd)/.kubeconfig
kubectl config set-context --current --namespace=$NAMESPACE # kubectl config set-context --current --namespace=$NAMESPACE
cd examples/backends/$FRAMEWORK # cd examples/backends/$FRAMEWORK
export FRAMEWORK_RUNTIME_IMAGE="${{ secrets.AZURE_ACR_HOSTNAME }}/ai-dynamo/dynamo:${{ github.sha }}-${FRAMEWORK}-amd64" # export FRAMEWORK_RUNTIME_IMAGE="${{ secrets.AZURE_ACR_HOSTNAME }}/ai-dynamo/dynamo:${{ github.sha }}-${FRAMEWORK}-amd64"
export KUBE_NS=$NAMESPACE # export KUBE_NS=$NAMESPACE
export GRAPH_NAME=$(yq e '.metadata.name' $DEPLOYMENT_FILE) # export GRAPH_NAME=$(yq e '.metadata.name' $DEPLOYMENT_FILE)
echo "GRAPH_NAME=${GRAPH_NAME}" >> $GITHUB_ENV # echo "GRAPH_NAME=${GRAPH_NAME}" >> $GITHUB_ENV
# Update the deployment file in-place # # Update the deployment file in-place
yq -i '.spec.services.[].extraPodSpec.mainContainer.image = env(FRAMEWORK_RUNTIME_IMAGE)' $DEPLOYMENT_FILE # yq -i '.spec.services.[].extraPodSpec.mainContainer.image = env(FRAMEWORK_RUNTIME_IMAGE)' $DEPLOYMENT_FILE
# Debug: Show updated deployment file # # Debug: Show updated deployment file
echo "=== UPDATED DEPLOYMENT FILE ===" # echo "=== UPDATED DEPLOYMENT FILE ==="
cat $DEPLOYMENT_FILE # cat $DEPLOYMENT_FILE
# Apply the updated file # # Apply the updated file
kubectl apply -n $KUBE_NS -f $DEPLOYMENT_FILE # kubectl apply -n $KUBE_NS -f $DEPLOYMENT_FILE
# --- Wait for all pods in the dynamo graph deployment to be ready --- # # --- Wait for all pods in the dynamo graph deployment to be ready ---
sleep 20 # sleep 20
# Get the deployment name from the file # # Get the deployment name from the file
export GRAPH_NAME=$(yq e '.metadata.name' $DEPLOYMENT_FILE) # export GRAPH_NAME=$(yq e '.metadata.name' $DEPLOYMENT_FILE)
echo "Waiting for all pods with label nvidia.com/dynamo-graph-deployment-name: $GRAPH_NAME" # echo "Waiting for all pods with label nvidia.com/dynamo-graph-deployment-name: $GRAPH_NAME"
# Wait for all pods with the deployment label to be ready # # Wait for all pods with the deployment label to be ready
kubectl wait --for=condition=ready pod -l "nvidia.com/dynamo-graph-deployment-name=$GRAPH_NAME" -n ${KUBE_NS} --timeout=1300s # kubectl wait --for=condition=ready pod -l "nvidia.com/dynamo-graph-deployment-name=$GRAPH_NAME" -n ${KUBE_NS} --timeout=1300s
# Debug: Show final pod statuses for the deployment # # Debug: Show final pod statuses for the deployment
echo "=== FINAL POD STATUSES ===" # echo "=== FINAL POD STATUSES ==="
kubectl get pods -l "nvidia.com/dynamo-graph-deployment-name=$GRAPH_NAME" -n $KUBE_NS -o wide # kubectl get pods -l "nvidia.com/dynamo-graph-deployment-name=$GRAPH_NAME" -n $KUBE_NS -o wide
echo "" # echo ""
kubectl get all -n $KUBE_NS # kubectl get all -n $KUBE_NS
export FRONTEND_POD=$(kubectl get pods -n ${KUBE_NS} -l nvidia.com/dynamo-component-type=frontend,nvidia.com/dynamo-graph-deployment-name=${GRAPH_NAME} | tail -n1 | awk '{print $1}') # export FRONTEND_POD=$(kubectl get pods -n ${KUBE_NS} -l nvidia.com/dynamo-component-type=frontend,nvidia.com/dynamo-graph-deployment-name=${GRAPH_NAME} | tail -n1 | awk '{print $1}')
export CONTAINER_PORT=$(kubectl get pod $FRONTEND_POD -n ${KUBE_NS} -o jsonpath='{.spec.containers[0].ports[?(@.name=="http")].containerPort}') # export CONTAINER_PORT=$(kubectl get pod $FRONTEND_POD -n ${KUBE_NS} -o jsonpath='{.spec.containers[0].ports[?(@.name=="http")].containerPort}')
echo "Container port is ${CONTAINER_PORT}" # echo "Container port is ${CONTAINER_PORT}"
kubectl port-forward pod/$FRONTEND_POD 8000:${CONTAINER_PORT} -n ${KUBE_NS} & # kubectl port-forward pod/$FRONTEND_POD 8000:${CONTAINER_PORT} -n ${KUBE_NS} &
