问题
This runs locally (without specifying driver_path), but not on Heroku.
Code:
Selenium::WebDriver::Chrome.driver_path = ENV['GOOGLE_CHROME_SHIM']
browser = Watir::Browser.new :chrome
I've confirmed below values in heroku rails console
ENV['GOOGLE_CHROME_BIN'] => "/app/.apt/opt/google/chrome/chrome"
ENV['GOOGLE_CHROME_SHIM'] => "/app/.apt/usr/bin/google-chrome-stable"
Installed Buildpacks:
https://github.com/heroku/heroku-buildpack-chromedriver
https://github.com/heroku/heroku-buildpack-google-chrome
Current Error:
Selenium::WebDriver::Error::WebDriverError: unable to connect to chromedriver 127.0.0.1:9515
Searching unable to connect to chromedriver 127.0.0.1:9515 on SO returns a bunch of results but none mention heroku.
Also:
I considered phantomjs. Someone else got it working here, Using a headless browser with Heroku Rails Unicorn stack
But its deprecated. Below error when running it locally.
Selenium support for PhantomJS is deprecated. Use headless Chrome/Firefox or HTMLUnit instead.
Also Tried:
For transparency, I also tried the following.
Changing browser = Watir::Browser.new :chrome
To browser = Watir::Browser.new :chrome, headless: true
Though I didn't expect this to work.
Also Also Tried:
Removing:
https://github.com/heroku/heroku-buildpack-chromedriver
Adding:
https://github.com/heroku/heroku-buildpack-xvfb-google-chrome.
Adding headless gem.
And running the below script given in the watir gem page, http://watir.com/guides/headless/.
require 'watir'
require 'headless'
headless = Headless.new
headless.start
b = Watir::Browser.start 'www.google.com'
puts b.title
b.close
headless.destroy
Error:
Selenium::WebDriver::Error::UnknownError: unknown error: cannot find Chrome binary
I assume this failed because I didn't specify location of the chrome binary/shim. Couldn't find how to specify this when using headless in the docs.
Tried Per Suggestion:
heroku run /usr/bin/chromedriver --app app-name
Running /usr/bin/chromedriver on ⬢ app-name... up, run.2151
(Hobby) bash: /usr/bin/chromedriver: No such file or directory
Also see below logs that mention chrome when deploying to heroku:
remote: -----> chromedriver app detected
remote: -----> Looking up latest chromedriver version...
remote: -----> Downloading chromedriver v2.33...
remote: Archive: /tmp/chromedriver.zip
remote: inflating: /tmp/build_cd35072c5b766edaa2b565cbff57e5d6/.chromedriver/bin/chromedriver
remote: -----> Creating chromedriver export scripts...
remote: -----> Google Chrome app detected
...
remote: -----> Fetching https://dl.google.com/linux/direct/google-chrome-stable_current_amd64.deb
remote: -----> Installing google-chrome-stable_current_amd64.deb
...
remote: -----> Creating google-chrome shims
Some Progress:
If I ssh into the heroku server,
heroku run bash --app app-name
And search for files with the name chrome,
find /app/ -name "*chrome*"
/app/.profile.d/chromedriver.sh
/app/.profile.d/010_google-chrome.sh
/app/.apt/etc/cron.daily/google-chrome
/app/.apt/opt/google/chrome
/app/.apt/opt/google/chrome/chrome
/app/.apt/opt/google/chrome/chrome_100_percent.pak
/app/.apt/opt/google/chrome/chrome-sandbox
/app/.apt/opt/google/chrome/chrome_200_percent.pak
/app/.apt/opt/google/chrome/google-chrome
/app/.apt/opt/google/chrome/cron/google-chrome
/app/.apt/usr/bin/google-chrome-stable
/app/.apt/usr/bin/google-chrome
/app/.apt/usr/share/menu/google-chrome.menu
/app/.apt/usr/share/doc/google-chrome-stable
/app/.apt/usr/share/applications/google-chrome.desktop
/app/.apt/usr/share/gnome-control-center/default-apps/google-chrome.xml
/app/.apt/usr/share/man/man1/google-chrome.1
/app/.apt/usr/share/appdata/google-chrome.appdata.xml
/app/vendor/bundle/ruby/2.4.0/gems/selenium-webdriver-3.7.0/lib/selenium/webdriver/chrome
/app/vendor/bundle/ruby/2.4.0/gems/selenium-webdriver-3.7.0/lib/selenium/webdriver/chrome.rb
/app/vendor/bundle/ruby/2.4.0/gems/browser-2.4.0/test/unit/chrome_test.rb
/app/vendor/bundle/ruby/2.4.0/gems/browser-2.4.0/lib/browser/platform/chrome_os.rb
/app/vendor/bundle/ruby/2.4.0/gems/browser-2.4.0/lib/browser/chrome.rb
/app/.chromedriver
/app/.chromedriver/bin/chromedriver
I can see chromedriver binary at /app/.chromedriver/bin/chromedriver.
