Chromium in docker times out unexpectedly on various API call

[ankur22]

Brief summary

When working with chromium in a docker container, it will sometimes time out unexpectedly on API calls, such as typing something into an input text box.

Uncaught (in promise) typing "hello world" in "#input-field": timed out after 30s executor=shared-iterations

When the same script is ran locally without docker, it runs without any issues.

When a screenshot is taken before the input box typing API call, it results in a blank page.

Looking at the compute resource usage, it is off the scale:

Increasing the compute resources will result in Chromium taking less resources, but the issue still persists.

Currently, when a navigation fails, the goto doesn't return the correct error and usually navigates to an error page, could this be the issue? Linked PR: #4862

k6 version

NA

OS

NA

Docker version and image (if applicable)

No response

Steps to reproduce the problem

See brief

Expected behaviour

The page loads and API calls work as expected without timing out.

Actual behaviour

Navigation seems to fail which causes subsequent calls to fail.

[ankur22]

• focus seems to help, but this shouldn't be necessary.
• Increasing the time out to 2 minutes from 30 seconds also helps, but again this shouldn't be needed.

[ankur22]

I'm pretty sure this is an issue with Chromium running in docker. It seems to go in a degraded state if there's no 3D hardware support, and when a website needs rasterization support headless chromium in docker then struggles with software rendering.

I tried working with playwright too and it falls victim to the same issue.

I found this issue which looked promising, but alas that didn't help either.

The one thing that helped a little was to replace type with fill. type performs many requests against chromium (down and up on each character), whereas fill is a considerably less, which seems to have been part of the issue.