This is a modified copy of a ticket I summited early today that was automatically marked as Resolved.
Ticket number: #2397462
Hello,
A member of the Cloudflare Discord encouraged me to reach out, in an attempt to troubleshoot an error page that several of us are intermittently encountering.
What is the Problem:
Randomly receive “503 Service Temporarily Unavailable - Cloudflare”, however reloading the page the request immediately goes through. Unfortunately there is no Ray-ID visible on the page.
Why I suspect it is a problem with CF:
The error has shown up on a Cloudflare pages.dev site multiple times, along with remotely hosted sites(only when the CF proxy is enabled on the domain). If the CF proxy is disabled, the error doesn’t occur. A common denominator seems to be the Cloudflare Toronto location.
First timestamps seen:
My site monitor(Uptime Kuma) on VPS(Vultr)
Timestamp - Domain
2022-02-25 09:43:46 - https://progate.ca
I have also setup an experimental VPS on Vultr located in New Jersey to test my hypothesize that it is related to CF Toronto. So far my results of 8 hours running the same site monitor as on my other Vultr VPS in Toronto has turned up no 503 error, where as my Toronto VPS has gotten 10 503 errors across 3 addresses it checks every 60 seconds.
I seem to be hitting this issue as well randomly. I’m using uptime kuma to keep an eye on my stuff and periodically (as of late around the time that the original post was posted) I’m getting 503 reports that only last a few moments before it’s reported that the service is operating fine.
After running an experiment using a VPS in New Jersey over 30hrs, I found that that Uptime Kuma instance reported 0 failures. [ http://upload.ranni.ca/uploads/ayden/chrome_ZGkJAgtTtK.png ]
Concurrently(over the same 30hrs) I ran two instances that connect to CF Toronto, one residential(not my local LAN) and one hosted on an hourly VPS - both reported 503 errors numerous times.
FMCore: I’m trying to figure out if this is related to a specific CF datacenter. Does your monitoring instance happen to be connecting to CF Toronto or CF Chicago?
The 503 error hasn’t shown up since “2022-03-20 14:03:32” so I guess this problem has been resolved.
I noticed that in the time frame when the 503 errors were showing up I would get a ray id ending in ORD(Chicago) or YYZ(Toronto), but doing a traceroute it seem to only always end at YYZ datacenter. Now when I look at the HTTP header it only shows YYZ, so I am guessing configuration change was made at the Toronto datacenter that seem to resolve the issue.
