Missing browserCriClient in connectToNewSpec

[jazpearson]

Current behavior

Unfortunately, this issue is intermittent and only happens in our CI pipeline. But i wanted to see if could raise the issue in case there's something obvious.

We run cypress with the following: cypress-run cross-env NODE_OPTIONS=--max-http-header-size=1000000 cypress run --browser chrome. And we run in a docker container, using the cypress/included images.

This error happens mid-way running through our test suites. Our tests can take 30 minutes, and the following error seems to occur when ending a test suite (or conversely, when starting the next test suite).

Timed out waiting for the browser to connect. Retrying...
Missing browserCriClient in connectToNewSpec
Error: Missing browserCriClient in connectToNewSpec
    at Object.connectToNewSpec (<embedded>:2744:242702)
    at Object.connectToNewSpec (<embedded>:2750:24813)
    at process.processTicksAndRejections (node:internal/process/task_queues:96:5)
    at async v.relaunchBrowser (<embedded>:2802:346537)

When this happens, the test run completely exits and the remaining suites do not run.

Desired behavior

For the test run not to crash

Test code to reproduce

Really difficult to do, sorry. It's intermittent, and it's a fairly large, sensitive, code-base.

Cypress Version

11.0.1

Node version

16.16.0

Operating System

Jenkins CI - CentOS (but runs the cypress docker image)

Debug Logs

No response

Other

We've seen this happening in v10 and v11.

[rsouthgate]

@jazpearson Don't know if this helps but stumbled across this when trying to figure out why our own CI pipeline was breaking with this error message after upgrading agents to 10.11.0. In our case the error was being caused by our test director (sorry-cypress) and ultimately was due to this off-by-one error sorry-cypress/sorry-cypress#613 which has since been fixed.

[nagash77]

Hi @jazpearson , I'm sorry to hear you are having problems using Cypress. Unfortunately without a reproducible example, we wont be able to further assist you. If you can figure out a way to reproduce this issue that you can share with us please do comment here and we can reopen the issue.

[tommy0831]

I am also seeing this error with Cypress 10.11 when running on Jenkins pipeline. This happens very rarely.

3:11:26 Timed out waiting for the browser to connect. Retrying... 13:11:26 Missing browserCriClient in connectToNewSpec 13:11:26 Error: Missing browserCriClient in connectToNewSpec 13:11:26 at Object.connectToNewSpec (/root/.cache/Cypress/10.11.0/Cypress/resources/app/packages/server/lib/browsers/chrome.js:451:19) 13:11:26 at Object.connectToNewSpec (/root/.cache/Cypress/10.11.0/Cypress/resources/app/packages/server/lib/browsers/index.js:96:31) 13:11:26 at processTicksAndRejections (node:internal/process/task_queues:96:5) 13:11:26 at OpenProject.relaunchBrowser (/root/.cache/Cypress/10.11.0/Cypress/resources/app/packages/server/lib/open_project.js:150:24)

[nagash77]

Hi @tommy0831 , can you provide a reproducible example that our engineers can use to debug further?

[jazpearson]

One thing i did do was to update these two libraries that i had installed:

• @cypress/webpack-preprocessor from 4.0.3 to 5.15.7
• cypress-terminal-report from 4.1.2 to 5.0.0.

I don't know if it was coincidence or not, but that did help improve things. It still happens, but it's reasonably rare at the minute. Maybe happens 1/20 times, whereas before it was sometimes as bad as 1/4.

I wish i could provide a sample, but i'm under NDA.

This is currently some of our reporting 3rd party components.

    "cypress-mochawesome-reporter": "3.2.3",
    "cypress-multi-reporters": "^1.6.2",
    "cypress-terminal-report": "^5.0.0"
    "mocha-junit-reporter": "^2.2.0"

[matanbanin]

Happened to me also. Anything new on this issue?

[nagash77]

Hi @matanbanin , We still do not have a reproducible example for our Cypress devs to use to debug this issue. If you can provide one I will reopen the ticket for further investigation.

