How to autoupdate chromeDriver & geckDriver in selenium

Question

I have selenium grid setup with multiple node machines , where in I am manually downloading chromeDriver & geckoDriver on all selenium node machines & using them for chrome & firefox browsers respectively.

Now here chrome & firefox browsers (on all selenium node machines) are set on 'Automatic Update' (which is required as I want my application to be tested always on latest browser versions) , because of this browsers on my node machines keep getting updated more often & since respective driver updates is a manual process , it forces me to log in to each selenium node machine & update them manually.

Can this process be automated ?

PS : I know that dockerized selenium grid can be used to fetch/pull latest browser images & their drivers , however switching from traditional selenium grid to dockerized selenium grid is another thing & will take some time to implement.

Gipsy

Not sure what approach you finally took, I am in a similar situation and found a library that manages the driver automatically: github.com/bonigarcia/webdrivermanager . Am yet to explore this approach, but wanted to send a note to you if it helps.. – Superelevation

Answer 1

First @Asyranok is right, even when implemented auto update code will not work 100% of the time. However, for many of us, this occasional downtime is "ok" as long as it's just a few days.

I've found that manually updating X servers every few months to be incredibly irritating and while there are well written instructions on the selenium website on how to "auto update" the driver I've yet to see one openly available non-library implementation of this guide.

My answer is specific to C#, for this language the solution typically suggested is to use NuGet to pull the latest driver automatically, this has two issues:

1. You need to deploy at the frequency of chrome updating (most companies aren't there yet, neither are we) or your application will be "broken" for the time between chrome updating and your "new" version of the application deploying, and again this is only if you release on a schedule, if you release ad-hoc your going to have to go through a series of manual steps to update, build, release etc. to get your application working again.

2. You need (typically, without a work around) to pull the latest chromedrive from NuGet by hand, again a manual process.

What would be nice would be what python has and @leminhnguyenHUST suggests which is using a library that will automatically pull the latest chromedriver on runtime. I've looked around and haven't yet found anything for C# that does this, so I decided to roll my own and build that into my application:

public void DownloadLatestVersionOfChromeDriver()
{
    string path = DownloadLatestVersionOfChromeDriverGetVersionPath();
    var version = DownloadLatestVersionOfChromeDriverGetChromeVersion(path);
    var urlToDownload = DownloadLatestVersionOfChromeDriverGetURLToDownload(version);
    DownloadLatestVersionOfChromeDriverKillAllChromeDriverProcesses();
    DownloadLatestVersionOfChromeDriverDownloadNewVersionOfChrome(urlToDownload);
}

public string DownloadLatestVersionOfChromeDriverGetVersionPath()
{
    //Path originates from here: https://chromedriver.chromium.org/downloads/version-selection            
    using (RegistryKey key = Registry.LocalMachine.OpenSubKey("SOFTWARE\\Microsoft\\Windows\\CurrentVersion\\App Paths\\chrome.exe"))
    {
        if (key != null)
        {
            Object o = key.GetValue("");
            if (!String.IsNullOrEmpty(o.ToString()))
            {
                return o.ToString();
            }
            else
            {
                throw new ArgumentException("Unable to get version because chrome registry value was null");
            }
        }
        else
        {
            throw new ArgumentException("Unable to get version because chrome registry path was null");
        }
    }
}

public string DownloadLatestVersionOfChromeDriverGetChromeVersion(string productVersionPath)
{
    if (String.IsNullOrEmpty(productVersionPath))
    {
        throw new ArgumentException("Unable to get version because path is empty");
    }

    if (!File.Exists(productVersionPath))
    {
        throw new FileNotFoundException("Unable to get version because path specifies a file that does not exists");
    }

    var versionInfo = FileVersionInfo.GetVersionInfo(productVersionPath);
    if (versionInfo != null && !String.IsNullOrEmpty(versionInfo.FileVersion))
    {
        return versionInfo.FileVersion;
    }
    else
    {
        throw new ArgumentException("Unable to get version from path because the version is either null or empty: " + productVersionPath);
    }
}

public string DownloadLatestVersionOfChromeDriverGetURLToDownload(string version)
{
    if (String.IsNullOrEmpty(version))
    {
        throw new ArgumentException("Unable to get url because version is empty");
    }

