Closed avikha closed 3 years ago
👋 Hi there! Thank you for creating this issue.
I am the Selenium Assistant Bot 🤖, I triage issues in this repository. If I can't do it, I label it to help maintainers identify issues that need triaging.
I am an Open Source project 🙌, post bugs or ideas here!
❗️ It seems this issue is not using any of the supported templates
💡 Supported issue types are (they start with):
Issue templates help this project to stay in shape, please use them and fill them out completely. By doing that you are helping the project because the community and maintainers can provide prompt feedback, and potentially solve the issue.
If you are asking a question, a better way to address this is:
If you think this is incorrect, please feel free to open a new issue.
Thank you for your contributions.
🐛 Bug Report
We are seeing
chrome not reachable
exception while running selenium ui tests on aws lambda.I wonder if some one else met this problem? Solved it? Worked around it? Maybe met similar problems on dockers or virtual machines?
Any body that has any suggestion of how to continue and explore this problem, please share. Also, if there is any new information that I can supply or provide, please let me know.
Disclaimer: About five months ago, I published a similar bug on chromium bug list. Unfortunately, no happy news there. So I am trying my luck here as well.
To Reproduce
Step 1: Create AWS account, and then create the AWS lambda running selenium. For more information you can read docs docs
Step 2: Then you can run the following test
Or
Expected behavior
We expected the tests to finish the run, the driver to be terminated, the logs to be printed (published), and everyone to be happy ever since. However, the reality is slightly different - it occurs that the test "fails" right after 60-90 seconds into the run.
We get the following message
Sometimes, the chrome also outputs the following message
More info
Not consistent problem
It is an interesting fact that the problem is not a consistent one, but rather statistical. For instance, running the tests above 30 times may result in 12
chrome not reachable
failures and 18 successes. There are tests where some of them will file due to crashes for 5/30 and some 20/30.The memory issue
We struggle to understand what memory runs out? At first, we thought that it might be the RAM that is used by Lambda. So, what we did is to increase the lambda to MAX of 3Gb RAM - no help, still same problems.
We also noticed that the max memory usage of the lambda is approx 1.1Gb - chrome, chromedriver, java (general), and java tests/infra. Everything that runs on lambda. So we concluded that it is probably not the lambda memory issues.
Local/BrowserStack Runs
We could not recreate the problem running chrome heedlessly/full on "regular" local computer. We tried on Linux, Windows, and Macs. No success there. The same goes for BrowserStack - we can run the same test on the browser stack 30 times and it will pass. No crashes there.
Chrome versions tested and test relations
We also thought that there might be some relevance to some of our tests. However, we saw that on each browser version and each run there are different type of tests that fail. So no consistency there.
Environment
Operating System
Name: Linux Architecture: amd64 OS Version: 4.14.214-164.339.amzn2.x86_64
Browser
Vendor: Chromium Version: >= 79
Selenium/Language
Java: 1.8.0_201 Binding version: 3.141.0
Chrome flags