TypeError: Failed to fetch

[sentry-io]

Is this a real error for the user, or can we stop sending it to Sentry?

If it is a real error, can we display a message to the user instead of sending it to Sentry?

Sentry Issue: MICROMASTERS-3X1

TypeError: Failed to fetch

[umarmughal824]

@pdpinch could I see the complete console log of the dashboard page in the production server as I have attached? It's quite ambiguous to know which specific part of the page is throwing that error.

[umarmughal824]

@pdpinch it's only being produced for CORS on the local machine while adding support for discussions. But in production, it's already showing 200 and all the discussions are lading. So that means it's not the issue actually and all the APIs are giving 200 on the dashboard page on Production as my observation and sentry logs. So that's hard to say anything about it that why it's throwing an error. By the way, I will keep an eye on it as soon as I find something helpful, I will fix it.

It could be the cause of the discussion server down, overloaded queue, and late response but I am not sure about that.

@blarghmatey / @shaidar could you please confirm that have we ever faced a situation when the discussion server is down or overloaded or so on a recurrent basis?

[pdpinch]

@umarmughal824 have you looked for any Sentry events in Open (Discussions) that are recorded at the same time?

[umarmughal824]

@pdpinch I don't have access to the open discussions project on sentry.

[umarmughal824]

@pdpinch Here is my findings. There are some events are occurring against that ambiguous error that is talking about Network Error or promises not handled. So adding a PR for that. So as per my suggestion, mostly we are getting errors for loading discussion posts so this is something related to the server not responding or so.

Note: If we want to debug it more, I need admin rights to check against specific users on production if feasible that are producing that errors. I wanna see if we further see the traceback or console logs while visiting on behalf of the user via user sessions or you could load the dashboard page and could show the console logs against a few selected users that I would request.

[umarmughal824]

This is also a related error and also on the same page. #4934

[briangrossman]

I worked with @umarmughal824 on this today and tested a list of users. When I hijacked the accounts, none of them showed errors in the console.

Note that when the accounts are hijacked, it may the case that I am still logged into Open as myself. I can see Discussion postings for a channel that only I would have access to. I'm not sure if this could affect the findings.

There was one use whose account doesn't seem to be working properly. I got the following error when I loaded their account:

400 Sorry, we were unable to load the data necessary to process your request. Please reload the page.

If the error persists, please contact [email protected] specifying this entire error message.

Note that the only messages I saw in the console were a warning and a verbose message:

• Warning: [Facebook Pixel] - Removed URL query parameters due to potential violations.
• Verbose: [Violation] Avoid using document.write(). https://developers.google.com/web/updates/2016/08/removing-document-write zendesk_widget.js:43

[umarmughal824]

If the error persists, please contact [email protected] specifying this entire error message.

@pdpinch as per the findings by @briangrossman, this error can be ignored because there had been any such error we could have encountered. So this might be the cause of an unstable internet connection or so. In 95% of cases, we have faced issues loading the discussion posts as that takes some time because of hitting the other domain that needs a stable connection to be functional.

As per the sentry forum discussion that's it is recommended that we should ignore such errors.
https://forum.sentry.io/t/unhandledrejection-non-error-promise-rejection-captured-with-value/14062/14

By the way, we have handled the promises rejections in this #4934 PR, so it shouldn't stop posting such errors on sentry.

[pdpinch]

I think we need to upgrade our Sentry account in order to be able to user the IgnoreErrors feature. I'll look into it.