Skip to content

Run PR check for guided notebooks #32

Run PR check for guided notebooks

Run PR check for guided notebooks #32

name: Guided notebooks tests
on:
pull_request:
branches:
- main
- 'release-*'
paths-ignore:
- 'docs/**'
- '**.adoc'
- '**.md'
- 'LICENSE'
push:
branches:
- main
- 'release-*'
paths-ignore:
- '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"
permissions:
id-token: write # This is required for requesting the JWT
contents: read
jobs:
kubernetes:
runs-on: ubuntu-20.04-4core-gpu
steps:
- name: Checkout code
uses: actions/checkout@v4
with:
submodules: recursive
- 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: ${{ secrets.GITHUB_TOKEN }}
- name: Set up specific Python version
uses: actions/setup-python@v5
with:
python-version: '3.9'
cache: 'pip' # caching pip dependencies
- name: Setup NVidia GPU environment for KinD
uses: ./common/github-actions/nvidia-gpu-setup
- name: Setup and start KinD cluster
uses: ./common/github-actions/kind
- name: Install NVidia GPU operator for KinD
uses: ./common/github-actions/nvidia-gpu-operator
with:
enable-time-slicing: 'true'
- name: Deploy CodeFlare stack
id: deploy
run: |
cd codeflare-operator
echo Setting up CodeFlare stack
make setup-e2e
echo Deploying CodeFlare operator
make deploy -e IMG="${CODEFLARE_OPERATOR_IMG}" -e ENV="e2e"
kubectl wait --timeout=120s --for=condition=Available=true deployment -n openshift-operators codeflare-operator-manager
cd ..
- name: Install MINIO
run: |
kubectl apply -f ./tests/e2e/minio_deployment.yaml
kubectl wait --timeout=120s --for=condition=Available=true deployment -n default minio
- 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: |
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 --resource=rayclusters
kubectl create clusterrolebinding sdk-user-raycluster-creator --clusterrole=raycluster-creator --user=sdk-user
kubectl create clusterrole appwrapper-creator --verb=get,list,create,delete,patch --resource=appwrappers
kubectl create clusterrolebinding sdk-user-appwrapper-creator --clusterrole=appwrapper-creator --user=sdk-user
kubectl create clusterrole resourceflavor-creator --verb=get,list,create,delete --resource=resourceflavors
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
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
kubectl create clusterrolebinding sdk-user-localqueue-creator --clusterrole=localqueue-creator --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 config use-context sdk-user
- name: Setup Guided notebooks execution
run: |
echo "Installing papermill and dependencies..."
pip install poetry papermill ipython ipykernel
# Disable virtualenv due to problems using packaged in virtualenv in papermill
poetry config virtualenvs.create false
echo "Installing SDK..."
poetry install --with test,docs
- name: Run 0_basic_ray.ipynb
run: |
set -euo pipefail
# Remove login/logout cells, as KinD doesn't support authentication using token
jq -r 'del(.cells[] | select(.source[] | contains("Create authentication object for user permissions")))' 0_basic_ray.ipynb > 0_basic_ray.ipynb.tmp && mv 0_basic_ray.ipynb.tmp 0_basic_ray.ipynb
jq -r 'del(.cells[] | select(.source[] | contains("auth.logout()")))' 0_basic_ray.ipynb > 0_basic_ray.ipynb.tmp && mv 0_basic_ray.ipynb.tmp 0_basic_ray.ipynb
# Run notebook
# poetry run papermill 0_basic_ray.ipynb 0_basic_ray_out.ipynb --log-output --execution-timeout 600
working-directory: demo-notebooks/guided-demos
- name: Run 1_cluster_job_client.ipynb
run: |
set -euo pipefail
# Remove login/logout cells, as KinD doesn't support authentication using token
jq -r 'del(.cells[] | select(.source[] | contains("Create authentication object for user permissions")))' 1_cluster_job_client.ipynb > 1_cluster_job_client.ipynb.tmp && mv 1_cluster_job_client.ipynb.tmp 1_cluster_job_client.ipynb
