This is a weird issue I’m baffled to not find any clues online so I’m not sure anyone outside our Okta host is experiencing.
Prerequisites:
Okta Client Application with email authenticator (can be a MFA)
Either manually embedded Okta Sign In widget or the redirect sign in page from the Okta host. This can be reproduced accessing the Okta Admin Panel as well.
Repro steps
Sign in to the client app and select email as authenticator factor.
The polling requests fails only once, at which point the error appears on UI. The error does not go away afterwards.
The error is a ERR_HTTP2_PROTOCOL_ERROR.
I have tried incognito and cleaning cookies and local cache.
This only happens on chromium based browsers (I am on version 143.0.7499.40, but it is safe to assume this happens on multiple versions).
Many of our users are experiencing the same.
This is happening in both our Okta Preview and Main hosts.
This does not happen on my personal trial host I use for testing (integrator plan).
The actual sign in workflow works without issues through the email link.
I don’t think this is a configuration issue on our side, the very same configs in my trial host are replicated exactly the same on our Okta Preview and Main hosts.
Any ideas? Or is this something that needs to be handled with the support team?
Typically when I’ve seen this error occur, there’s been some sort of network connection issue (for example, I’ve been able to trigger it by enabling/disabling a VPN in the middle of the login flow)
When you mention this not occurring on the integrator org you use for testing, are you also able reproduce the issue with your prod/preview orgs while on the same network/machine? If it only happens on your prod/preview orgs, is there anything that could be impacting the network connections when those particular orgs are accessed that doesn’t occur when you access the integrator org?
Yes, I have executed all my tests on the same machine/network configurations using two different browsers, a chromium-based and a firefox-based one. Not only I can always reproduce on chromium, where just a single poll request fails every time, it never happens on firefox.
The other comment I have regarding this would be why does the widget does not recover when the subsequent poll requests are successful? I mean the widget is still fully functional and works fine after a single failed poll request, we are mostly worried about our customers opinion having an error pop up each time and staying there, which is confusing.
The only difference between my integrator plan and our preview/prod hosts I can think of that could be impacting connections would be the custom domain we have on both the latter, but I think it would be a far-fetched possibility.
I am experiencing this exact issue in an android webview - /poll requests regularly fails after being on the mfa-verify-passcode controller for a few seconds; the “Failed to fetch” error does not disappear even though subsequent requests to the /pollendpoint are successful.
I have been discussing this with my team and at the very least we believe the widget should recover itself, I’ll be creating a github issue in the sign in widget repo, let me know if there would be anything else I can do from my side.
