ERS PATC implementation #114
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| name: End-to-end tests on AWS GPU instance | |
| on: | |
| pull_request: | |
| branches: [ "main" ] | |
| max-concurrency: 1 | |
| # Allows to run this workflow manually from the Actions tab | |
| workflow_dispatch: | |
| max-concurrency: 1 | |
| jobs: | |
| start-gpu-instance: | |
| name: Start GPU instance | |
| runs-on: ubuntu-latest | |
| outputs: | |
| instance_ip: ${{ steps.trigger_instance.outputs.instance_ip }} | |
| steps: | |
| - name: Configure AWS credentials | |
| uses: aws-actions/configure-aws-credentials@v4 | |
| with: | |
| aws-access-key-id: ${{ secrets.FALCON_ACCESS_KEY_ID }} | |
| aws-secret-access-key: ${{ secrets.FALCON_SECRET_ACCESS_KEY }} | |
| aws-region: ${{ secrets.EAGLE_INSTANCE_REGION }} | |
| - uses: actions/checkout@v4 | |
| - name: Trigger the instance | |
| id: trigger_instance | |
| run: | | |
| set -o pipefail | |
| $GITHUB_WORKSPACE/build-automation/trigger_ec2_instance.sh ${{ secrets.EAGLE_INSTANCE_ID }} | tee trigger_out.txt | |
| ip_address=$(cat trigger_out.txt | tail -n 1) | |
| echo "instance_ip=$ip_address" >> $GITHUB_OUTPUT | |
| sleep 5 | |
| end-to-end-testing: | |
| name: End-to-end testing | |
| runs-on: ubuntu-latest | |
| needs: start-gpu-instance | |
| steps: | |
| - uses: actions/checkout@v4 | |
| with: | |
| submodules: 'recursive' | |
| - name: Create SSH key | |
| env: | |
| SSH_PRIVATE_KEY: ${{ secrets.EAGLE_SSH_KEY }} | |
| run: | | |
| key_location=~/.ssh/ | |
| mkdir -p $key_location | |
| key_path=$key_location/id_rsa | |
| echo "$SSH_PRIVATE_KEY" > $key_path | |
| sudo chmod 600 $key_path | |
| ssh-keygen -f $key_path -y > $key_path.pub | |
| - name: Transfer package and create e2e docker image | |
| id: transfer_package | |
| env: | |
| INSTANCE_USERNAME: ${{ secrets.EAGLE_USERNAME }} | |
| run: | | |
| instance_ip=${{ needs.start-gpu-instance.outputs.instance_ip }} | |
| rm -rf $GITHUB_WORKSPACE/.git | |
| scp -o "StrictHostKeyChecking no" -r $GITHUB_WORKSPACE $INSTANCE_USERNAME@$instance_ip:/tmp/ | |
| ssh -o "StrictHostKeyChecking no" $INSTANCE_USERNAME@$instance_ip "CUDAARCHS=\"50;52\" CMAKE_BUILD_TYPE=Release /tmp/asar-focus/build-automation/create_exe_in_container.sh /tmp/asar-focus cgialus/alus-ootpa-devel-extra" | |
| # To RAM disk | |
| ssh -o "StrictHostKeyChecking no" $INSTANCE_USERNAME@$instance_ip "cp -r /home/$INSTANCE_USERNAME/e2e /dev/shm/" | |
| # Do not use previous run results here. When skipping/failing would possibly use valid results for validation. | |
| ssh -o "StrictHostKeyChecking no" $INSTANCE_USERNAME@$instance_ip "rm -rf /dev/shm/e2e/last_run/*" | |
| - uses: actions/checkout@v4 | |
| - name: End-to-end tests | |
| id: e2e-run | |
| continue-on-error: true | |
| env: | |
| INSTANCE_USERNAME: ${{ secrets.EAGLE_USERNAME }} | |
| ACCESS_KEY_ID: ${{ secrets.PALMY_ACCESS_KEY_ID }} | |
| SECRET_ACCESS_KEY: ${{ secrets.PALMY_SECRET_ACCESS_KEY }} | |
| run: | | |
| instance_ip=${{ needs.start-gpu-instance.outputs.instance_ip }} | |
| exe_image_name="alus-ootpa-devel-extra-asar-focus-exe" | |
| # JFYI | |
| exe_container_name="alus-ootpa-devel-extra-asar-focus-exe-e2e" | |
| echo "[default]" > /tmp/cred | |
| echo "aws_access_key_id=${ACCESS_KEY_ID}" >> /tmp/cred | |
| echo "aws_secret_access_key=${SECRET_ACCESS_KEY}" >> /tmp/cred | |
| scp -o "StrictHostKeyChecking no" /tmp/cred $INSTANCE_USERNAME@$instance_ip:/tmp/ | |
| ssh -o "StrictHostKeyChecking no" $INSTANCE_USERNAME@$instance_ip "/tmp/asar-focus/build-automation/run_e2e_container.sh /dev/shm/e2e $exe_image_name /tmp/cred" | |
| - name: Stash assets from ramdisk | |
| env: | |
| INSTANCE_USERNAME: ${{ secrets.EAGLE_USERNAME }} | |
| run: | | |
| instance_ip=${{ needs.start-gpu-instance.outputs.instance_ip }} | |
| ssh -o "StrictHostKeyChecking no" $INSTANCE_USERNAME@$instance_ip "sudo chown -R $INSTANCE_USERNAME:$INSTANCE_USERNAME /dev/shm/e2e" | |
| ssh -o "StrictHostKeyChecking no" $INSTANCE_USERNAME@$instance_ip "rsync -av /dev/shm/e2e /home/$INSTANCE_USERNAME/" | |
| ssh -o "StrictHostKeyChecking no" $INSTANCE_USERNAME@$instance_ip "rm -rf /dev/shm/e2e" | |
| # This is needed because 'continue-on-error: true' for 'e2e-run' would not make this step failure in actions UI | |
| - name: Report E2E run error | |
| if: steps.e2e-run.outcome != 'success' | |
| run: exit 1 | |
| stop-gpu-instance: | |
| name: Stop GPU instance | |
| runs-on: ubuntu-latest | |
| needs: [start-gpu-instance, end-to-end-testing] | |
| if: | | |
| always() && | |
| needs.start-gpu-instance.result == 'success' | |
| steps: | |
| - name: Configure AWS Credentials | |
| uses: aws-actions/configure-aws-credentials@v4 | |
| with: | |
| aws-access-key-id: ${{ secrets.FALCON_ACCESS_KEY_ID }} | |
| aws-secret-access-key: ${{ secrets.FALCON_SECRET_ACCESS_KEY }} | |
| aws-region: ${{ secrets.EAGLE_INSTANCE_REGION }} | |
| - uses: actions/checkout@v4 | |
| - name: Stop GPU instance | |
| run: | | |
| $GITHUB_WORKSPACE/build-automation/stop_ec2_instance.sh ${{ secrets.EAGLE_INSTANCE_ID }} |