Page occasionally does not load from cy.visit()

[drewbrend]

Current behavior:

99+% of the time this is not an issue, but once in a while the cy.visit() at the beginning of a test will not load the page, no matter how long of a timeout I set, no XHR requests are sent, nothing is loaded in the page. This is only ever seen in cypress tests and has never been seen manually by our testers so I am sure it is a bug with Cypress.

CypressError: Timed out after waiting '60000ms' for your remote page to load.

Your page did not fire its 'load' event within '60000ms'.

You can try increasing the 'pageLoadTimeout' value in 'cypress.json' to wait longer.

Browsers will not fire the 'load' event until all stylesheets and scripts are done downloading.

When this 'load' event occurs, Cypress will continue running commands.

Desired behavior:

Always load the page.

Steps to reproduce: (app code and test code)

I realize I am not providing a reproducible example, but it is not triggered by any specific test and is not reliably reproducible. I have seen in other tickets that members of the cypress team have asked for additional logging to be turned on. I would like to turn on additional logging that may help diagnose this and provide you the logs next time this happens.

Versions

Cypress 3.1.3

[jennifer-shehane]

Does this happen during cypress open, cypress run, or both?

[lukeapage]

We also have the same problem - I've been spending alot of time investigating this - I wrote up my findings so far on this issue:

#1311 (comment)

but I've definitely ruled out visiting the same url as the cause of the problem - I was about to raise a new issue when I saw this.

I've only ever seen the problem on cypress run but for us (as above) it is approximately 1/200 times that visit fails.

As I say in the link - our server definitely responds with the correct and identical html, but its stalled.

It looks like it is something to do with the way in which the initial result is redirected to the proxy - but my knowledge of the inner workings of cypress are preventing me getting further. I'm happy to add any logging or run a special debug version.

[lukeapage]

I've tried:

1. navigating to a different page first
2. turning on full logging
3. waiting for a second before cy.visit

Is there some way to catch the visit failure and try visiting again?

[drewbrend]

Thanks for the additional info @lukeapage!

@jennifer-shehane cypress run is where we are seeing it most often, because we run that very frequently. I estimate we are averaging 1,000 calls to cy.visit() per day in CI with cypress run, probably more.

I believe I have seen it with cypress open a while ago, but it is very infrequent due to the difference in the amount we run in CI vs local development where one person only runs the few tests they are working on.

[lukeapage]

I’ve ended up solving our problems by retrying :
#1313 (comment)

[drewbrend]

That is a decent workaround. I'm hesitant to implement auto retries because it will allow for sloppier test writing. If there was a way to implement it for just this issue I'd be all over it.

[lukeapage]

If there was a way to implement it for just this issue I'd be all over it.

You may be able to - if the cli returns the reason for the failures, you can alter that script to only retry tests where this is the problem.

My plan is that for CI and getting PR approval, I will allow retry. But I also have a scheduled run of 20x on master where I will disable retry. Any intermittant failures coming up from that can then be investigated.

However I am wondering how even that will work out - I've spent alot of time fixing intermittant failures the last 5 days and #695 is a big source of problems (click does not retry the get if the element is not visible) and quite often there are situations where it is very difficult to assert with a should in order to make sure the click definitely won't fail.. thats alot of effort that I am not sure if it is worth it, espescially if the cypress team fix #695 . So far it looks like this issue and that issue are the main causes and everything else is a real app problem that should be fixed.

[drewbrend]

We're in the same situation, one of the commenters on #695 is my teammate. Besides that issue we've been able to fix all intermittent failures with either a cy.route()/cy.wait() or a .should().

Thanks, I may look into modifying the retry script in January - higher priority things right now, then we're shut down for the holidays.

[Wolsten]

I get the same error running on my development computer using MAMP. The error arises in different specs but only when I choose to "Run all specs". The error is somewhat random as different specs fail in different places, though some seem to fail more than others. This is happening where there is a valid target to click() using cy.get() (rather than using cy.visit() to a specific url). I find that the url is always opened correctly when I run each spec individually. I'll take a look at the work around, but it doesn't appear to be ideal.

[drewbrend]

@Wolsten that sounds like a different issue. This issue is about a page not loading from cy.visit(), not cy.get().click().

[Wolsten]

@drewbrend, thanks for the feedback. I tried replacing the get click with visit(url) and I see the same error, it doesn't appear to matter how the url is determined. The test in question runs fine every time I run in isolation but always times out when run in a sequence of tests. The strange thing is going back through the snapshots, the one for the page I requested through visit(url) is correctly displayed and the following get([name=title] succeeds as shown in the attached screenshot.

[Wolsten]

A question about a possible work around. As the specs run without problem individually, is there a "before" command I could use to reinitialise the state of Cypress? I tried putting in a cy.wait(10000) in the "before" block but this didn't help.

[jennifer-shehane]

There may be an intermittent issue involving the page load event, as I myself have this issue with one of our Cypress projects - this is the isolating factor though. I only have it on our Dashboard project, no other project. I've isolated a few instances in the past 6 months from our develop branch:

CircleCI IDs:
#16251
#15371
#14318
#13936
#8119

Video of failure - looks like the url may have gotten in a weird state before beginning test?
https://dashboard.cypress.io/#/projects/5b79905a-f894-4e91-b60c-fff4d94c218a/runs/47652/specs

Printed errors:

[lukeapage]

@jennifer-shehane you have marked it "needs information" - is there anything we can add? If I were you I would enhance logging, release a new version and then ask for logs of when it occurs again.