[nightly] Create Nightly Pipeline, make docker-nightly-publish.yml & integration.yml more modular

.github/workflows/docker-nightly-publish.yml

@@ -7,6 +7,11 @@ on:
         description: 'release/nightly/temp, default is nightly'
         required: true
         default: 'nightly'
+      skip_nightly_integ_test:
+        description: 'buld and push the nightly without running integ test'
+        required: false
+        default: false
+        type: boolean
   workflow_call:
     inputs:
       mode:
@@ -21,6 +26,10 @@ permissions:
   id-token: write
   contents: read
 
+env:
+  AWS_ECR_REPO: "185921645874.dkr.ecr.us-east-1.amazonaws.com/djl-ci-temp"
+  DOCKER_HUB_REPO: "deepjavalibrary/djl-serving"
+
 jobs:
   nightly-build:
     runs-on: ubuntu-latest
@@ -70,29 +79,25 @@ jobs:
         run: |
           ./gradlew --refresh-dependencies :serving:dockerDeb -Psnapshot
       - name: Build and push nightly docker image
-        if: ${{ inputs.mode == '' || inputs.mode == 'nightly' }}
         working-directory: serving/docker
         run: |
           export NIGHTLY="-nightly"
           docker compose build --no-cache \
           --build-arg djl_version=${{ env.DJL_VERSION }}-SNAPSHOT \
           --build-arg djl_serving_version=${{ env.SERVING_VERSION }}-SNAPSHOT \
           ${{ matrix.arch }}
-          docker compose push ${{ matrix.arch }}
-      - name: Build and push temp image
-        if: ${{ inputs.mode == 'temp' }}
+      - name: Tag and push temp image to ECR repo
+        if: ${{ !inputs.skip_nightly_integ_test && inputs.mode == 'nightly' ||  inputs.mode == 'temp'}}
         working-directory: serving/docker
         run: |
-          export NIGHTLY="-nightly"
-          docker compose build --no-cache \
-          --build-arg djl_version=${{ env.DJL_VERSION }}-SNAPSHOT \
-          --build-arg djl_serving_version=${{ env.SERVING_VERSION }}-SNAPSHOT \
-          ${{ matrix.arch }}
-          repo="185921645874.dkr.ecr.us-east-1.amazonaws.com/djl-ci-temp"
-          aws ecr get-login-password --region us-east-1 | docker login --username AWS --password-stdin $repo
-          tempTag="$repo:${{ matrix.arch }}-${GITHUB_SHA}"
-          docker tag deepjavalibrary/djl-serving:${{ matrix.arch }}-nightly $tempTag
+          aws ecr get-login-password --region us-east-1 | docker login --username AWS --password-stdin ${{env.AWS_ECR_REPO}}
+          tempTag="${{ env.AWS_ECR_REPO }}:${{ matrix.arch }}-${GITHUB_SHA}"
+          docker tag ${{ env.DOCKER_HUB_REPO }}:${{ matrix.arch }}-nightly $tempTag
           docker push $tempTag
+      - name: Push nightly to dockerhub
+        if: ${{ inputs.skip_nightly_integ_test && inputs.mode == 'nightly' }}
+        run: |
+            docker push ${{ env.DOCKER_HUB_REPO }}:${{ matrix.arch }}-nightly
       - name: Build and push release docker image
         if: ${{ inputs.mode == 'release' }}
         working-directory: serving/docker
@@ -111,6 +116,34 @@ jobs:
           docker tag deepjavalibrary/djl-serving:${{ env.SERVING_VERSION }} deepjavalibrary/djl-serving:latest
           docker push deepjavalibrary/djl-serving:latest
 
+  run-integration-tests:
+    if: ${{ inputs.mode == 'nightly' && !inputs.skip_integ_test }}
+    needs: [nightly-build, nightly-aarch64]
+    uses: ./.github/workflows/integration.yml
+    secrets: inherit
+    with:
+      djl-version: temp
+
+  push-to-dockerhub:
+    runs-on: ubuntu-latest
+    needs: [run-integration-tests]
+    strategy:
+      matrix:
+        arch: [ cpu, cpu-full, pytorch-inf2, pytorch-gpu, tensorrt-llm, lmi ]
+    steps:
+      - name: Login to Docker
+        uses: docker/login-action@v3
+        with:
+          username: ${{ secrets.DOCKER_USERNAME }}
+          password: ${{ secrets.DOCKER_PASSWORD }}
+      - name: Pull Image from ECR and Push it to Dockerhub
+        run: |
+          aws ecr get-login-password --region us-east-1 | docker login --username AWS --password-stdin ${{env.AWS_ECR_REPO}}
+          tempTag=" ${{env.AWS_ECR_REPO}}:${{ matrix.arch }}-${GITHUB_SHA}"
+          docker pull $tempTag
+          docker tag $tempTag ${{ env.DOCKER_HUB_REPO }}:${{ matrix.arch }}-nightly
+          docker push ${{ env.DOCKER_HUB_REPO }}:${{ matrix.arch }}-nightly
+
   create-runner:
     runs-on: [ self-hosted, scheduler ]
     steps:
@@ -186,7 +219,7 @@ jobs:
           aarch64
           docker compose push aarch64
       - name: Build and push temp image
-        if: ${{ inputs.mode == 'temp' }}
+        if: ${{ inputs.mode == 'temp' || inputs.mode == 'nightly' }}
         working-directory: serving/docker
         run: |
           export NIGHTLY="-nightly"
@@ -195,8 +228,8 @@ jobs:
           --build-arg djl_serving_version=${{ env.SERVING_VERSION }}-SNAPSHOT \
           aarch64
           repo="185921645874.dkr.ecr.us-east-1.amazonaws.com/djl-ci-temp"
-          aws ecr get-login-password --region us-east-1 | docker login --username AWS --password-stdin $repo
-          tempTag="$repo:aarch64-${GITHUB_SHA}"
+          aws ecr get-login-password --region us-east-1 | docker login --username AWS --password-stdin ${{env.AWS_ECR_REPO}}
+          tempTag="${{env.AWS_ECR_REPO}}:aarch64-${GITHUB_SHA}"
           docker tag deepjavalibrary/djl-serving:aarch64-nightly $tempTag
           docker push $tempTag
       - name: Build and push release docker image

