fquiz/.github/workflows/main.yml

name: fquiz 镜像构建与部署

on:
  push:
    branches:
      - dev
  workflow_dispatch:

concurrency:
  group: fquiz-build-and-deploy
  cancel-in-progress: true

permissions:
  contents: read

env:
  REGISTRY: ${{ vars.REGISTRY || 'crpi-u265r07n4blchcqo.cn-shanghai.personal.cr.aliyuncs.com' }}
  REGISTRY_NAMESPACE: ${{ vars.REGISTRY_NAMESPACE || 'ck-registry' }}

jobs:
  build-and-push:
    runs-on: ubuntu-latest
    outputs:
      api_image: ${{ steps.vars.outputs.api_image }}
      web_image: ${{ steps.vars.outputs.web_image }}
      image_tag: ${{ steps.vars.outputs.image_tag }}
    steps:
      - name: 拉取代码
        uses: actions/checkout@v6

      - name: 生成镜像变量
        id: vars
        shell: bash
        run: |
          NS="${{ env.REGISTRY_NAMESPACE }}"
          echo "api_image=${{ env.REGISTRY }}/${NS}/fquiz-api" >> "$GITHUB_OUTPUT"
          echo "web_image=${{ env.REGISTRY }}/${NS}/fquiz-web" >> "$GITHUB_OUTPUT"
          echo "image_tag=${GITHUB_SHA}" >> "$GITHUB_OUTPUT"

      - name: 安装 Buildx
        uses: docker/setup-buildx-action@v4
        with:
          driver-opts: |
            image=docker.m.daocloud.io/moby/buildkit:buildx-stable-1

      - name: 登录镜像仓库
        uses: docker/login-action@v4
        with:
          registry: ${{ env.REGISTRY }}
          username: ${{ secrets.REGISTRY_USERNAME }}
          password: ${{ secrets.REGISTRY_PASSWORD }}

      - name: 构建并推送 API 镜像
        uses: docker/build-push-action@v7
        with:
          context: ./api
          file: ./api/Dockerfile
          pull: true
          push: true
          build-args: |
            PIP_INDEX_URL=${{ secrets.PIP_INDEX_URL || vars.PIP_INDEX_URL || 'https://pypi.org/simple' }}
            PIP_DEFAULT_TIMEOUT=${{ vars.PIP_DEFAULT_TIMEOUT || '120' }}
          tags: |
            ${{ steps.vars.outputs.api_image }}:${{ steps.vars.outputs.image_tag }}
            ${{ steps.vars.outputs.api_image }}:latest
          cache-from: type=gha,scope=fquiz-api
          cache-to: type=gha,mode=max,scope=fquiz-api

      - name: 构建并推送 Web 镜像
        uses: docker/build-push-action@v7
        with:
          context: ./web
          file: ./web/Dockerfile
          pull: true
          push: true
          build-args: |
            NEXT_PUBLIC_API_BASE_URL=${{ vars.NEXT_PUBLIC_API_BASE_URL || 'http://127.0.0.1:8000' }}
            NEXT_PUBLIC_APP_BASE_PATH=${{ vars.NEXT_PUBLIC_APP_BASE_PATH || '/fl' }}
          tags: |
            ${{ steps.vars.outputs.web_image }}:${{ steps.vars.outputs.image_tag }}
            ${{ steps.vars.outputs.web_image }}:latest
          cache-from: type=gha,scope=fquiz-web
          cache-to: type=gha,mode=max,scope=fquiz-web

