I’ve been using Cloudflare (free) for a couple of years now but last week it started to act up for my WordPress website. I am accessing the website from the Toronto Canada area (it is also where my actual Vultr server is located).
After I purge the Cloudflare cache for some reason it can take a resource file for a page 20.17 seconds (give or take a .05s) to load. The file is random it can be a css file, js, or an image. It sometimes loads the page fine, it sometimes is only 1 file and sometimes it can be up to 5 (if the page is loading a lot of resources).
This also affects my WordPress backend which I have a page rule enabled to disable caching for.
Once the page and all resource files are cached by Cloudflare the page will load fine (unless it has ajax queries then it may have issues again or you are in the backend as the issue will keep happening).
Turning on Cloudflare Developer mode makes all the pages load normally again (obviously from my actual web server). The backend also works fine.
Does anyone know what is going on here? When a file takes time to load it literally takes 20.1 seconds. This seems to indicate to me a timeout issue? Is this a Cloudflare DNS issue? I wonder this since my WordPress backend has issues and that should not be cached by Cloudflare.
I can confirm this is still an issue for us.
We have turned off CF proxy on half the sites now and are looking at alternatives.
It’s too bad, we appreciated CF, it’s been a good experience for years.
But after 100 sites intermittently down for a week and still no response from support, CF probably cannot remain part of our solution.
Hmm I searched before posting the issue… I am surprised I didn’t find yours. I also checked via VPN this morning and accessing the website from the states doesn’t seem to produce the problem so that does indicate a Toronto issue.
I am finding that Developer Mode doesn’t always resolve the issue.
I have gotten 20-second delays that way, too - with cf-cache-status BYPASS.
This looks like it’s not necessarily cache-related. Request from origin and pass-through can still have the same issues.
Turning off proxy for the hostname (orange cloud) always works, be it after a while, which is longer than TTL.