    //URL's originates from here: https://chromedriver.chromium.org/downloads/version-selection
    string html = string.Empty;
    string urlToPathLocation = @"https://chromedriver.storage.googleapis.com/LATEST_RELEASE_" + String.Join(".", version.Split('.').Take(3));

    HttpWebRequest request = (HttpWebRequest)WebRequest.Create(urlToPathLocation);
    request.AutomaticDecompression = DecompressionMethods.GZip;

    using (HttpWebResponse response = (HttpWebResponse)request.GetResponse())
    using (Stream stream = response.GetResponseStream())
    using (StreamReader reader = new StreamReader(stream))
    {
        html = reader.ReadToEnd();
    }

    if (String.IsNullOrEmpty(html))
    {
        throw new WebException("Unable to get version path from website");
    }

    return "https://chromedriver.storage.googleapis.com/" + html + "/chromedriver_win32.zip";
}

public void DownloadLatestVersionOfChromeDriverKillAllChromeDriverProcesses()
{
    //It's important to kill all processes before attempting to replace the chrome driver, because if you do not you may still have file locks left over
    var processes = Process.GetProcessesByName("chromedriver");
    foreach (var process in processes)
    {
        try
        {
            process.Kill();
        }
        catch
        {
            //We do our best here but if another user account is running the chrome driver we may not be able to kill it unless we run from a elevated user account + various other reasons we don't care about
        }
    }
}

public void DownloadLatestVersionOfChromeDriverDownloadNewVersionOfChrome(string urlToDownload)
{
    if (String.IsNullOrEmpty(urlToDownload))
    {
        throw new ArgumentException("Unable to get url because urlToDownload is empty");
    }

    //Downloaded files always come as a zip, we need to do a bit of switching around to get everything in the right place
    using (var client = new WebClient())
    {
        if (File.Exists(System.IO.Path.GetDirectoryName(Assembly.GetEntryAssembly().Location) + "\\chromedriver.zip"))
        {
            File.Delete(System.IO.Path.GetDirectoryName(Assembly.GetEntryAssembly().Location) + "\\chromedriver.zip");
        }

        client.DownloadFile(urlToDownload, "chromedriver.zip");

        if (File.Exists(System.IO.Path.GetDirectoryName(Assembly.GetEntryAssembly().Location) + "\\chromedriver.zip") && File.Exists(System.IO.Path.GetDirectoryName(Assembly.GetEntryAssembly().Location) + "\\chromedriver.exe"))
        {
            File.Delete(System.IO.Path.GetDirectoryName(Assembly.GetEntryAssembly().Location) + "\\chromedriver.exe");
        }

        if (File.Exists(System.IO.Path.GetDirectoryName(Assembly.GetEntryAssembly().Location) + "\\chromedriver.zip"))
        {
            System.IO.Compression.ZipFile.ExtractToDirectory(System.IO.Path.GetDirectoryName(Assembly.GetEntryAssembly().Location) + "\\chromedriver.zip", System.IO.Path.GetDirectoryName(Assembly.GetEntryAssembly().Location));
        }
    }
}

Then usually I'll stick this very hacky invocation at the beginning of my application to invoke this feature and ensure that the latest chromedriver is available for my application:

//This is a very poor way of determining if I "need" to update the chromedriver,     
//however I've yet to figure out a better way of doing this...
try
{
    using (var chromeDriver = SetupChromeDriver())
    {
        chromeDriver.Navigate().GoToUrl("www.google.com");
        chromeDriver.Quit();
    }
}
catch
{
    DownloadLatestVersionOfChromeDriver();
}

I'm sure this could be improved significantly, but so far it's worked for me.

Note: Cross Posted Here

Interstate

THIS WORKED FOR ME LIKE A CHARM. THANKS. – Literary

Thank you David handy easy solution. This will work till the chromedriver version 114. I found one solution at which uses github.com/GoogleChromeLabs/chrome-for-testing and download the binaries. github.com/Swimburger/DownloadChromeDriverSample – Constantino

Answer 2

I would argue that your current approach is not a feasible approach. New versions of browsers are released with zero consideration for Selenium (or any other drivers). As soon as a new browser update is released, there is a reasonably high chance that there will be NO existing driver that works with that version. It often takes days for Selenium teams to release updated drivers to match the newest version of a browser.

