CD
Cloudflare Developers•4w ago
Sergey

CF is randomly loosing port number in HTTP header "host"

I'm not sure if this is the right place to report a bug. Cloudflare is randomly sending wrong "host" HTTP header to end server (loosing port number). Once per every approx. 15-20 requests. More details: My HTTPS server is listening on port 8443. Traffic is routed through Cloudflare. Here is an example URL: https://check-cf.vezha.com:8443/testCrawl Normally my webserver is connected to port 8443 and HTTP header "host" is filled with check-cf.vezha.com:8443. But once per approximately 15-20 GET requests, my server is getting request with incorrect "host" header: check-cf.vezha.com. The port number is missing. The mentioned webserver is configured specially to show the problem for you. It is using the following trivial Nginx config for the domain check-cf.vezha.com: 
server {
  server_name check-cf.vezha.com;
  listen      0.0.0.0:8443 ssl http2;
  root        /var/www/html;

  include     commonSSL.conf;

  location    /
  {
    add_header Content-Type text/plain;
    return 200 'healthy; http_host: $http_host;';
  }
}
server {
  server_name check-cf.vezha.com;
  listen      0.0.0.0:8443 ssl http2;
  root        /var/www/html;

  include     commonSSL.conf;

  location    /
  {
    add_header Content-Type text/plain;
    return 200 'healthy; http_host: $http_host;';
  }
}
Here is how I automate executing HTTP requests: 
set -eu

while true
do
  curl https://check-cf.vezha.com:8443/testCrawl
  echo ""
done
set -eu

while true
do
  curl https://check-cf.vezha.com:8443/testCrawl
  echo ""
done
Here is a typical Bash script output: 
c:\temp\crawl_vezha_for_error>bash crawl.sh
healthy; http_host: check-cf.vezha.com:8443;
healthy; http_host: check-cf.vezha.com:8443;
healthy; http_host: check-cf.vezha.com;
healthy; http_host: check-cf.vezha.com:8443;
healthy; http_host: check-cf.vezha.com:8443;
healthy; http_host: check-cf.vezha.com;
healthy; http_host: check-cf.vezha.com:8443;
c:\temp\crawl_vezha_for_error>bash crawl.sh
healthy; http_host: check-cf.vezha.com:8443;
healthy; http_host: check-cf.vezha.com:8443;
healthy; http_host: check-cf.vezha.com;
healthy; http_host: check-cf.vezha.com:8443;
healthy; http_host: check-cf.vezha.com:8443;
healthy; http_host: check-cf.vezha.com;
healthy; http_host: check-cf.vezha.com:8443;
Here is an example of HTTP response headers for the case when port number was lost by CF: 
< HTTP/2 200
< date: Tue, 04 Feb 2025 04:07:00 GMT
< content-type: text/plain
< content-length: 39
< server: cloudflare
< cf-cache-status: DYNAMIC
< cf-ray: 90c7ca93d9fcbbde-WAW
< alt-svc: h3=":8443"; ma=86400
<
healthy; http_host: check-cf.vezha.com;* Connection #0 to host check-cf.vezha.com left intact
< HTTP/2 200
< date: Tue, 04 Feb 2025 04:07:00 GMT
< content-type: text/plain
< content-length: 39
< server: cloudflare
< cf-cache-status: DYNAMIC
< cf-ray: 90c7ca93d9fcbbde-WAW
< alt-svc: h3=":8443"; ma=86400
<
healthy; http_host: check-cf.vezha.com;* Connection #0 to host check-cf.vezha.com left intact
1 Reply
Walshy
Walshy•4w ago
Hey, Looks like a bug in a new platform update we're rolling out - I've reported it to the team. Thanks for the report!

Did you find this page helpful?