export LLM_URL="http://localhost:8000" # export LLM_URL="http://localhost:8000"
sleep 10 # Give port-forward time to establish the connection # sleep 10 # Give port-forward time to establish the connection
echo "LLM URL: ${LLM_URL}" # echo "LLM URL: ${LLM_URL}"
echo "MODEL NAME: ${MODEL_NAME}" # echo "MODEL NAME: ${MODEL_NAME}"
# Wait until the model is available in the /v1/models response # # Wait until the model is available in the /v1/models response
MAX_ATTEMPTS=30 # MAX_ATTEMPTS=30
ATTEMPT=1 # ATTEMPT=1
while [ $ATTEMPT -le $MAX_ATTEMPTS ]; do # while [ $ATTEMPT -le $MAX_ATTEMPTS ]; do
MODELS_RESPONSE=$(curl -s --retry 5 --retry-delay 2 --retry-connrefused "${LLM_URL}/v1/models") # MODELS_RESPONSE=$(curl -s --retry 5 --retry-delay 2 --retry-connrefused "${LLM_URL}/v1/models")
if echo "$MODELS_RESPONSE" | jq -e --arg MODEL_NAME "$MODEL_NAME" '.data[]?.id == $MODEL_NAME' >/dev/null 2>&1; then # if echo "$MODELS_RESPONSE" | jq -e --arg MODEL_NAME "$MODEL_NAME" '.data[]?.id == $MODEL_NAME' >/dev/null 2>&1; then
echo "Model $MODEL_NAME is available in /v1/models" # echo "Model $MODEL_NAME is available in /v1/models"
break # break
fi # fi
echo "Waiting for model $MODEL_NAME to be available in /v1/models... (attempt $ATTEMPT/$MAX_ATTEMPTS)" # echo "Waiting for model $MODEL_NAME to be available in /v1/models... (attempt $ATTEMPT/$MAX_ATTEMPTS)"
sleep 5 # sleep 5
ATTEMPT=$((ATTEMPT + 1)) # ATTEMPT=$((ATTEMPT + 1))
done # done
if [ $ATTEMPT -gt $MAX_ATTEMPTS ]; then # if [ $ATTEMPT -gt $MAX_ATTEMPTS ]; then
echo "Model $MODEL_NAME not found in /v1/models after $MAX_ATTEMPTS attempts" # echo "Model $MODEL_NAME not found in /v1/models after $MAX_ATTEMPTS attempts"
echo "Last response: $MODELS_RESPONSE" # echo "Last response: $MODELS_RESPONSE"
exit 1 # exit 1
fi # fi
RESPONSE=$(curl -s -N --no-buffer --retry 10 --retry-delay 5 --retry-connrefused -X POST "${LLM_URL}/v1/chat/completions" \ # RESPONSE=$(curl -s -N --no-buffer --retry 10 --retry-delay 5 --retry-connrefused -X POST "${LLM_URL}/v1/chat/completions" \
-H 'accept: text/event-stream' \ # -H 'accept: text/event-stream' \
-H 'Content-Type: application/json' \ # -H 'Content-Type: application/json' \
-d '{ # -d '{
"model": "'"${MODEL_NAME:-Qwen/Qwen3-0.6B}"'", # "model": "'"${MODEL_NAME:-Qwen/Qwen3-0.6B}"'",
"messages": [ # "messages": [
{ # {
"role": "user", # "role": "user",
"content": "In the heart of Eldoria, an ancient land of boundless magic and mysterious creatures, lies the long-forgotten city of Aeloria. Once a beacon of knowledge and power, Aeloria was buried beneath the shifting sands of time, lost to the world for centuries. You are an intrepid explorer, known for your unparalleled curiosity and courage, who has stumbled upon an ancient map hinting at ests that Aeloria holds a secret so profound that it has the potential to reshape the very fabric of reality. Your journey will take you through treacherous deserts, enchanted forests, and across perilous mountain ranges. Your Task: Character Background: Develop a detailed background for your character. Describe their motivations for seeking out Aeloria, their skills and weaknesses, and any personal connections to the ancient city or its legends. Are they driven by a quest for knowledge, a search for lost familt clue is hidden." # "content": "In the heart of Eldoria, an ancient land of boundless magic and mysterious creatures, lies the long-forgotten city of Aeloria. Once a beacon of knowledge and power, Aeloria was buried beneath the shifting sands of time, lost to the