torqvoice/.github/workflows/deploy-prod.yml

name: "[Prod] Deploy"

on:
  workflow_dispatch:
    inputs:
      tag:
        description: "Version tag to deploy (e.g. v1.2.39). Leave blank to deploy the newest published release. 'latest' is refused."
        required: false
        default: ""

jobs:
  deploy:
    runs-on: [self-hosted, Linux, X64, hetzner]

    steps:
      - name: Resolve version tag
        id: tag
        run: |
          TAG="${{ inputs.tag }}"
          if [ -z "$TAG" ]; then
            # Newest published release (drafts are excluded, so publishing the
            # release in GitHub is the gate that makes it deployable).
            TAG=$(curl -fsSL -H "Authorization: Bearer ${{ secrets.GITHUB_TOKEN }}" \
              "https://api.github.com/repos/${{ github.repository }}/releases/latest" \
              | grep -m1 '"tag_name"' | cut -d'"' -f4)
            echo "Resolved newest published release: $TAG"
          fi
          if [ -z "$TAG" ] || [ "$TAG" = "latest" ]; then
            echo "::error::No deployable version tag ('latest' is refused; rollback needs a named target)."
            exit 1
          fi
          echo "value=$TAG" >> "$GITHUB_OUTPUT"

      - name: Log in to GitHub Container Registry
        run: echo "${{ secrets.GHCR_TOKEN }}" | docker login ghcr.io -u ${{ github.actor }} --password-stdin

      - name: Create deployment files
        env:
          VIRTUAL_HOST: ${{ secrets.CLOUD_VIRTUAL_HOST }}
          DATABASE_URL: ${{ secrets.CLOUD_DATABASE_URL }}
          BETTER_AUTH_SECRET: ${{ secrets.CLOUD_BETTER_AUTH_SECRET }}
          APP_URL: ${{ secrets.CLOUD_APP_URL }}
          STRIPE_SECRET_KEY: ${{ secrets.CLOUD_STRIPE_SECRET_KEY }}
          STRIPE_WEBHOOK_SECRET: ${{ secrets.CLOUD_STRIPE_WEBHOOK_SECRET }}
          STRIPE_PRO_PRICE_ID: ${{ secrets.CLOUD_STRIPE_PRO_PRICE_ID }}
          STRIPE_ENTERPRISE_PRICE_ID: ${{ secrets.CLOUD_STRIPE_ENTERPRISE_PRICE_ID }}
          DATA_PATH: ${{ secrets.DATA_PATH }}
          POSTHOG_HOST: ${{ secrets.NEXT_PUBLIC_POSTHOG_HOST }}
          POSTHOG_KEY: ${{ secrets.NEXT_PUBLIC_POSTHOG_KEY }}
          BACKUP_HEARTBEAT_TOKEN: ${{ secrets.BACKUP_HEARTBEAT_TOKEN }}
        run: |
          mkdir -p $HOME/torqvoice-deploy/prod
          cd $HOME/torqvoice-deploy/prod

          env | grep -E '^(VIRTUAL_HOST|DATA_PATH|DATABASE_URL|BETTER_AUTH_SECRET|APP_URL|STRIPE_|POSTHOG_|BACKUP_)' > .env

          cat > docker-compose.yml << 'COMPOSE'
          networks:
            proxy:
              external: true
              name: proxy
          services:
            torqvoice-app:
              container_name: torqvoice-app
              image: ghcr.io/torqvoice/torqvoice:${APP_TAG:-latest}
              restart: unless-stopped
              volumes:
                - ${DATA_PATH}/torqvoice_data:/app/data
              expose:
                - "3000"
              environment:
                VIRTUAL_HOST: ${VIRTUAL_HOST}
                VIRTUAL_PORT: "3000"
                LETSENCRYPT_HOST: ${VIRTUAL_HOST}
                DATABASE_URL: ${DATABASE_URL}
                BETTER_AUTH_SECRET: ${BETTER_AUTH_SECRET}
                NEXT_PUBLIC_APP_URL: ${APP_URL}
                TORQVOICE_MODE: cloud
                STRIPE_SECRET_KEY: ${STRIPE_SECRET_KEY}
                STRIPE_WEBHOOK_SECRET: ${STRIPE_WEBHOOK_SECRET}
                STRIPE_PRO_PRICE_ID: ${STRIPE_PRO_PRICE_ID}
                STRIPE_ENTERPRISE_PRICE_ID: ${STRIPE_ENTERPRISE_PRICE_ID}
                POSTHOG_HOST: ${POSTHOG_HOST}
                POSTHOG_KEY: ${POSTHOG_KEY}
                BACKUP_HEARTBEAT_TOKEN: ${BACKUP_HEARTBEAT_TOKEN}
              networks:
                - proxy
          COMPOSE

