QI want to run test suite collection with parallel chrome headless mode in KRE command line but it stopping for fetching driver details … can someone help
Note: It works for below test suite collection
Hi there, and thanks for posting in the Katalon community! 
To help you faster, please review our guide on Chrome Headless here:
Double-checking the steps and configurations might resolve the issue.
If the doc doesn’t help, feel free to provide more details, and a community member will assist you soon.
Thanks for being a part of our community!
Best,
Elly Tran
Could you share the error log?
No error thrown… it didn’t execute after this line
Checking WebDriverManager
Verify built-in web driver manager
Using built-in file
share the debug logs from your KRE run
To resolve issues with running KRE test suites in parallel headless Chrome mode, follow these steps:
chromedriver version matches your installed Chrome browser version.chromedriver is correctly configured in KRE’s settings.npm update -g kre
selenium-webdriver, chromedriver).Explicitly define headless Chrome options in your KRE configuration file (e.g., kre.config.js):
const { Options } = require('selenium-webdriver/chrome');
module.exports = {
browsers: {
chrome_headless: {
browserName: 'chrome',
headless: true,
chromeOptions: new Options()
.addArguments('--headless=new') // Use new headless mode
.addArguments('--no-sandbox')
.addArguments('--disable-dev-shm-usage')
.addArguments('--disable-gpu')
}
},
parallel: true, // Enable parallel execution
maxParallelThreads: 4 // Set based on your system's CPU cores
};
module.exports = {
maxParallelThreads: 2 // Start with a lower number
};
kre run --config ./kre.config.js --log-level=debug
WebDriverError: unknown error: DevToolsActivePort file doesn't exist (add --remote-debugging-port=9222 to Chrome arguments).Unable to discover open pages (ensure Chrome flags are configured correctly).kre run --browsers=chrome_headless --parallel --workers=2
--headless=new with --headless=chrome./usr/bin/google-chrome on Linux).Let me ask you a question.
In the Katalon documentation, the word “parallel” means 2 different features.
These 2 features use a single word “parallel” but these are completely different. You shouldn’t confuse them.
Which “parallel” do you mean?
I guess, you mean the #1 — Parallel mode execution of a single Test Suite Collection. Am I right?
As far as I see, Katalon designed the #1 “executing Test Suite Collection in parallel mode” poorly. It brought a lot of problems to me. So I am determined not to touch it. Instead, I just prefer the sequential mode — it works stable, as you already noticed.
Yes.. I mean the #1 in KRE Command line execution
For the #1 at least 4instances should work right
Or
Any maximum instance suggestion…
I don’t know, as I do not use #1, and I have no KRE licenses.
I could execute a Test Suite Collection with parallel mode using a single license of Katalon Studio GUI Free while specifiying 3 as the maxium parallel processes. ---- though I specified “parallel” but actually the TSC ran slowly, so that no quicker than the sequential mode.
This proves that Test Suite Collection with parallel mode requires only 1 license of Katalon Studio GUI Free.
Based on this observation, I guess that a execution of Test Suite Collection with parallel mode requires only 1 active license. Rather, I believe that the number of licenses is not checked at all in the case of #1 parallel mode.
I guess that the number of KRE license has nothing to do with your current problem.
As far as I see, the “Test Suite Collection in parallel mode” is unstable, problematic. I would recommend you not to touch it. It would waste your time.
If you are eager for “parallelism”, you should look at the #2 parallelism. Actually I do not know it at all, but as far as I read the Katalon’s doc, #2 seems to be better (better tested possibly) than #1.
In the #2, if you want 4 TestOps Agents to work in parallel, I believe that you need at least 4 KRE licenses. Simple, isn’t it?
Parallel execution of bulky Test Suites on a single machine won’t run any quicker than a sequential execution. On a single box, even if you tell a Test Suite Collection to execute multiple Test Suites in parallel, in fact the test suites run nearly sequentially. Why? Each individual Test Suite makes the machine fully occupied. So OS will allocate the CPUs+memory resources just sequentially to the Test Suites one after the other.
If @vasanthi.jayakumar wants process the bulk of Test Suites more quickly, he definitely needs to allocate multiple machines. Each machines should equip enough CPUs and memory. And he wants to distribute the bulky Test Suites into the nodes; monitor the distributed execution; collect the results and compile a final report. The more number of machines, the quicker the processing duration could be.
On the Internet, I could find some articles how to implement a distributed processing of Selenium tests. For example, with AWS, we can distribute selenium tests over mutliple nodes instanciated using ECS Fargate:
Browserstack publishes the following article:
How can we do the similar distributed processing using Katalon? — I don’t know. I have never seen an article published by Katalon about this issue.
Input by Katalon people is desired.
→ @Elly_Tran , @Shin, @vu.tran
Then, you would want to purchace at least 4 KRE licenses to implement the #2 parallelism. Can you afford of it?
Beware, a license of KRE is payable monthly for fixed amount of money regardless how long seconds/minutes/hours you actually used. It is not a pay-as-you-go system.
Hi all,
Thank you for sharing the issue. I will create the ticket for my team to investigate. Back to you soon
I think that the parallel execution of Test Suites is not the first approach you should try for faster and stable test exectuion. Seeking for parallelism would trigger difficult troubles and would waste your time.
Rather, I would recommend you to rewrite each individual test cases so that they run faster and more stable.
For example, if you use “WebUI.delay(n)” keyword in your test cases, you should avoid it. You should replace “WebUI.delay(n)” with “WebUI.waitForElementVisible(testobject, n)”. Why? “WebUI.delay(10)” will always delay your test case for 10 seconds. On the other hand, “WebUI.waitForElementVisible(testobject, 10)” is likely to pass in just 1 seconds. Avoiding “WebUI.delay” will certainly result a faster execution.
You should check the execution log line by line and identify which statement takes long seconds. You should try to find out a way to make it run quickly.
In my humble opinion, there is no automagical way that makes your tests run faster. You need to refactor your test cases one by one. You need to devote yourself for tuning every single test cases to make it run faster and more stable.
When you finished refining all your test cases, a sequential execution of your bulky test suites might run in an acceptable duration. You do not need to pay more money for this approach.
Once you have done fine-tuning your test cases, it would be an idea to try the parallelism #2 for even faster execution while paying for additional multiple machines and additional KRE licenses.
Hi,
I got the answer back from my team that If we want to run faster, we should setup Selenium Grid to distribute the test to other machines rather run all tests on the same machine:
@vasanthi.jayakumar, please follow up with all the suggestions above. Let us know if any works for you. Otherwise, after 2 weeks no reply from you, I will process to close this. Thank you
??? I believe that Selenium Grid has nothing to do with the “parallel” execution. I am afraid that the suggestion from the Katalon team does not help @vasanthi.jayakumar. It rather confuses him.
The following session might be a response from Katalon.