Selenium: Chromedriver leaves forked Chrome instances hanging in background after test run (with large CPU load)

[garrymann]

Meta

OS:
Microsoft Windows [Version 10.0.16299.248] x64
Selenium Version:
2.53.1
Browser:
Chrome 65 x64
Browser Version:
Version 65.0.3325.146 (Official Build) (64-bit)
Additional rerequisites:

• ChromeDriver (2.36)
• NodeJS (v8.10.0)
• [email protected]
• [email protected]

Expected Behavior -

1. The chromedriver.exe should stop and its Chrome instances as well.

Actual Behavior -

1. The chromedriver.exe stops 4-5 secs after the tests completed. Then the forked Chrome instances stay in the background as running processes with huge CPU usage.

Steps to reproduce -

1. Run at least 1 failing test with CucumberJS and Protractor.
   note: I'm running e2e tests against an Angular 1.6 application
2. Examine the processes in Task manager.

Note:

This issue was introduced by the latest Chrome and ChromeDriver update. No symptomps were present in previous versions.

Screenshot:

[cgoldberg]

Run at least 1 failing test with CucumberJS and Protractor.

does your code always call quit() on the driver instance when a test fails? (usually done as a teardown/cleanup fixture in your framework)

[garrymann]

@cgoldberg Hi!
The .quit() is handled by Protractor as far as I know. So it is not necessary since runner teardownEnv will take care of that for you.

[barancev]

If you (or Protractor) called quit, but chromedriver could not stop the browser -- report this issue to chromedriver: https://sites.google.com/a/chromium.org/chromedriver/help

[garrymann]

@barancev Done. Thanks.
https://bugs.chromium.org/p/chromedriver/issues/detail?id=2311

Things I've tried but did not resolve the issue:
reinstall chrome (v65 x64)
disable and remove extensions
run it without a single chrome instance open in the background
delete npm-cache
redo npm i
update to npm 5.7 (latest)
update to node 8 (latest LTS)
update windows 10: Version 10.0.16299.248

Any other ideas?

[cgoldberg]

it is not necessary since runner teardownEnv
will take care of that for you

and what happens when something crashes unexpectedly and your teardown isn't called?

[garrymann]

tried to call the quit manually as a hook but it can't reach the instance by this time:
this.registerHandler('AfterFeatures', () => { logger.warn('Force closing ChromeDriver.'); return browser.driver.quit(); });

So in conclusion nothing happens. After trying many workarounds (and sticking to the latest versions) I guess Google Chrome needs to fix stuff with Chrome and Chromedriver then.

Thanks anyway!

[farmazonne]

FYI: Don't know why but deleting arg "--no-sandbox" fixed the issue for me.

[nolmit]

removing "no-sandbox" helped for me too

[Vishwa1]

I was facing the same issue and was so annoyed with it. Had to always keep an eye on CPU usage and kill the process. However, removing "no-sandbox" resolved my problem too.

[elsticky]

Saw the same problem, also removed --no-sandbox from my protractor config and the high CPU load disappeared. Took a long time to find this out since the problem started occurring previous week and the --no-sandbox place was already in there for a month or so.

[gtgunarathna]

i don't have --no-sandbox in any of Protractor settings file.

When executing test CPU goes to 100%.

is there any other place I should remove the this config

[jmhdez]

I don't know why, but when I add --disable-gpu the problem goes away.

[TARIFABBAS]

After removing --no-sandbox from chrome options, there is coming problem in opening chrome browser on windows 7 while on windows 8 and above it is working fine.
Anyone, have encounter the same problem and what's the solution?