      - name: Ensure data directory permissions
        run: |
          DATA_DIR="${{ secrets.DATA_PATH }}/torqvoice_data"
          mkdir -p "$DATA_DIR"
          docker run --rm -v "$DATA_DIR":/data alpine chown -R 1001:1001 /data

      - name: Check free disk space
        run: |
          MIN_GB=5
          DOCKER_ROOT=$(docker info --format '{{.DockerRootDir}}' 2>/dev/null || echo /var/lib/docker)
          AVAIL_KB=$(df -Pk "$DOCKER_ROOT" | awk 'NR==2 {print $4}')
          AVAIL_GB=$(awk "BEGIN {printf \"%.1f\", $AVAIL_KB/1024/1024}")
          echo "Free space on $DOCKER_ROOT: ${AVAIL_GB} GB (minimum ${MIN_GB} GB)"
          if [ "$AVAIL_KB" -lt $((MIN_GB * 1024 * 1024)) ]; then
            echo "::error::Not enough disk space to pull image: ${AVAIL_GB} GB free, need ${MIN_GB} GB. Aborting deploy."
            exit 1
          fi

      - name: Back up production database
        env:
          DATABASE_URL: ${{ secrets.CLOUD_DATABASE_URL }}
        run: |
          set -euo pipefail

          DB_USER=$(echo "$DATABASE_URL" | sed -E 's|^[^:]+://([^:]+):.*|\1|')
          DB_NAME=$(echo "$DATABASE_URL" | sed -E 's|.*/([^/?]+)(\?.*)?$|\1|')

          BACKUP_DIR="${{ secrets.DATA_PATH }}/db-backups/$DB_NAME"
          # The volume is root-owned and the runner is not root. Binding the
          # path into a container makes Docker create it, and the chown hands
          # it to the runner so the dump redirect below can write there.
          docker run --rm -v "$BACKUP_DIR":/b alpine chown "$(id -u):$(id -g)" /b

          # The Postgres data directory lives on this filesystem: filling it
          # takes the database down, not just the backup.
          MIN_GB=2
          AVAIL_KB=$(df -Pk "$BACKUP_DIR" | awk 'NR==2 {print $4}')
          if [ "$AVAIL_KB" -lt $((MIN_GB * 1024 * 1024)) ]; then
            echo "::error::Only $(awk "BEGIN {printf \"%.1f\", $AVAIL_KB/1024/1024}") GB free where the database lives. Refusing to dump."
            exit 1
          fi

          DUMP="$BACKUP_DIR/${DB_NAME}-$(date +%Y%m%d-%H%M%S)-pre-${{ steps.tag.outputs.value }}.dump"
          echo "Dumping $DB_NAME to $DUMP"
          docker exec torqvoice-db pg_dump -U "$DB_USER" -d "$DB_NAME" -Fc > "$DUMP"

          # A dump nobody can read is worse than no dump. No backup, no deploy.
          docker exec -i torqvoice-db pg_restore --list < "$DUMP" > /dev/null
          echo "Verified $(du -h "$DUMP" | cut -f1) backup"

          # Keep the 10 most recent pre-deploy dumps; hourly dumps in the same
          # folder belong to the backup cron and age out there.
          ls -t "$BACKUP_DIR"/*-pre-*.dump 2>/dev/null | tail -n +11 | xargs -r rm --

      - name: Deploy
        run: |
          cd $HOME/torqvoice-deploy/prod
          APP_TAG=${{ steps.tag.outputs.value }} docker compose up -d --pull always

      - name: Health check
        run: |
          echo "Waiting for prod container to be ready..."
          for i in $(seq 1 12); do
            STATUS=$(docker inspect --format='{{.State.Status}}' torqvoice-app 2>/dev/null || echo "not found")
            if [ "$STATUS" = "running" ]; then
              echo "Deployed ${{ steps.tag.outputs.value }}."
              exit 0
            fi
            echo "Status: $STATUS, retrying in 5s..."
            sleep 5
          done
          echo "::error::Container failed to start"
          docker logs torqvoice-app --tail 50
          exit 1

      - name: Prune old images and build cache
        run: docker image prune -f && docker builder prune -f