Yes, our team has tested this 520 error.
The load of cpu and database is close to 100%.
But we don’t know what the specific impact of this load situation on the 520 error is.
Because it won’t happen without going through the cloudflare proxy.
1.GCP access_log has no 520 error log (looks all good)
2.Driect connection to GCP without any 520 error (with same connection session)
3.Using CF proxy with cache make some 520 (looks random file url)
4.Using CF with developer mode is fine, no 520 error
5.we check GCP CPU and SQL load below 50% still got 520 error
6.when doing load test at GCP will get 100% CPU load for 2~3 secs, and upload 100Mbps+ content
The HTTP 520 status code is more like Cloudflare’s proprietary status code, so it won’t be returned from your GCP instance. But what you need to take a look at is the possibility of TCP connection reset from your instance, invalid HTTP response returned, empty response returned, missing response headers in HTTP response, and more.
If you notice the above behaviour occasionally, then it could be the cause for Error 520 generated from Cloudflare.
This seems to be a good way to figure out this error.
We will collect more information and log for this,
Does Cloudflare able to find more detail log for error520 by RayID
or does any tool for troubleshooting include dump data for response or trace error connection package? or do we need to upgrade to Pro account for this tool.