IE 11 and Edge users API calls terminated possibly due Throttling / Immediate spike

I have chat website setup with Cloudflare and due to the nature of application, the traffic will be low most times but spikes up instantly (upto 100s of requests at same time), when a new visitor engages in the portal. Things work fine in Chrome/FF but IE is causing the browser networks access issues. ie., the website or HTML pages/scripts or API calls from the same IE tab is no longer accessible. Users have to open new IE tab to continue the work. I checked Cloudflare settings and disabled “Web Application Framework” to disable DDOS mode, but it does not appear to have affect. Any suggestions would be helpful.

Sounds like it’s something to do with the javascript you use to run the chat, not the network.

If it’s something with the request timing out/failing/etc (you didn’t give much info) I would recommend making sure it reconnects if the connection fails.

If there’s some error in the javascript console, check that out. That would likely be a problem with the library/script you’re using for live chat.

Also, maybe see if pausing CF, or creating a “dev” subdomain that’s not proxied (dns record is :grey:), makes it work with the trident engine. If one of these fixes the problem, then it would be a Cloudflare issue.

Thanks for the response and suggestion. Cloudflare DNS proxy is now set to disabled (since other configurations did not work and great to know about Dev mode). Will monitor it for a day to keep the group posted.

It is possible Chat/JS issue and more likely IE internal issue. Some pointers that make it look like network issue:

  • It occurs primarily from 1 location (same IP) where there are 100 agents trying to grab a visitor chat. Outside of same IP no issue. Domain is whitelisted in company firewalls.
  • Since all users use IE 11, they are requested as separate connections (as opposed to Chrome or FF that use existing connection).
  • After the network connections are failed, existing websockets continue to transmit data and screens work normally, only new API calls are hung.

This topic was automatically closed after 30 days. New replies are no longer allowed.