.github/workflows/integration.yml

@@ -7,9 +7,15 @@ on:
         description: 'The released version of DJL'
         required: false
         default: ''
-  schedule:
-    - cron: '0 15 * * *'
-
+  workflow_call:
+    inputs:
+      djl-version:
+        description: 'The released version of DJL'
+        required: false
+        type: string
+        default: ''
+env:
+  AWS_ECR_REPO: "185921645874.dkr.ecr.us-east-1.amazonaws.com/djl-ci-temp"
 
 jobs:
   create-runners:
@@ -175,12 +181,18 @@ jobs:
           wget https://publish.djl.ai/awscurl/awscurl
           chmod +x awscurl
           mkdir outputs
+      - name: Configure AWS Credentials
+        uses: aws-actions/configure-aws-credentials@v4
+        with:
+          role-to-assume: arn:aws:iam::185921645874:role/github-actions-djl-serving
+          aws-region: us-east-1
       - name: Test
         working-directory: tests/integration
         env:
           TEST_DJL_VERSION: ${{ inputs.djl-version }}
         run: |
-          python -m pytest -k ${{ matrix.test.test }} tests.py
+          aws ecr get-login-password --region us-east-1 | docker login --username AWS --password-stdin ${{env.AWS_ECR_REPO}}
+          python -m pytest -s -k ${{ matrix.test.test }} tests.py
       - name: Cleanup
         working-directory: tests/integration
         run: |
@@ -224,11 +236,25 @@ jobs:
           python-version: '3.10.x'
       - name: Install pip dependencies
         run: pip3 install requests numpy pillow wheel
+      - name: Configure AWS Credentials
+        uses: aws-actions/configure-aws-credentials@v4
+        with:
+          role-to-assume: arn:aws:iam::185921645874:role/github-actions-djl-serving
+          aws-region: us-east-1
       - name: Build container name
-        run: ./serving/docker/scripts/docker_name_builder.sh pytorch-inf2 ${{ github.event.inputs.djl-version }}
+        run: |
+          ./serving/docker/scripts/docker_name_builder.sh pytorch-inf2 ${{ github.event.inputs.djl-version }}
       - name: Download models and dockers
         run: |
-          docker pull deepjavalibrary/djl-serving:$DJLSERVING_DOCKER_TAG
+          if [ ${{ github.event.inputs.djl-version }} == "temp"  ]; then
+            DOCKER_IMAGE_URI="185921645874.dkr.ecr.us-east-1.amazonaws.com/djl-ci-temp:pytorch-inf2-${GITHUB_SHA}"
+          else
+            DOCKER_IMAGE_URI="deepjavalibrary/djl-serving:$DJLSERVING_DOCKER_TAG"
+          fi
+          echo "DOCKER_IMAGE_URI=$DOCKER_IMAGE_URI" >>$GITHUB_ENV
+          aws ecr get-login-password --region us-east-1 | docker login --username AWS --password-stdin ${{env.AWS_ECR_REPO}}
+          echo $DOCKER_IMAGE_URI
+          docker pull $DOCKER_IMAGE_URI
       - name: Run djl_python unit/integration tests on container
         working-directory: engines/python/setup
         run: |
@@ -241,7 +267,7 @@ jobs:
           -v $PWD/:/opt/ml/model/ \
           -w /opt/ml/model \
           --device=/dev/neuron0:/dev/neuron0 \
-          deepjavalibrary/djl-serving:$DJLSERVING_DOCKER_TAG \
+          $DOCKER_IMAGE_URI \
           /bin/bash -c "'pip install /opt/ml/model/dist/*.whl pytest' && \
           pytest djl_python/tests/neuron_test_scripts/ | tee logs/results.log"