  deploy:
    runs-on: ubuntu-latest
    needs: build-and-push
    if: github.ref == 'refs/heads/dev'
    steps:
      - name: 校验部署参数
        env:
          SERVER_HOST: ${{ secrets.SERVER_HOST || vars.SERVER_HOST }}
          SERVER_USER: ${{ secrets.SERVER_USER || vars.SERVER_USER }}
          SERVER_SSH_KEY: ${{ secrets.SERVER_SSH_KEY }}
          SERVER_PASSWORD: ${{ secrets.SERVER_PASSWORD }}
          REGISTRY_USERNAME: ${{ secrets.REGISTRY_USERNAME }}
          REGISTRY_PASSWORD: ${{ secrets.REGISTRY_PASSWORD }}
        run: |
          set -euo pipefail
          [ -n "${SERVER_HOST}" ] || { echo "::error::缺少 SERVER_HOST（请在 Secrets 或 Variables 中配置）"; exit 1; }
          [ -n "${SERVER_USER}" ] || { echo "::error::缺少 SERVER_USER（请在 Secrets 或 Variables 中配置）"; exit 1; }
          if [ -z "${SERVER_SSH_KEY}" ] && [ -z "${SERVER_PASSWORD}" ]; then
            echo "::error::缺少登录凭据：请至少配置 SERVER_SSH_KEY 或 SERVER_PASSWORD"
            exit 1
          fi
          [ -n "${REGISTRY_USERNAME}" ] || { echo "::error::缺少 REGISTRY_USERNAME"; exit 1; }
          [ -n "${REGISTRY_PASSWORD}" ] || { echo "::error::缺少 REGISTRY_PASSWORD"; exit 1; }

      - name: 拉取代码
        uses: actions/checkout@v6

      - name: 准备远端部署目录
        uses: appleboy/ssh-action@v1.2.0
        env:
          DEPLOY_PATH: ${{ secrets.DEPLOY_PATH || vars.DEPLOY_PATH }}
        with:
          host: ${{ secrets.SERVER_HOST || vars.SERVER_HOST }}
          username: ${{ secrets.SERVER_USER || vars.SERVER_USER }}
          port: ${{ secrets.SERVER_PORT || vars.SERVER_PORT || 22 }}
          timeout: 120s
          command_timeout: 10m
          key: ${{ secrets.SERVER_SSH_KEY }}
          password: ${{ secrets.SERVER_PASSWORD }}
          envs: DEPLOY_PATH
          script: |
            set -euo pipefail
            DEPLOY_DIR="${DEPLOY_PATH:-/opt/fquiz}"
            mkdir -p "${DEPLOY_DIR}"

      - name: 同步 deploy/pro-deploy 配置到服务器
        uses: appleboy/scp-action@v1.0.0
        with:
          host: ${{ secrets.SERVER_HOST || vars.SERVER_HOST }}
          username: ${{ secrets.SERVER_USER || vars.SERVER_USER }}
          port: ${{ secrets.SERVER_PORT || vars.SERVER_PORT || 22 }}
          key: ${{ secrets.SERVER_SSH_KEY }}
          password: ${{ secrets.SERVER_PASSWORD }}
          timeout: 120s
          command_timeout: 10m
          source: "deploy/pro-deploy"
          target: ${{ secrets.DEPLOY_PATH || vars.DEPLOY_PATH || '/opt/fquiz' }}
          overwrite: true

      - name: 通过 SSH 拉取并更新容器
        uses: appleboy/ssh-action@v1.2.0
        env:
          DEPLOY_PATH: ${{ secrets.DEPLOY_PATH || vars.DEPLOY_PATH }}
          API_IMAGE: ${{ needs.build-and-push.outputs.api_image }}
          WEB_IMAGE: ${{ needs.build-and-push.outputs.web_image }}
          IMAGE_TAG: ${{ needs.build-and-push.outputs.image_tag }}
          NEXT_PUBLIC_API_BASE_URL: ${{ vars.NEXT_PUBLIC_API_BASE_URL || 'http://127.0.0.1:8000' }}
          FLOWER_BASIC_AUTH: ${{ secrets.FLOWER_BASIC_AUTH || vars.FLOWER_BASIC_AUTH || 'admin:admin' }}
          REGISTRY: ${{ env.REGISTRY }}
          REGISTRY_USERNAME: ${{ secrets.REGISTRY_USERNAME }}
          REGISTRY_PASSWORD: ${{ secrets.REGISTRY_PASSWORD }}
        with:
          host: ${{ secrets.SERVER_HOST || vars.SERVER_HOST }}
          username: ${{ secrets.SERVER_USER || vars.SERVER_USER }}
          port: ${{ secrets.SERVER_PORT || vars.SERVER_PORT || 22 }}
          timeout: 120s
          command_timeout: 45m
          key: ${{ secrets.SERVER_SSH_KEY }}
          password: ${{ secrets.SERVER_PASSWORD }}
          envs: DEPLOY_PATH,API_IMAGE,WEB_IMAGE,IMAGE_TAG,NEXT_PUBLIC_API_BASE_URL,FLOWER_BASIC_AUTH,REGISTRY,REGISTRY_USERNAME,REGISTRY_PASSWORD
          script: |
            set -euo pipefail
            export DOCKER_CLIENT_TIMEOUT="${DOCKER_CLIENT_TIMEOUT:-600}"
            export COMPOSE_HTTP_TIMEOUT="${COMPOSE_HTTP_TIMEOUT:-600}"

            DEPLOY_DIR="${DEPLOY_PATH:-/opt/fquiz}"
            cd "${DEPLOY_DIR}"
            [ -f deploy/pro-deploy/compose.yml ] || { echo "[error] deploy/pro-deploy/compose.yml 不存在"; exit 1; }
            [ -f deploy/pro-deploy/.env ] || { echo "[error] deploy/pro-deploy/.env 不存在"; exit 1; }
            [ -f deploy/pro-deploy/.env.prod ] || { echo "[error] deploy/pro-deploy/.env.prod 不存在"; exit 1; }
            mkdir -p \
              deploy/pro-deploy/data/postgres \
              deploy/pro-deploy/data/redis \
              deploy/pro-deploy/data/minio \
              deploy/pro-deploy/data/app/celery

            cat > .images.env <<ENV
            API_IMAGE=${API_IMAGE}:${IMAGE_TAG}
            WEB_IMAGE=${WEB_IMAGE}:${IMAGE_TAG}
            NEXT_PUBLIC_API_BASE_URL=${NEXT_PUBLIC_API_BASE_URL}
            NEXT_PUBLIC_APP_BASE_PATH=/fl
            FLOWER_BASIC_AUTH=${FLOWER_BASIC_AUTH}
            ENV

            echo "${REGISTRY_PASSWORD}" | docker login "${REGISTRY}" -u "${REGISTRY_USERNAME}" --password-stdin

            COMPOSE_CMD="docker compose"
            if ! docker compose version >/dev/null 2>&1; then
              COMPOSE_CMD="docker-compose"
            fi
            COMPOSE_PROJECT_NAME="$(awk -F= '/^COMPOSE_PROJECT_NAME=/{print $2; exit}' deploy/pro-deploy/.env | tr -d '[:space:]')"
            COMPOSE_PROJECT_NAME="${COMPOSE_PROJECT_NAME:-fquiz}"

            compose_run() {
              ${COMPOSE_CMD} -p "${COMPOSE_PROJECT_NAME}" --env-file deploy/pro-deploy/.env --env-file deploy/pro-deploy/.env.prod --env-file .images.env -f deploy/pro-deploy/compose.yml "$@"
            }

            pull_with_retry() {
              local max_retries=3
              local attempt=1
              while true; do
                if compose_run pull; then
                  break
                fi
                if [ "${attempt}" -ge "${max_retries}" ]; then
                  echo "[error] docker compose pull failed after ${max_retries} attempts."
                  return 1
                fi
                local sleep_seconds=$((attempt * 20))
                echo "[warn] docker compose pull failed (attempt ${attempt}/${max_retries}), retrying in ${sleep_seconds}s..."
                sleep "${sleep_seconds}"
                attempt=$((attempt + 1))
              done
            }

            pull_with_retry
            if ! compose_run up -d --remove-orphans; then
              echo "[error] docker compose up failed, collecting container diagnostics..."
              compose_run ps || true
              compose_run logs --no-color --tail 300 api || true
              compose_run logs --no-color --tail 200 db || true
              compose_run logs --no-color --tail 200 redis || true
              compose_run logs --no-color --tail 200 celery-worker || true
              compose_run logs --no-color --tail 200 celery-beat || true
              compose_run logs --no-color --tail 200 flower || true
              exit 1
            fi
            compose_run ps