world for centuries. You are an intrepid explorer, known for your unparalleled curiosity and courage, who has stumbled upon an ancient map hinting at ests that Aeloria holds a secret so profound that it has the potential to reshape the very fabric of reality. Your journey will take you through treacherous deserts, enchanted forests, and across perilous mountain ranges. Your Task: Character Background: Develop a detailed background for your character. Describe their motivations for seeking out Aeloria, their skills and weaknesses, and any personal connections to the ancient city or its legends. Are they driven by a quest for knowledge, a search for lost familt clue is hidden."
} # }
], # ],
"stream":false, # "stream":false,
"max_tokens": 30, # "max_tokens": 30,
"temperature": 0.0 # "temperature": 0.0
}' 2>&1) # }' 2>&1)
echo "Response: $RESPONSE" # echo "Response: $RESPONSE"
TEST_RESULT=0 # TEST_RESULT=0
if ! echo "$RESPONSE" | jq -e . >/dev/null 2>&1; then # if ! echo "$RESPONSE" | jq -e . >/dev/null 2>&1; then
echo "Test failed: Response is not valid JSON" # echo "Test failed: Response is not valid JSON"
echo "Got: $RESPONSE" # echo "Got: $RESPONSE"
TEST_RESULT=1 # TEST_RESULT=1
elif ! echo "$RESPONSE" | jq -e '.choices[0].message.role == "assistant"' >/dev/null 2>&1; then # elif ! echo "$RESPONSE" | jq -e '.choices[0].message.role == "assistant"' >/dev/null 2>&1; then
echo "Test failed: Message role is not 'assistant'" # echo "Test failed: Message role is not 'assistant'"
echo "Got: $(echo "$RESPONSE" | jq '.choices[0].message.role')" # echo "Got: $(echo "$RESPONSE" | jq '.choices[0].message.role')"
TEST_RESULT=1 # TEST_RESULT=1
elif ! echo "$RESPONSE" | jq -e '.model == "'"${MODEL_NAME}"'"' >/dev/null 2>&1; then # elif ! echo "$RESPONSE" | jq -e '.model == "'"${MODEL_NAME}"'"' >/dev/null 2>&1; then
echo "Test failed: Model name is incorrect" # echo "Test failed: Model name is incorrect"
echo "Got: $(echo "$RESPONSE" | jq '.model')" # echo "Got: $(echo "$RESPONSE" | jq '.model')"
TEST_RESULT=1 # TEST_RESULT=1
elif ! echo "$RESPONSE" | jq -e '.choices[0].message.content | length > 100' >/dev/null 2>&1; then # elif ! echo "$RESPONSE" | jq -e '.choices[0].message.content | length > 100' >/dev/null 2>&1; then
echo "Test failed: Response content length is not greater than 100 characters" # echo "Test failed: Response content length is not greater than 100 characters"
echo "Got length: $(echo "$RESPONSE" | jq '.choices[0].message.content | length')" # echo "Got length: $(echo "$RESPONSE" | jq '.choices[0].message.content | length')"
TEST_RESULT=1 # TEST_RESULT=1
else # else
echo "Test passed: Response matches expected format and content" # echo "Test passed: Response matches expected format and content"
fi # fi
exit $TEST_RESULT # exit $TEST_RESULT
- name: Cleanup # - name: Cleanup
if: always() # if: always()
timeout-minutes: 5 # timeout-minutes: 5
env: # env:
NAMESPACE: ${{ needs.deploy-operator.outputs.NAMESPACE }} # NAMESPACE: ${{ needs.deploy-operator.outputs.NAMESPACE }}
run: | # run: |
set -x # set -x
export KUBECONFIG=$(pwd)/.kubeconfig # export KUBECONFIG=$(pwd)/.kubeconfig
kubectl config set-context --current --namespace=$NAMESPACE # kubectl config set-context --current --namespace=$NAMESPACE
# For debugging purposes, list all the resources before we delete # # For debugging purposes, list all the resources before we delete
kubectl get all # kubectl get all
echo "Deleting DynamoGraphDeployments for this job in namespace $NAMESPACE..." # echo "Deleting DynamoGraphDeployments for this job in namespace $NAMESPACE..."