jq -r 'del(.cells[] | select(.source[] | contains("auth.logout()")))' 1_cluster_job_client.ipynb > 1_cluster_job_client.ipynb.tmp && mv 1_cluster_job_client.ipynb.tmp 1_cluster_job_client.ipynb
# Replace async logs with waiting for job to finish, async logs don't work properly in papermill
JOB_WAIT=$(jq -r '.' ${GITHUB_WORKSPACE}/.github/resources/wait_for_job_cell.json)
jq --argjson job_wait "$JOB_WAIT" -r '(.cells[] | select(.source[] | contains("async for lines in client.tail_job_logs"))) |= $job_wait' 1_cluster_job_client.ipynb > 1_cluster_job_client.ipynb.tmp && mv 1_cluster_job_client.ipynb.tmp 1_cluster_job_client.ipynb
# Run notebook
# poetry run papermill 1_cluster_job_client.ipynb 1_cluster_job_client_out.ipynb --log-output --execution-timeout 1200
working-directory: demo-notebooks/guided-demos
- name: Run 2_basic_interactive.ipynb
run: |
set -euo pipefail
# Remove login/logout cells, as KinD doesn't support authentication using token
jq -r 'del(.cells[] | select(.source[] | contains("Create authentication object for user permissions")))' 2_basic_interactive.ipynb > 2_basic_interactive.ipynb.tmp && mv 2_basic_interactive.ipynb.tmp 2_basic_interactive.ipynb
jq -r 'del(.cells[] | select(.source[] | contains("auth.logout()")))' 2_basic_interactive.ipynb > 2_basic_interactive.ipynb.tmp && mv 2_basic_interactive.ipynb.tmp 2_basic_interactive.ipynb
# Rewrite cluster_uri() to local_client_url() to retrieve client URL available out of cluster, as the test is executed outside of cluster
sed -i "s/cluster_uri()/local_client_url()/" 2_basic_interactive.ipynb
# Set explicit namespace as SDK need it (currently) to resolve local queues
sed -i "s/head_cpus=1,/head_cpus=1, namespace='default',/" 2_basic_interactive.ipynb
# Add MINIO related modules to runtime environment
sed -i "s/transformers/s3fs\\\\\", \\\\\"pyarrow\\\\\", \\\\\"transformers/" 2_basic_interactive.ipynb
# Replace markdown cell with remote configuration for MINIO
MINIO_CONFIG=$(jq -r '.' ${GITHUB_WORKSPACE}/.github/resources/minio_remote_config_cell.json)
jq --argjson minio_config "$MINIO_CONFIG" -r '(.cells[] | select(.source[] | contains("Now that we are connected"))) |= $minio_config' 2_basic_interactive.ipynb > 2_basic_interactive.ipynb.tmp && mv 2_basic_interactive.ipynb.tmp 2_basic_interactive.ipynb
# Configure persistent storage for Ray trainer
sed -i -E "s/# run_config.*\)/, run_config=ray.get(get_minio_run_config.remote())/" 2_basic_interactive.ipynb
# Run notebook
poetry run papermill 2_basic_interactive.ipynb 2_basic_interactive_out.ipynb --log-output --execution-timeout 1200
env:
GRPC_DNS_RESOLVER: "native"
working-directory: demo-notebooks/guided-demos
- name: Switch to kind-cluster context to print logs
if: always() && steps.deploy.outcome == 'success'
run: kubectl config use-context kind-cluster
- name: Print debug info
if: always() && steps.deploy.outcome == 'success'
run: |
echo "Printing debug info"
kubectl describe pods -n default
- name: Print CodeFlare operator logs
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 ${TEMP_DIR}/codeflare-operator.log
- name: Print Kueue operator logs
if: always() && steps.deploy.outcome == 'success'
run: |
echo "Printing Kueue operator logs"
KUEUE_CONTROLLER_POD=$(kubectl get pods -n kueue-system | grep kueue-controller | awk '{print $1}')
kubectl logs -n kueue-system --tail -1 ${KUEUE_CONTROLLER_POD} | tee ${TEMP_DIR}/kueue.log
- name: Print KubeRay operator logs
if: always() && steps.deploy.outcome == 'success'
run: |
echo "Printing KubeRay operator logs"
kubectl logs -n ray-system --tail -1 -l app.kubernetes.io/name=kuberay | tee ${TEMP_DIR}/kuberay.log
- name: Export all KinD pod logs
uses: ./common/github-actions/kind-export-logs
if: always() && steps.deploy.outcome == 'success'
with:
output-directory: ${TEMP_DIR}
- name: Upload logs
uses: actions/upload-artifact@v4
if: always() && steps.deploy.outcome == 'success'
with:
name: logs
retention-days: 10
path: |
${{ env.TEMP_DIR }}/**/*.log