Parallel Processing with Nightwatch and ChromeDriver without Selenium

[cortezcristian]

Is it possible to use this framework without java and selenium?
Using nightwatch parallel auto feature directly avoiding selenium? Or is selenium a core component of Magellan?

[archlichking]

Nope, selenium isn't a must to the magellan eco-system. There is an option to disable selenium auto launch in nightwatch.json. Similarly like http://testarmada.io/#manageAppiumLifecycle, selenium.start_process can be false so that nightwatch won't automatically launch selenium server for you.

[cortezcristian]

I'm running a local executor, if i set start_process: false, obviously it start failing.

Where should be the place to start the chromedriver server? I mean calling chromedriver.start/stop in magellan? This is from the official guide: http://nightwatchjs.org/gettingstarted under ChromeDriver section 3.

Because i added the start/stop on the before/after hooks on the lib/globals.js file and the chrome instances don't show up.

Changed the port to be 9515 instead of 4444, but still failing.

[archlichking]

i'm afraid you need to follow all the instructions especially this one

We also need to clear the default_path_prefix, as it is set by default to /wd/hub, which is what selenium is using.

with all the changes required by the nightwatch, test works fine in my local without selenium server. the entry i use for chrome

"chrome": {
      "selenium_port"  : 9515,
      "selenium_host"  : "localhost",
      "default_path_prefix" : "",
      "desiredCapabilities": {
        "browserName": "chrome"
      },
      "selenium": {
        "start_process": false
      }
    },

[cortezcristian]

I did set "default_path_prefix": "", to be empty under test_settings/default. Still failing

[archlichking]

when you see error retrieving a new session ..., it means nightwatch is still looking for selenium server. you may still have selenium.start_process enabled somewhere. use --debug to have local executor to output the actual desiredCapabilities.

[cortezcristian]

I did that:

[DEBUG][Local Executor] executor config: {
	"desiredCapabilities": {
		"browserName": "chrome",
		"chromeOptions": {
			"args": ["--no-sandbox", "start-maximized", "disable-web-security"],
			"prefs": {
				"credentials_enable_service": false,
				"profile.password_manager_enabled": false,
				"profile.managed_default_content_settings.images": 2
			}
		},
		"acceptSslCerts": true
	}
}

That's it the default settings were being overrided

It does work now, but i found another issue. Every time the cases run starts a new instance of chrome driver on a different port.

And they end all using the same one on port 9515. How can I attach the different set of workers to those separate intances of chromedriver? (I guess that's not possible)

The only solution that worked for me, was to run chromedriver as a single process before starting the suite, to avoid the the problem of multiple chromedriver instances in that scenario (when the test run in parallel).