kubectl delete dynamographdeployments ${GRAPH_NAME} -n $NAMESPACE || true # kubectl delete dynamographdeployments ${GRAPH_NAME} -n $NAMESPACE || true
deploy-test-sglang: # deploy-test-sglang:
runs-on: cpu-amd-m5-2xlarge # runs-on: cpu-amd-m5-2xlarge
if: needs.changed-files.outputs.has_code_changes == 'true' # if: needs.changed-files.outputs.has_code_changes == 'true'
needs: [changed-files, deploy-operator, sglang] # needs: [changed-files, deploy-operator, sglang]
permissions: # permissions:
contents: read # contents: read
strategy: # strategy:
fail-fast: false # fail-fast: false
max-parallel: 1 # max-parallel: 1
matrix: # matrix:
profile: # profile:
- agg # - agg
- agg_router # - agg_router
name: deploy-test-sglang (${{ matrix.profile }}) # name: deploy-test-sglang (${{ matrix.profile }})
env: # env:
FRAMEWORK: sglang # FRAMEWORK: sglang
DYNAMO_INGRESS_SUFFIX: dev.aire.nvidia.com # DYNAMO_INGRESS_SUFFIX: dev.aire.nvidia.com
DEPLOYMENT_FILE: "deploy/${{ matrix.profile }}.yaml" # DEPLOYMENT_FILE: "deploy/${{ matrix.profile }}.yaml"
MODEL_NAME: "Qwen/Qwen3-0.6B" # MODEL_NAME: "Qwen/Qwen3-0.6B"
steps: *deploy-test-steps # steps: *deploy-test-steps
deploy-test-trtllm: # deploy-test-trtllm:
runs-on: cpu-amd-m5-2xlarge # runs-on: cpu-amd-m5-2xlarge
if: needs.changed-files.outputs.has_code_changes == 'true' # if: needs.changed-files.outputs.has_code_changes == 'true'
needs: [changed-files, deploy-operator, trtllm] # needs: [changed-files, deploy-operator, trtllm]
permissions: # permissions:
contents: read # contents: read
strategy: # strategy:
fail-fast: false # fail-fast: false
max-parallel: 1 # max-parallel: 1
matrix: # matrix:
profile: # profile:
- agg # - agg
- agg_router # - agg_router
- disagg # - disagg
- disagg_router # - disagg_router
name: deploy-test-trtllm (${{ matrix.profile }}) # name: deploy-test-trtllm (${{ matrix.profile }})
env: # env:
FRAMEWORK: trtllm # FRAMEWORK: trtllm
DYNAMO_INGRESS_SUFFIX: dev.aire.nvidia.com # DYNAMO_INGRESS_SUFFIX: dev.aire.nvidia.com
DEPLOYMENT_FILE: "deploy/${{ matrix.profile }}.yaml" # DEPLOYMENT_FILE: "deploy/${{ matrix.profile }}.yaml"
MODEL_NAME: "Qwen/Qwen3-0.6B" # MODEL_NAME: "Qwen/Qwen3-0.6B"
steps: *deploy-test-steps # steps: *deploy-test-steps
cleanup: # cleanup:
runs-on: cpu-amd-m5-2xlarge # runs-on: cpu-amd-m5-2xlarge
if: always() # if: always()
needs: [changed-files, deploy-operator, deploy-test-trtllm, deploy-test-sglang, deploy-test-vllm] # needs: [changed-files, deploy-operator, deploy-test-trtllm, deploy-test-sglang, deploy-test-vllm]
steps: # steps:
- name: Output Node Name # - name: Output Node Name
shell: bash # shell: bash
run: | # run: |
echo ${K8S_NODE_NAME} # echo ${K8S_NODE_NAME}
- uses: actions/checkout@v4 # - uses: actions/checkout@v4