So I tried
heroku run /app/.chromedriver/bin/chromedriver --app app-name
Result:
Running /app/.chromedriver/bin/chromedriver on ⬢ app-name... up, run.2067 (Hobby)
Starting ChromeDriver 2.33.506092 (733a02544d189eeb751fe0d7ddca79a0ee28cce4) on port 9515
Only local connections are allowed.
But then running heroku run rake selenium_namespace:task_one --app app-name gives the same result.
Selenium::WebDriver::Error::WebDriverError: unable to connect to chromedriver 127.0.0.1:9515 ... /app/vendor/ruby-2.4.1/lib/ruby/2.4.0/net/http.rb:906:in `rescue in block in connect': Failed to open TCP connection to 127.0.0.1:9515 (Connection refused - connect(2) for "127.0.0.1" port 9515) (Errno::ECONNREFUSED) ...
回答1:
This is possible on Heroku.
Confusing chrome and chromedriver
Your configuration is mixing up chromedriver and Chrome. GOOGLE_CHROME_SHIM points to the Chrome executable google-chrome-stable, not to chromedriver. The line below results in Selenium executing the wrong binary, which results in the misleading error message.
Selenium::WebDriver::Chrome.driver_path = ENV['GOOGLE_CHROME_SHIM'] # WRONG!
As of writing this (Jan '18) the chromedriver build pack automatically adds /app/.chromedriver/bin to the $PATH variable. If you delete the above line Selenium should again be able to find chromedriver automatically.
And then?
You probably added the line above to fix Selenium not being able to find the Chrome binary. The error message for that would have looked something like:
Selenium::WebDriver::Error::UnknownError: unknown error: cannot find Chrome binary
You can fix this by telling Selenium where the Chrome binary is located using Selenium::WebDriver::Chrome::Options. The following code should accomplish that.
options = Selenium::WebDriver::Chrome::Options.new
chrome_bin_path = ENV.fetch('GOOGLE_CHROME_SHIM', nil)
options.binary = chrome_bin_path if chrome_bin_path # only use custom path on heroku
options.add_argument('--headless') # this may be optional
driver = Selenium::WebDriver.for :chrome, options: options
driver.navigate.to "https://stackoverflow.com"
Buildpacks
This should all be possible with the standard chrome and chromedriver build packs:
https://github.com/heroku/heroku-buildpack-google-chrome.git
https://github.com/heroku/heroku-buildpack-chromedriver.git
You may need heroku-buildpack-xvfb-google-chrome instead of vanilla chrome if you're automating clicks in the browser, but that should not be required just to get headless chrome running.
回答2:
I am quoting Ilya Vassilevsky from this post
ChromeDriver is just a driver for Chrome. It needs the actual Chrome browser installed on the same machine to actually work.
Heroku doesn't have Chrome installed on its dynos by default. You need to use a buildpack that installs Chrome. For example:
https://github.com/dwayhs/heroku-buildpack-chrome
You can see how it fetches Chrome:
https://github.com/dwayhs/heroku-buildpack-chrome/blob/master/bin/compile#L36-38
Then I read their discussion in the comments:
Petr Gazarov says
I tried this buildpack and it didn't work. I'm suspecting installing google chrome (or any browser) on heroku might be more involved.
Ilya Vassilevsky replies
Yes, Heroku is a very opinionated and closed platform. It should be much easier to set up Chrome with ChromeDriver on your own VM on AWS, Linode, or DigitalOcean.
Petr Gazarov replies
Thanks for your answer Ilya. I ended up re-writing with Watir with phantomjs because I couldn't get Heroku to install Chrome.
You can read more info in that question. If something comes to my mind, I will post it.
回答3:
I've been struggling with this one for more than a few hours ....
It has been very frustrating to try to debug on Heroku CI. This ticket has helped me a lot to find a solution.
I had this piece of code in the spec_helper.rb that was causing trouble (for if it helps someone):
config.before(:each, type: :system, js: true) do
driven_by :selenium_chrome_headless
end
It was bypassing all the capybara setup...
回答4:
Yes, I agree with Fabrizio. DigitalOcean must be used. ChromeDriver requires a Chrome browser, which creates a less malleable program and more complex coding
来源：https://stackoverflow.com/questions/47318564/heroku-unable-to-connect-to-chromedriver-127-0-0-19515-when-using-watir-seleni