EC2实例中ChromeDriver调用Google Chrome时启动失败问题排查

在Ubuntu 18.04.4 LTS的EC2测试服务器上，将chromedriver与Google Chrome同步更新至109.0.5414.74版本（版本匹配），受限于Ruby 2.3.6，同步升级selenium-webdriver至3.142.7。原有代码新增Service实例后，运行时出现Chrome异常退出错误，提示DevToolsActivePort file doesn't exist，添加--disable-dev-shm-usage等参数后问题仍未解决。

• selenium-webdriver: 3.142.7
• capybara: 2.7.1
• 操作系统: Ubuntu 18.04.4 LTS
• 内核: Linux 5.4.0-1093-aws
• 架构: x86-64

终端版本验证输出：

$ which chromedriver
/usr/bin/chromedriver
$ which google-chrome
/usr/bin/google-chrome
$ chromedriver --version
ChromeDriver 109.0.5414.74 (e7c5703604daa9cc128ccf5a5d3e993513758913-refs/branch-heads/5414@{#1172})
$ google-chrome --version
Google Chrome 109.0.5414.74

def set_defaults
  options = set_options
  capabilities = Selenium::WebDriver::Remote::Capabilities.chrome(chromeOptions: options)
  service = Selenium::WebDriver::Service.chrome(path: '/usr/bin/chromedriver')
  selenium_options = { browser: :chrome, desired_capabilities: capabilities, service: service }
  Capybara.register_driver :chrome do |app|
    Capybara::Selenium::Driver.new(app, selenium_options)
  end
  Capybara.javascript_driver = :chrome
end

def set_options
  user_agent = 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10.10; rv:36.0) Gecko/20100101 Firefox/36.0'
  options = Selenium::WebDriver::Chrome::Options.new
  options.add_argument('--no-sandbox')
  options.add_argument('--headless')
  options.add_argument("--user-agent='#{user_agent}'")
  options.add_argument('--disable-popup-blocking')
  options.add_argument('--disable-gpu')
  options.add_argument('--window-size=1280,1024')
  options.add_argument('--disable-dev-shm-usage') # 参考其他回答添加，但无效
  options
end

Selenium::WebDriver::Error::UnknownError: unknown error: Chrome failed to start: exited abnormally.
  (unknown error: DevToolsActivePort file doesn't exist)
  (The process started from chrome location /usr/bin/google-chrome is no longer running, so ChromeDriver is assuming that Chrome has crashed.)
    from #0 0x55b693029303 <unknown>
    ...
    from script/rails:21:in `<main>'

1. 修正User-Agent参数格式
   当前代码中--user-agent参数被多余的单引号包裹，导致Chrome无法正确解析，进而触发启动异常。修改set_options中的对应行：

   # 原代码
   options.add_argument("--user-agent='#{user_agent}'")
   # 修改为
   options.add_argument("--user-agent=#{user_agent}")
   

2. 优化Chrome启动参数组合
   针对服务器headless环境，补充调整参数避免端口分配异常：

   def set_options
     user_agent = 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10.10; rv:36.0) Gecko/20100101 Firefox/36.0'
     options = Selenium::WebDriver::Chrome::Options.new
     options.add_argument('--no-sandbox')
     options.add_argument('--headless=new') # Chrome 109+推荐使用新版headless模式
     options.add_argument("--user-agent=#{user_agent}")
     options.add_argument('--disable-popup-blocking')
     options.add_argument('--disable-gpu')
     options.add_argument('--window-size=1280,1024')
     options.add_argument('--disable-dev-shm-usage')
     options.add_argument('--remote-debugging-port=9222') # 指定固定调试端口
     options.add_argument('--single-process') # 强制单进程运行适配服务器环境
     options
   end
   

3. 检查文件权限与依赖

   • 确保运行代码的用户对Chrome和chromedriver有执行权限：

     sudo chmod +x /usr/bin/chromedriver /usr/bin/google-chrome
     

   • 安装Chrome运行必需的依赖库：

     sudo apt-get install -y libnss3 libgconf-2-4 libxss1 libappindicator1 libindicator7
     

4. 适配旧版Capybara的驱动注册方式
   由于Capybara 2.7.1版本较旧，直接传递service参数可能存在兼容性问题，改为手动初始化驱动：

   def set_defaults
     Capybara.register_driver :chrome do |app|
       options = set_options
       driver = Selenium::WebDriver.for :chrome, options: options, service: Selenium::WebDriver::Service.chrome(path: '/usr/bin/chromedriver')
       Capybara::Selenium::Driver.new(app, browser: :chrome, driver: driver)
     end
     Capybara.javascript_driver = :chrome
   end
   

内容的提问来源于stack exchange，提问作者Magpie