And since you are automatically updating your browsers, then you are possibly automatically breaking your Selenium tests until a new driver version is released, or until you downgrade the browsers.

Now, you may be fine with that, and are okay with disabling a browser's tests until the most current Selenium drivers work with the most current browser version. If that is the case, then here are some solutions:

1) If you are using C#, store your drivers in the test solution as a Nuget package, or in a dependencies folder. Then, have the automation reference that driver no matter where it is running. When you need to update the driver, you literally only need to update it in one place, and check in the changes. All client machines will, through your CI process, pull down the latest code, which includes that new driver.

2) If for some reason you do not want the driver in your project as a Nuget package or a manually-saved dependency, then have your CI handle the update process. Point your automation code to a driver located in some common directory on whatever client machine it is currently running on -> wherever your machine stores the dependencies after downloading them. For example; downloading selenium files via console on a Windows machine will put them somewhere in %APPDATA% "C:\Users\xxxxxx\AppData\Roaming\npm\node_modules". That is where your test solution should look.

Then, in your CI scripts, before running any tests, download the latest driver. The syntax is nearly the same, if not identical between Windows and Linux/Unix Kernels. This assumes you have npm installed.

npm install -g selenium

If you already have latest, then nothing will happen. If you don't the latest driver will be downloaded by your CI script before running tests. Then, your test solution will be pointing to where the driver is stored on the client, and it will automatically be using the newest driver.

Epistemic

Thanks for elaborated answer. Appreciate your time. Here I am running my tests through jenkins , so architecture/flow is like jenkins (triggering selenium suite) -> jenkins slave machine(pulling qa repo containing tests , running them through gradle/maven) -> selenium hub machine -> selenium nodes machines(having geckodriver/chromedriver) Considering above architecture in mind , let me ponder on above solutions, – Gipsy

#1 suggests packaging drivers(chrome/gecko) within tests repo , if i do so it will stay at my jenkins slave machie as jenkins slave is the machine which pulls qa repo , however I want those drivers to be existing on selenium node machines. So packaging drivers with git/nuget manager won't help in my case (However keeping those drivers at some shared folder, which can be accessible by all selenium nodes & then updating drivers under shared folder manually once new drivers are released will work , but still it will be manual action) – Gipsy

#2 suggests using NPM to update selenium/chromedriver/geckodriver (like npm install chromedriver / npm install geckodriver) before running tests through jenkins. I have to run these commands specifically on selenium nodes by making them as jenkins slave (i.e selenium node machine should act as jenkins slave so that I can execute commands on them), however NPM would be the prerequisite to be installed on each selenium node machine. Well I will try both of these options & will see what suits me more. – Gipsy

Answer 3

I know its a bit of an old question, but I thought this could also be helpful to people: https://github.com/rosolko/WebDriverManager.Net. Its a Nuget package (https://www.nuget.org/packages/WebDriverManager/) that appears to solve the problem using .NET.

Ind

Perfekt solution and easy to use. Thank you so much! Just add: new DriverManager().SetUpDriver(new EdgeConfig(), VersionResolveStrategy.MatchingBrowser); – Brilliant

Answer 4

One option for Java based solution is the Bonigarcia Webdrivermanager. Seems like it has inbuilt support for remote web driver, using this drivermanager is able to download the latest browsers on the Hub of the selenium grid . Check this post.

Kipp

Answer 5

One possible solution - write a script that downloads the latest Webdriver version and schedule its daily launch.