- name: Setup Kubeconfig # - name: Setup Kubeconfig
env: # env:
NAMESPACE: ${{ needs.deploy-operator.outputs.NAMESPACE }} # NAMESPACE: ${{ needs.deploy-operator.outputs.NAMESPACE }}
run: | # run: |
set -x # set -x
# Setup kubeconfig # # Setup kubeconfig
echo "${{ secrets.AZURE_AKS_CI_KUBECONFIG_B64 }}" | base64 -d > .kubeconfig # echo "${{ secrets.AZURE_AKS_CI_KUBECONFIG_B64 }}" | base64 -d > .kubeconfig
chmod 600 .kubeconfig # chmod 600 .kubeconfig
export KUBECONFIG=$(pwd)/.kubeconfig # export KUBECONFIG=$(pwd)/.kubeconfig
kubectl config set-context --current --namespace=$NAMESPACE --kubeconfig "${KUBECONFIG}" # kubectl config set-context --current --namespace=$NAMESPACE --kubeconfig "${KUBECONFIG}"
kubectl config current-context # kubectl config current-context
- name: Cleanup # - name: Cleanup
timeout-minutes: 5 # timeout-minutes: 5
env: # env:
NAMESPACE: ${{ needs.deploy-operator.outputs.NAMESPACE }} # NAMESPACE: ${{ needs.deploy-operator.outputs.NAMESPACE }}
run: | # run: |
set -x # set -x
export KUBECONFIG=$(pwd)/.kubeconfig # export KUBECONFIG=$(pwd)/.kubeconfig
kubectl config set-context --current --namespace=$NAMESPACE # kubectl config set-context --current --namespace=$NAMESPACE
echo "${{ secrets.AZURE_AKS_CI_KUBECONFIG_B64 }}" | base64 -d > .kubeconfig # echo "${{ secrets.AZURE_AKS_CI_KUBECONFIG_B64 }}" | base64 -d > .kubeconfig
chmod 600 .kubeconfig # chmod 600 .kubeconfig
export KUBECONFIG=$(pwd)/.kubeconfig # export KUBECONFIG=$(pwd)/.kubeconfig
kubectl config set-context --current --namespace=$NAMESPACE --kubeconfig "${KUBECONFIG}" # kubectl config set-context --current --namespace=$NAMESPACE --kubeconfig "${KUBECONFIG}"
# For debugging purposes, list all the resources before we uninstall # # For debugging purposes, list all the resources before we uninstall
kubectl get all # kubectl get all
echo "Deleting all DynamoGraphDeployments in namespace $NAMESPACE..." # echo "Deleting all DynamoGraphDeployments in namespace $NAMESPACE..."
kubectl delete dynamographdeployments --all -n $NAMESPACE || true # kubectl delete dynamographdeployments --all -n $NAMESPACE || true
# Uninstall the helm chart # # Uninstall the helm chart
helm ls # helm ls
helm uninstall dynamo-platform --namespace $NAMESPACE || true # helm uninstall dynamo-platform --namespace $NAMESPACE || true
echo "Namespace $NAMESPACE deletion initiated, proceeding with cleanup..." # echo "Namespace $NAMESPACE deletion initiated, proceeding with cleanup..."
kubectl delete namespace $NAMESPACE || true # kubectl delete namespace $NAMESPACE || true
echo "Namespace $NAMESPACE completed." # echo "Namespace $NAMESPACE completed."
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment