preview-build-deploy-images.yml

name: Preview

on:
  pull_request:
    types: [opened, synchronize, reopened]
    branches-ignore:
      - "main"

concurrency:
  group: pr-${{ github.event.pull_request.number }}-build-images
  cancel-in-progress: true

jobs:
  create_deployment:
    name: Create Deployment
    runs-on: blacksmith-2vcpu-ubuntu-2204
    if: github.head_ref != 'qa' || github.base_ref != 'main'
    permissions:
      contents: read
      pull-requests: write
      issues: write
      deployments: write
    outputs:
      deployment_id: ${{ steps.deployment.outputs.id }}
    steps:
      - name: Find Build & Deploy Images Comment
        uses: peter-evans/find-comment@v3
        id: preview-build-deploy-images-comment
        with:
          issue-number: ${{ github.event.pull_request.number }}
          comment-author: "github-actions[bot]"
          body-includes: "Preview - Build & Deploy Images"

      - name: Create initial status comment
        uses: peter-evans/create-or-update-comment@v3
        with:
          comment-id: ${{ steps.preview-build-deploy-images-comment.outputs.comment-id }}
          issue-number: ${{ github.event.pull_request.number }}
          body: |
            ## Preview - Build & Deploy Images
            ⏳ Building images...
            ⏳ Deploy images

            🔍 Track progress in the [workflow run](${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }})
          edit-mode: replace

      - name: Deactivate previous deployments
        run: |
          gh api \
            --method GET \
            -H "Accept: application/vnd.github+json" \
            /repos/${{ github.repository }}/deployments \
            -f ref="${{ github.event.pull_request.head.ref }}" \
            -f environment=preview-pr-${{ github.event.pull_request.number }} \
            --jq '.[] | .id' | while read -r id; do
              gh api \
                --method POST \
                -H "Accept: application/vnd.github+json" \
                /repos/${{ github.repository }}/deployments/$id/statuses \
                -f state="inactive" \
                -f environment="preview-pr-${{ github.event.pull_request.number }}" \
                -f description="Superseded by newer deployment"
            done
        env:
          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}

      - name: Create deployment
        id: deployment
        run: |
          DEPLOYMENT_ID=$(gh api \
            --method POST \
            -H "Accept: application/vnd.github+json" \
            /repos/${{ github.repository }}/deployments \
            -f ref="${{ github.event.pull_request.head.ref }}" \
            -f description="Building and deploying PR #${{ github.event.pull_request.number }} (${GITHUB_SHA::7})" \
            -F auto_merge=false \
            -f required_contexts\[\] \
            -f environment=preview-pr-${{ github.event.pull_request.number }} \
            -F transient_environment=true \
            -F production_environment=false \
            --jq '.id')
          echo "id=$DEPLOYMENT_ID" >> $GITHUB_OUTPUT
        env:
          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}

      - name: Update deployment status (building images)
        run: |
          gh api \
            --method POST \
            -H "Accept: application/vnd.github+json" \
            /repos/${{ github.repository }}/deployments/${{ steps.deployment.outputs.id }}/statuses \
            -f state="in_progress" \
            -f description="Building Docker images..." \
            -f environment=preview-pr-${{ github.event.pull_request.number }} \
            -f log_url="${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}"
        env:
          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}

  build_images:
    name: Build Images
    needs: create_deployment
    permissions:
      contents: read
      pull-requests: write
      issues: write
      deployments: write
    runs-on: ${{ matrix.runner }}
    if: github.head_ref != 'qa' || github.base_ref != 'main'
    timeout-minutes: 30
    strategy:
      matrix:
        include:
          - service: client
            runner: blacksmith-8vcpu-ubuntu-2204
          - service: api
            runner: blacksmith-2vcpu-ubuntu-2204
          - service: connection
            runner: blacksmith-4vcpu-ubuntu-2204
          - service: files
            runner: blacksmith-4vcpu-ubuntu-2204
          - service: multiplayer
            runner: blacksmith-4vcpu-ubuntu-2204
      fail-fast: true
    steps:
      - name: Checkout code
        uses: actions/checkout@v4

      - name: Generate Build Metadata
        id: build-metadata
        run: |
          echo "BUILD_TIME=$(date -u +'%Y-%m-%dT%H:%M:%SZ')" >> $GITHUB_OUTPUT
          echo "GIT_SHA_SHORT=$(git rev-parse --short HEAD)" >> $GITHUB_OUTPUT
          echo "BRANCH_NAME=$(echo "${{ github.head_ref }}" | tr '/' '-')" >> $GITHUB_OUTPUT

      - name: Configure AWS Credentials
        uses: aws-actions/configure-aws-credentials@v4
        with:
          aws-access-key-id: ${{ secrets.AWS_ACCESS_KEY_ID_DEVELOPMENT }}
          aws-secret-access-key: ${{ secrets.AWS_SECRET_ACCESS_KEY_DEVELOPMENT }}
          aws-region: ${{ secrets.AWS_REGION }}

      - name: Login to Amazon ECR Private
        id: login-ecr
        uses: aws-actions/amazon-ecr-login@v2

      - name: Define repository name
        id: repo-name
        run: |
          echo "REPO_NAME=quadratic-${{ matrix.service }}" >> $GITHUB_OUTPUT

      - name: Create Private ECR Repository
        id: create-ecr
        env:
          REPO_NAME: ${{ steps.repo-name.outputs.REPO_NAME }}
        run: |
          # Try to describe the repository first
          if ! aws ecr describe-repositories --repository-names $REPO_NAME 2>/dev/null; then
            # Repository doesn't exist, create it
            aws ecr create-repository --repository-name $REPO_NAME || true
          fi

          # Get the repository URI either way
          REPO_INFO=$(aws ecr describe-repositories --repository-names $REPO_NAME)
          ECR_URL=$(echo $REPO_INFO | jq -r '.repositories[0].repositoryUri')
          echo "ECR_URL=$ECR_URL" >> $GITHUB_OUTPUT

      - name: Set up Docker Buildx
        uses: docker/setup-buildx-action@v3
        with:
          driver-opts: |
            image=moby/buildkit:latest
            network=host

      - name: Build and push
        uses: useblacksmith/build-push-action@v1
        with:
          context: .
          file: quadratic-${{ matrix.service }}/Dockerfile
          push: true
          tags: ${{ steps.create-ecr.outputs.ECR_URL }}:pr-${{ github.event.pull_request.number }}
          build-args: |
            BUILD_TIME=${{ steps.build-metadata.outputs.BUILD_TIME }}
            GIT_SHA_SHORT=${{ steps.build-metadata.outputs.GIT_SHA_SHORT }}
            BRANCH_NAME=${{ steps.build-metadata.outputs.BRANCH_NAME }}
            PR_NUMBER=${{ github.event.pull_request.number }}
          labels: |
            org.opencontainers.image.created=${{ steps.build-metadata.outputs.BUILD_TIME }}
            org.opencontainers.image.revision=${{ steps.build-metadata.outputs.GIT_SHA_SHORT }}

      - name: Find Build & Deploy Images Comment
        if: failure()
        uses: peter-evans/find-comment@v3
        id: preview-build-deploy-images-comment
        with:
          issue-number: ${{ github.event.pull_request.number }}
          comment-author: "github-actions[bot]"
          body-includes: "Preview - Build & Deploy Images"

      - name: Update comment on build images failure
        if: failure()
        uses: peter-evans/create-or-update-comment@v3
        with:
          comment-id: ${{ steps.preview-build-deploy-images-comment.outputs.comment-id }}
          issue-number: ${{ github.event.pull_request.number }}
          body: |
            ## Preview - Build & Deploy Images
            ❌ Build images
            ❌ Deploy images

            🔍 Please check the [workflow run](${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}) for details.
          edit-mode: replace

      - name: Update deployment status (failure)
        if: failure()
        run: |
          gh api \
            --method POST \
            -H "Accept: application/vnd.github+json" \
            /repos/${{ github.repository }}/deployments/${{ needs.create_deployment.outputs.deployment_id }}/statuses \
            -f state="failure" \
            -f description="Failed to build ${{ matrix.service }} image" \
            -f environment=preview-pr-${{ github.event.pull_request.number }} \
            -f log_url="${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}"
        env:
          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}

  deploy_images:
    name: Deploy Images
    needs: [create_deployment, build_images]
    permissions:
      contents: read
      pull-requests: write
      issues: write
      deployments: write
    runs-on: blacksmith-2vcpu-ubuntu-2204
    if: github.head_ref != 'qa' || github.base_ref != 'main'
    timeout-minutes: 30
    env:
      STACK_NAME: pr-${{ github.event.pull_request.number }}
      MAX_ATTEMPTS: 50
    steps:
      - name: Find Build & Deploy Images Comment
        uses: peter-evans/find-comment@v3
        id: preview-build-deploy-images-comment-start
        with:
          issue-number: ${{ github.event.pull_request.number }}
          comment-author: "github-actions[bot]"
          body-includes: "Preview - Build & Deploy Images"

      - name: Update comment on deploy images start
        uses: peter-evans/create-or-update-comment@v3
        with:
          comment-id: ${{ steps.preview-build-deploy-images-comment-start.outputs.comment-id }}
          issue-number: ${{ github.event.pull_request.number }}
          body: |
            ## Preview - Build & Deploy Images
            ✅ Build images
            ⏳ Deploying images...

            🔍 Track progress in the [workflow run](${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }})
          edit-mode: replace

      - name: Update deployment status (deploying)
        run: |
          gh api \
            --method POST \
            -H "Accept: application/vnd.github+json" \
            /repos/${{ github.repository }}/deployments/${{ needs.create_deployment.outputs.deployment_id }}/statuses \
            -f state="in_progress" \
            -f description="Deploying images..." \
            -f environment=preview-pr-${{ github.event.pull_request.number }} \
            -f log_url="${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}"
        env:
          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}

      - name: Configure AWS Credentials
        uses: aws-actions/configure-aws-credentials@v4
        with:
          aws-access-key-id: ${{ secrets.AWS_ACCESS_KEY_ID_DEVELOPMENT }}
          aws-secret-access-key: ${{ secrets.AWS_SECRET_ACCESS_KEY_DEVELOPMENT }}
          aws-region: ${{ secrets.AWS_REGION }}

      - name: Wait for stack deployment
        id: check-stack
        run: |
          ATTEMPTS=0
          echo "Waiting for stack deployment..."
          while [ $ATTEMPTS -lt ${{ env.MAX_ATTEMPTS }} ]; do
            if ! STATUS=$(aws cloudformation describe-stacks \
              --stack-name ${{ env.STACK_NAME }} \
              --query 'Stacks[0].StackStatus' \
              --output text 2>&1); then
              echo "Error getting stack status: $STATUS"
              echo "Stack might not exist yet. Waiting..."
              sleep 30
              ATTEMPTS=$((ATTEMPTS + 1))
              continue
            fi
            
            echo "Current stack status: $STATUS"
            
            # Fail if stack is in a failed or rollback state
            if [[ $STATUS == *FAILED* ]] || [[ $STATUS == *ROLLBACK* ]]; then
              echo "::error::Stack is in a failed or rollback state: $STATUS"
              exit 1
            fi
            
            # Continue if stack is ready
            if [[ $STATUS == "CREATE_COMPLETE" ]] || [[ $STATUS == "UPDATE_COMPLETE" ]]; then
              echo "::notice::Stack is ready with status: $STATUS"
              break
            fi
            
            # Wait and check again if stack is still being created/updated
            if [[ $STATUS == *IN_PROGRESS* ]]; then
              echo "Stack operation in progress. Waiting 30 seconds..."
              sleep 30
              ATTEMPTS=$((ATTEMPTS + 1))
              continue
            fi
          done

          if [ $ATTEMPTS -eq ${{ env.MAX_ATTEMPTS }} ]; then
            echo "::error::Timeout waiting for stack to be ready"
            exit 1
          fi

      - name: Get EC2 Instance ID
        id: get-instance
        run: |
          INSTANCE_ID=$(aws cloudformation describe-stack-resources \
            --stack-name ${{ env.STACK_NAME }} \
            --logical-resource-id EC2Instance \
            --query 'StackResources[0].PhysicalResourceId' \
            --output text)
          if [ -z "$INSTANCE_ID" ]; then
            echo "::error::Failed to get EC2 instance ID"
            exit 1
          fi
          echo "instance_id=$INSTANCE_ID" >> $GITHUB_OUTPUT

      - name: Wait for instance to be ready
        run: |
          aws ec2 wait instance-status-ok \
            --instance-ids ${{ steps.get-instance.outputs.instance_id }}

      - name: Run deployment script on EC2
        id: deploy
        run: |
          COMMAND_ID=$(aws ssm send-command \
            --instance-ids ${{ steps.get-instance.outputs.instance_id }} \
            --document-name "AWS-RunShellScript" \
            --parameters commands=["cd /quadratic-selfhost && ./login.sh && ./pull_start.sh"] \
            --comment "Deploying new images after build" \
            --query 'Command.CommandId' \
            --output text)

          # Wait for command completion
          ATTEMPTS=0
          echo "Waiting for deployment command to complete..."
          while [ $ATTEMPTS -lt ${{ env.MAX_ATTEMPTS }} ]; do
            STATUS=$(aws ssm get-command-invocation \
              --command-id "$COMMAND_ID" \
              --instance-id ${{ steps.get-instance.outputs.instance_id }} \
              --query "Status" \
              --output text 2>/dev/null || echo "Pending")
            
            echo "Command status: $STATUS"
            
            if [ "$STATUS" = "Success" ]; then
              echo "Deployment completed successfully"
              exit 0
            elif [ "$STATUS" = "Failed" ] || [ "$STATUS" = "Cancelled" ] || [ "$STATUS" = "TimedOut" ]; then
              echo "Deployment failed with status: $STATUS"
              
              # Get command output for debugging
              aws ssm get-command-invocation \
                --command-id "$COMMAND_ID" \
                --instance-id ${{ steps.get-instance.outputs.instance_id }} \
                --query "StandardOutputContent" \
                --output text
                
              exit 1
            fi
            
            echo "Waiting for deployment to complete... (Attempt $((ATTEMPTS + 1))/${{ env.MAX_ATTEMPTS }})"
            sleep 30
            ATTEMPTS=$((ATTEMPTS + 1))
          done

          echo "Deployment timed out after ${{ env.MAX_ATTEMPTS }} attempts"
          exit 1

      - name: Delete untagged (old) images
        run: |
          for service in client api connection files multiplayer; do
            REPO_NAME="quadratic-${service}"
            
            # Get list of untagged image digests
            UNTAGGED_IMAGES=$(aws ecr list-images \
              --repository-name "$REPO_NAME" \
              --filter "tagStatus=UNTAGGED" \
              --query 'imageIds[*].imageDigest' \
              --output json)
            
            # Delete untagged images if any exist
            if [ "$UNTAGGED_IMAGES" != "[]" ]; then
              echo "Deleting untagged images from $REPO_NAME"
              aws ecr batch-delete-image \
                --repository-name "$REPO_NAME" \
                --image-ids "$(echo $UNTAGGED_IMAGES | jq -r 'map({imageDigest: .}) | @json')"
            else
              echo "No untagged images found in $REPO_NAME"
            fi
          done

      - name: Generate Deploy Metadata
        id: deploy-metadata
        run: |
          # Sanitize branch name for DNS
          BRANCH_NAME="${{ github.event.pull_request.head.ref }}"
          SANITIZED_BRANCH_NAME=$(echo "$BRANCH_NAME" | tr '[:upper:]' '[:lower:]' | sed 's/[^a-z0-9-]/-/g' | sed 's/-\+/-/g' | sed 's/^-\|-$//')

          echo "WebsiteURL=https://${SANITIZED_BRANCH_NAME}.quadratic-preview.com" >> $GITHUB_OUTPUT
          echo "DEPLOY_TIME=$(date -u +'%b %d, %Y at %I:%M %p UTC')" >> $GITHUB_OUTPUT

      - name: Find Build & Deploy Images Comment
        if: always()
        uses: peter-evans/find-comment@v3
        id: preview-build-deploy-images-comment-update
        with:
          issue-number: ${{ github.event.pull_request.number }}
          comment-author: "github-actions[bot]"
          body-includes: "Preview - Build & Deploy Images"

      - name: Update comment on deploy images success
        if: success()
        uses: peter-evans/create-or-update-comment@v3
        with:
          comment-id: ${{ steps.preview-build-deploy-images-comment-update.outputs.comment-id }}
          issue-number: ${{ github.event.pull_request.number }}
          body: |
            ## Preview - Build & Deploy Images
            ✅ Build images
            ✅ Deploy images

            🕒 Last deployed: ${{ steps.deploy-metadata.outputs.DEPLOY_TIME }}

            🔗 URL: ${{ steps.deploy-metadata.outputs.WebsiteURL }}
          edit-mode: replace

      - name: Update comment on deploy images failure
        if: failure()
        uses: peter-evans/create-or-update-comment@v3
        with:
          comment-id: ${{ steps.preview-build-deploy-images-comment-update.outputs.comment-id }}
          issue-number: ${{ github.event.pull_request.number }}
          body: |
            ## Preview - Build & Deploy Images
            ✅ Build images
            ❌ Deploy images

            🔍 Please check the [workflow run](${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}) for details.
          edit-mode: replace

      - name: Update deployment status (success)
        if: success()
        run: |
          gh api \
            --method POST \
            -H "Accept: application/vnd.github+json" \
            /repos/${{ github.repository }}/deployments/${{ needs.create_deployment.outputs.deployment_id }}/statuses \
            -f state="success" \
            -f environment_url="${{ steps.deploy-metadata.outputs.WebsiteURL }}" \
            -f description="Deployment successful!" \
            -f environment=preview-pr-${{ github.event.pull_request.number }} \
            -f log_url="${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}"
        env:
          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}

      - name: Update deployment status (failure)
        if: failure()
        run: |
          gh api \
            --method POST \
            -H "Accept: application/vnd.github+json" \
            /repos/${{ github.repository }}/deployments/${{ needs.create_deployment.outputs.deployment_id }}/statuses \
            -f state="failure" \
            -f description="Deployment failed" \
            -f environment=preview-pr-${{ github.event.pull_request.number }} \
            -f log_url="${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}"
        env:
          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}