We are currently facing an issue where we receive a 15 second delay in serving pages on one of our platforms.
We have been doing load testing and have suddenly started experiencing 15 second delays on pages being served - the delay being there whether we are busy load testing or not.
- This delay seems to be rather constant - it’s like whatever the normal response time is + 15 seconds
- The weird thing is that we experience this delay only from certain machines. I.e. when I access the platform from my machine it loads immediately, when a colleague tries to load the same page there’s a 15 second delay. Same thing, doing a wget from some of our servers results in a 15 second delay from some, not others.
- We are currently experiencing this delay from our load testing provider (Gatling) which is currently impacting our ability to perform load testing.
- We do not have rate limiting enabled on our Cloudflare config for this platform.
The fact that it seems to be rather consistent w.r.t. sources it’s adding a delay, it points to me that it is a Cloudflare issue.
I have seen this 15 second issue being reported by multiple users, but without anyone finding a solution. I hope that someone has been able to resolve this, or at least been able to get Cloudflare to help provide support.
Additionally, we can confirm that it takes the server a couple of ms to process the request once it reaches it.
I’ve tested your 3 zones and am assuming you’re referring to your zone ending in
y.net. Currently its outputting a 522
From our logs:
“message”: " context: Fail to connect to x:443 cause: Connection timed out (os error 110)",
same for https://s*.*-*y.net/
You should make sure your DNS is correct, check for Cloudflare IP addresses blocking at your origin, and if none of this is the issue, check you origin logs for additional information.
Thank you for your response - we’ve spent some more time investigating the issue; it looks like for some reason some requests between Cloudflare and our load balancer (AWS ELB) are delayed.
Due to this being a public forum I am quite limited in the amount of information I feel comfortable sharing - happy to engage in a more private forum if needed.
We are not experiencing the issue on our .net domain, but for an instance on a subdomain on our .ai domain.
We are able to replicate the issue:
On two of our internet facing servers (each with a static IP) - for one server (let’s call it A) when we do a wget to the site home page we get a constant 15 second delay in receiving a response, from the other server (B) we receive the response immediately.
Upon reviewing the web application logs we see that for both requests the application is processing the request within milliseconds - this eliminates the web application as the cause of the issue.
Upon reviewing the load balancer logs, we can see that for server B the request is received by the load balancer immediately - for server A there is a 15 second delay between triggering the wget and the load balancer receiving the request.
This tells us there is an issue - for some IP addresses - for Cloudflare to route requests to our load balancer timeously.
We are not able to replicate the issue on any other subdomain in this same account and we are fairly confident that if we change the subdomain of this service the problem will disappear. However - we prefer to get to the bottom of it.
This issue is currently blocking some ongoing load testing efforts - please let us know if you are able to assist.
@riaand, we are experiencing the same. Did you ever get this resolved? Are you using Workers as part of your pipeline?
Also, is your backend IIS? We seem to only experience this on server-rendered pages, not static assets (even if those assets are not cached by cloudflare).