如何在GitHub Actions中当某一并行Job失败时终止其他Job？

GitHub Actions本身没有内置的跨Job终止机制，要实现「任一并行Job失败时立即终止另一个」的需求，需要借助GitHub REST API来主动取消运行中的Job，具体配置方案如下：

完整配置示例 #

name: Parallel Test Workflow
on: [push, pull_request]

# 必须赋予取消Job的权限
permissions:
  actions: write

jobs:
  Unit:
    runs-on: ubuntu-latest
    steps:
      - name: Checkout code
        uses: actions/checkout@v4

      - name: 后台监控Integration Job状态
        run: |
          while true; do
            # 获取当前Workflow运行ID
            WORKFLOW_RUN_ID=${{ github.run_id }}
            # 查询Integration Job的状态和结果
            INTEGRATION_JOB_INFO=$(gh api repos/${{ github.repository }}/actions/runs/$WORKFLOW_RUN_ID/jobs | jq -r '.jobs[] | select(.name == "Integration") | "\(.status)|\(.conclusion)"')
            STATUS=$(echo "$INTEGRATION_JOB_INFO" | cut -d'|' -f1)
            CONCLUSION=$(echo "$INTEGRATION_JOB_INFO" | cut -d'|' -f2)
            
            # 若Integration Job已失败，立即终止当前Job
            if [ "$STATUS" = "completed" ] && [ "$CONCLUSION" = "failure" ]; then
              echo "Integration测试已失败，终止Unit测试..."
              exit 1
            fi
            sleep 10
          done &
        env:
          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

      - name: 运行单元测试
        run: |
          # 替换成你的实际单元测试命令，比如 pytest、npm test 等
          npm test

      - name: 若Unit测试失败，取消Integration Job
        if: failure()
        run: |
          WORKFLOW_RUN_ID=${{ github.run_id }}
          # 获取Integration Job的ID
          INTEGRATION_JOB_ID=$(gh api repos/${{ github.repository }}/actions/runs/$WORKFLOW_RUN_ID/jobs | jq -r '.jobs[] | select(.name == "Integration") | .id')
          # 取消目标Job（仅当Job存在时执行）
          if [ -n "$INTEGRATION_JOB_ID" ]; then
            gh api --method POST repos/${{ github.repository }}/actions/jobs/$INTEGRATION_JOB_ID/cancel
            echo "已取消运行中的Integration Job"
          fi
        env:
          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

  Integration:
    runs-on: ubuntu-latest
    steps:
      - name: Checkout code
        uses: actions/checkout@v4

      - name: 后台监控Unit Job状态
        run: |
          while true; do
            WORKFLOW_RUN_ID=${{ github.run_id }}
            UNIT_JOB_INFO=$(gh api repos/${{ github.repository }}/actions/runs/$WORKFLOW_RUN_ID/jobs | jq -r '.jobs[] | select(.name == "Unit") | "\(.status)|\(.conclusion)"')
            STATUS=$(echo "$UNIT_JOB_INFO" | cut -d'|' -f1)
            CONCLUSION=$(echo "$UNIT_JOB_INFO" | cut -d'|' -f2)
            
            if [ "$STATUS" = "completed" ] && [ "$CONCLUSION" = "failure" ]; then
              echo "Unit测试已失败，终止Integration测试..."
              exit 1
            fi
            sleep 10
          done &
        env:
          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

      - name: 运行集成测试
        run: |
          # 替换成你的实际集成测试命令
          npm run test:integration

      - name: 若Integration测试失败，取消Unit Job
        if: failure()
        run: |
          WORKFLOW_RUN_ID=${{ github.run_id }}
          UNIT_JOB_ID=$(gh api repos/${{ github.repository }}/actions/runs/$WORKFLOW_RUN_ID/jobs | jq -r '.jobs[] | select(.name == "Unit") | .id')
          if [ -n "$UNIT_JOB_ID" ]; then
            gh api --method POST repos/${{ github.repository }}/actions/jobs/$UNIT_JOB_ID/cancel
            echo "已取消运行中的Unit Job"
          fi
        env:
          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

关键说明 #

1. 权限配置
   必须在Workflow开头添加permissions: actions: write，因为取消Job需要修改GitHub Actions的运行状态，默认的GITHUB_TOKEN没有该权限，需显式赋予。

2. 双向监控与取消

   • 每个Job启动一个后台进程，每隔10秒查询另一个Job的状态，一旦发现对方失败就立即终止自身。
   • 当当前Job失败时，执行取消另一个Job的操作，避免对方继续占用资源。

3. 工具依赖
   Ubuntu系统的GitHub Actions Runner默认已预装gh（GitHub CLI）和jq（JSON解析工具），无需额外安装。

4. 注意事项

   • 确保Job名称与脚本中select(.name == "XXX")的名称完全一致，否则无法匹配到目标Job。
   • API状态同步存在几秒延迟，无法做到完全实时终止，但能有效减少不必要的资源浪费。

内容的提问来源于stack exchange，提问作者margherita pizza