[swatinigam]

Hello, I am also seeing this error. I am running my tests locally (not in ci). These are the dev dependencies:
"cypress": "10.8.0",
"cypress-cucumber-preprocessor": "^4.2.0",
"cypress-real-events": "^1.7.1",
"cypress-repeat": "^2.3.3",
"cypress-xpath": "1.6.2"
Below is the error:

Timed out waiting for the browser to connect. Retrying...
Missing browserCriClient in connectToNewSpec
Error: Missing browserCriClient in connectToNewSpec
    at Object.connectToNewSpec (/Users/username/Library/Caches/Cypress/10.8.0/Cypress.app/Contents/Resources/app/packages/server/lib/browsers/chrome.js:425:19)
    at Object.connectToNewSpec (/Users/username/Library/Caches/Cypress/10.8.0/Cypress.app/Contents/Resources/app/packages/server/lib/browsers/index.js:95:31)
    at processTicksAndRejections (node:internal/process/task_queues:96:5)
    at OpenProject.relaunchBrowser (/Users/username/Library/Caches/Cypress/10.8.0/Cypress.app/Contents/Resources/app/packages/server/lib/open_project.js:148:24)

I have never seen this error before. All I did was run tests, but today I managed to see it a few times already.

[nagash77]

Hi @swatinigam , Do you have a reproducible example by chance? So far no one has been able to provide us with a reproducible example to help us debug this issue. Without being able to reproduce the issue it is extremely unlikely we will find a solution. If you can share one that would allow us to move this ticket forward.

[matanbanin]

It's happening just not reproducible. happens for about 2% of builds

[emilyrohrbough]

This seems to be the same issue as #22451, however the error is being thrown correctly here, where as it's getting one line of code further in #22451... we had identified it as a mistake on how we are re-launching a browser after the first spec if the browser process is corrupt.

This error is occurs here: https://github.com/cypress-io/cypress/blob/7721e2be9d7cd6695a16385f3a8965625da162c4/packages/server/lib/browsers/chrome.ts

[matanbanin]

Awesome! any idea on how much time will take to deploy a fix? Is there any safe version to rollback to?

[samixchoumi]

Hello, i've got the same error for no reason today :
My devDependencies from package.json looks like that :

"devDependencies": {
    "@badeball/cypress-cucumber-preprocessor": "^15.1.3",
    "@bahmutov/cypress-esbuild-preprocessor": "^2.1.5",
    "@shelex/cypress-allure-plugin": "^2.35.2",
    "@typescript-eslint/eslint-plugin": "^5.51.0",
    "@typescript-eslint/parser": "^5.51.0",
    "allure-js-commons": "^2.0.0-beta.26",
    "cypress": "^12.5.1",
    "cypress-multi-reporters": "^1.6.2",
    "cypress-real-events": "^1.7.6",
    "cypress-recurse": "^1.27.0",
    "cypress-slack-reporter": "^1.5.2",
    "cypress-slow-down": "^1.2.1",
    "cypress-time-marks": "^1.4.0",
    "esbuild": "^0.16.17",
    "eslint": "^8.34.0",
    "eslint-plugin-cypress": "^2.12.1",
    "fs-extra": "^11.1.0",
    "mocha": "^10.2.0",
    "mochawesome": "^7.1.3",
    "mochawesome-merge": "^4.2.2",
    "mochawesome-report-generator": "^6.2.0",
    "moment": "^2.29.4",
    "typescript": "^4.9.5"
  },

Any news about that ?
I use this new config into my cypress.config.ts file :
experimentalMemoryManagement: true,

[mkruisman2]

Any update on this issue? Since a few months we are experiencing the same and currently are getting this error multiple times per day, even with the test suites just running once per hour in CircleCI. We're on Cypress 12.3.0

We're also experiencing issue #747, in case it's related.

[chrisbreiding]

It's hard to say without a reliable way to reproduce this, but there's a chance this is fixed by #25898. It will be out with the next release, but if you'd like to check it out ahead of time, I'd recommend installing the prerelease build for the latest commit on the develop branch.

[samixchoumi]

@chrisbreiding hello, i still got the error only on gitlab (of course) after updating Cypress to 12.7.0, and i've no idea on how to reproduce it because, it works sometime, and some other time it fail.
I think it's linked to GitLab parallelization but it's just an idea.

[jazpearson]

Unfortunately we are still seeing this issue in 12.7.0

[hannahhaken]

We're still seeing this issue sporadically in our GoCD pipeline and we've upgraded to 12.8.1. No parallelisation being done at our end as well.

[jazpearson]

Is there anything we can add to help debug this? I know we want a reproducible repo, but it seems most of us aren't in a position to share our code.

This is starting to cripple us, as we've added more tests, the chances of this happening has gone up, and we are having to run tests multiple times to check for passes. Our tests take over an hour to run now.

Seeing this issue in 12.9.0

[narancshab]

It happens with "cypress": "^12.4.1", too:

other dependencies:

    "@badeball/cypress-cucumber-preprocessor": "15.1.3",
    "@bahmutov/cypress-esbuild-preprocessor": "2.1.3",
    "cucumber-html-reporter": "5.5.0",
    "cypress-file-upload": "5.0.8",
    "cypress-iframe": "^1.0.1",

I think it is somehow related to the cucumber-preprocessor.