For example, here is Python 3 script that downloads the latest Chromedriver for Linux of version that matches currently installed Chrome browser version

import os
import re
import requests
import zipfile

XML_INFO_URL = 'https://chromedriver.storage.googleapis.com/LATEST_RELEASE_'
DRIVER_PATH = 'https://chromedriver.storage.googleapis.com'
ZIPPED_DRIVER_FILE_NAME = 'chromedriver_linux64.zip'
UNZIPPED_DRIVER_FILE_NAME = 'chromedriver'

PATH_TO_CHROMEDRIVER = '.'

def get_browser_major_version():
    stream = os.popen('google-chrome --version')
    output = stream.read()
    version_info_str = re.search(r'\d+\.\d+\.\d+', output).group(0)
    return re.search(r'^\d+', version_info_str).group(0)


def get_driver_latest_version(browser_major_version):
    return requests.get(XML_INFO_URL + browser_major_version).text


def download_file(url, file_name):
    file = requests.get(url)
    with open(file_name, "wb") as code:
        code.write(file.content)


driver_file_url = DRIVER_PATH + '/' + get_driver_latest_version(get_browser_major_version()) + '/' + ZIPPED_DRIVER_FILE_NAME
download_file(driver_file_url, ZIPPED_DRIVER_FILE_NAME)

with zipfile.ZipFile(ZIPPED_DRIVER_FILE_NAME, 'r') as zip_ref:
    zip_ref.extractall(PATH_TO_CHROMEDRIVER)

os.chmod(PATH_TO_CHROMEDRIVER + '/' + UNZIPPED_DRIVER_FILE_NAME, 0o744)

os.remove(ZIPPED_DRIVER_FILE_NAME)

Fitch

Answer 6

Now the problem has been solved with webdrivermanager module from here

Description:

Python module to facilitate downloading and deploying WebDriver binaries. The classes in this module can be used to automatically search for and download the latest version (or a specific version) of a WebDriver binary and then extract it and place it by copying or symlinking it to the location where Selenium or other tools should be able to find it then.

Hope it helps !!!

Tva

Answer 7

I had the same situation. I came up with the below solution to automatically update the driver by checking chrome compatibility,

update_chromedriver.py

import zipfile

from bs4 import BeautifulSoup
import requests
import re
from selenium import webdriver
from selenium.common.exceptions import SessionNotCreatedException
from selenium.webdriver.chrome.options import Options

downloads_url = "https://chromedriver.chromium.org/downloads"
basic_downloads_path = 'https://chromedriver.storage.googleapis.com/{0}chromedriver_{1}.zip'
options = Options()
options.add_argument('--headless')
options.add_argument('--disable-gpu')

# os='win32'
os = 'linux64'
driver_executable = './driver/chromedriver'


def check_browser():
    driver = None
    try:
        driver = webdriver.Chrome(executable_path=driver_executable, options=options)
        driver.get('https://www.google.com/')
        return True
    except SessionNotCreatedException:
        return False
    finally:
        if driver is not None:
            driver.close()


def download_zip(_url):
    r = requests.get(_url, allow_redirects=True)
    open('./driver/chrome.zip', 'wb').write(r.content)
    with zipfile.ZipFile('./driver/chrome.zip') as z:
        z.extractall('./driver/')


def get_versions():
    html = requests.get(downloads_url).text
    soup = BeautifulSoup(html, "html.parser")
    versions_a = soup.find_all('a', href=re.compile('https://chromedriver.storage.googleapis.com/index.html\\?path='))
    versions = []
    for a in versions_a:
        version = a['href'].split('=', 1)[1]
        versions.append(version)
    return versions


def auto_update_driver():
    print('====> Checking current Chrome compatibility.')
    compatible = check_browser()
    print('====> It is ' + ('Compatible.' if compatible else 'Not Compatible.'))
    if not compatible:
        print('====> Trying to find alternate versions.')
        versions = get_versions()
        print(f'==> ({len(versions)}) found.')
        for version in versions:
            url = basic_downloads_path.format(version, os)
            download_zip(url)
            compatible = check_browser()
            print(f'==> Trying version {version} - {"Compatible" if compatible else "Not Compatible"}')
            if compatible:
                break
        if not compatible:
            raise Exception('Unable to upgrade chrome version. Probably you are using very old chrome version.'
                            'Please update chrome and try again!!')

Please note :

1. My chrome driver is in the current path ./driver/
2. This is in my ubuntu machine. Please change accordingly to your requirements

Usage of the above python script

from update_chromedriver import auto_update_driver

auto_update_driver()

Strick

Answer 8

you can use ansible or puppet to take update package on all of the node

Hug

If you could explain better, how to update chrome and firefox in a step-by-step way. Please review this question about how to write a good answer. stackoverflow.com/help/how-to-answer – Calvillo