在负载下运行Selenium WebDriver测试

Question

I have bunch of functional tests for a web application which were written using Java, Selenium WebDriver and TestNG. Currently my goal is to measure the time of each method in those tests under the load(for instance with 10-15 users working simultaneously). Did anyone face such problem and maybe has an idea how to do it? Or it is impossible to use WebDriver in case of load tests?

I was searching for resolution with jMeter + jUnit as jUnitRequest(rewriting my tests from TestNG to jUnit won't be difficult) but as far as I know jUnit doesn't support such tests when one method depends on some other(TestNG support this) and I have to merge all methods to only one(which imho is stupid idea).

Any hint how to handle this problem will be highly appreciated.

Answer 1

It's not that hard to do it with TestNG. But in order to run you tests in parallel, you'll have to manage your WebDriver with your threads. To do that, I would create a class with a ThreadLocal. With the TheadLocal class, you can easily manage your objects per thread. There is an example of such a class :

public class DriverManager {

    private static ThreadLocal<WebDriver> driver = new ThreadLocal<WebDriver>();

    public static WebDriver getDriver() {
        return driver.get();
    }

    public static void setWebDriver(WebDriver driver) {
        driver.set(driver);
    }
}

Whenever you have to use your driver, you just have to call it like that :

DriverManager.getDriver().doSomething

And then, with testNG, you can execute your tests in different threads.

Answer 2

I would recommend not doing load testing with selenium. Load testing with a browser does not make sense. Testing client side performance should only test performance with 1/2 browsers running as a client will only be running 1/2 instances of your application. They are never going to have 10-15 browsers open and running against your application. If you want to load test server side code you should use Jmeter and hit the API with users in parallel to determine how many parallel operations your server can handle before the server fails to respond.