I executed locally, in headless mode.

[RuudMeeuwsen]

I've also seen the issue in several 12.x versions (I'm currently on 12.9.0, where it also occurs).
I'm not using cypress-cucumber-preprocessor, and @jazpearson didn't mention it either, so I guess we can safely rule it out as a potential cause of the issue.
Unfortunately I'm also under NDA, so I can't share my codebase either :(.

[hannahhaken]

@nagash77 @chrisbreiding we're still seeing this issue. I'm not under any kind of NDA with my company's code and am happy to work with someone from Cypress to show you this issue and let you see our overall setup, if that helps at all?

[Stubennett1]

@chrisbreiding Any update on this issue? Unfortunately this is rapidly becoming a showstopper for us.

Can anyone provide any suggestions on how to mitigate the issue as we are seeing it almost 100% of our runs in the CI/CD environment.

[hannahhaken]

We're still seeing this issue 1 in every 10 builds, and have resorted to splitting off a small subset of our tests to run in just the prod stage of our GoCD pipeline instead of running them all just in preprod. We're yet to see the issue occur since we've done this. We have a quick enough feedback loop and release/rollback process so we can react to any issues flagged by our Cypress tests in prod quickly enough.

Not an ideal solution, and won't be an option for all experiencing this issue. However this does highlight that the length of the test suite has something to do with why we were experiencing the issue.

@chrisbreiding my offer still stands to work with you and show you our codebase, pipeline and test setup if any of this info might help you understand what's causing the issue.

[matanbanin]

Still happening in 12.10.0

[hannahhaken]

Hi everyone, thought I'd give an update on something we've changed which has so far stopped this issue from occurring. We've increased the mem_limit and mem_reservation in our docker image.

Previously it was set to:

mem_limit: 3500m
mem_reservation: 3250m

and now we've updated it to be:

mem_limit: 6000m
mem_reservation: 4000m

Like I said, since this change we haven't seen the browserCriClient issue once in our CI pipeline (we'd never encountered this locally).

Our original memory must have been set a while ago, and we haven't thought to change it until now. We thought it might need bumping up after spotting this in the "Hardware" section of getting started, the part about running tests reliably.
(https://docs.cypress.io/guides/getting-started/installing-cypress#Hardware)

Hopefully others experiencing this issue can give this a try and share if it works for them too. Good luck everyone.

[samixchoumi]

Any update ? We don't use docker image and it still appearing sometime.

[jazpearson]

We are running 12.11.0 and having to run builds mutliple times to get past this issue. It's becoming unworkable and we've invested so much time into creating these tests, We aren't using docker compose.

[jazpearson]

#25844 looks like another case was described here, but with reproducible code? but this issue was told it was a duplicate of this one....

[jazpearson]

I've submitted a PR in the hope to get some attention to this issue, because it's really hurting us now.

[przedab]

The same happens on version 12.13.0.

[AtofStryker]

Hey @jazpearson . I put together #27116 which I think might fix this issue. I will work on getting a binary built and mentioning it on this issue. In the meantime, I am going to close #26882 since I don't believe it is going to fix your issue as recreating the CRI client in that context will just give the wrong port and refuse the connection. I'll keep this issue up to date once I get a binary!

[jazpearson]

That's fantasic, thanks @AtofStryker. I wasn't convinced my PR would work, and was more to try and get the ball rolling with someone who knew what they were doing.

Thanks again - and will look forward to testing things out.

[AtofStryker]

I meant to update this on Friday, but the binaries with the fix proposed in #27116 are here. @jazpearson confirmed it looks to be working in #24650 (comment), but would love others to try and confirm!

[matanbanin]

YAY! Thanks!

[jazpearson]

In my testing, i did see the issue happening again, but the frequency seemed to be much reduced from before.

[cypress-bot]

Released in 12.17.0.

This comment thread has been locked. If you are still experiencing this issue after upgrading to
Cypress v12.17.0, please open a new issue.

[AtofStryker]

In my testing, i did see the issue happening again, but the frequency seemed to be much reduced from before.

good to know. There might be an edge case we are missing, but it's going to be hard to determine. If you are able to reproduce it, let us know and we can see how we can improve this fix!

[jennifer-shehane]

There is another fix to a similar error like this that will be coming out in 13.13.0, so please update to that version when it releases. PR: #29663