If I understand correctly, currently NPM passes the hostname from the original request through to the upstream server. In many use-cases (perhaps the majority) this is probably the desired behaviour.

Do you want the upstream server (S3 in your case) to see the value you entered in the UI for "Forward Hostname / IP"?

I believe with reverese proxying, the target should not receive the original request's Host header. The target should receive a new Host header specific to the target (ie the s3 host example above) and it should additionally receive an X-Forwarded-For header indicating the original request's host header (actually XFF is a chain of forwarding headers beginning with the original request's Host header.

You may just be right about that. I wonder if there is a public RFC which sets out the standard behaviour?

You could create a pull request for the change and see if the maintainers will accept it. You'd also need to check the same change is applied to custom location declarations too.

I met the same problem, I needed to use $proxy_host because I aim to speed up visiting a website.

For example, I visit abc.com slowly, so I use my VPS and my domain abc.mydomain.com to proxy it. In this case, I should use $proxy_host instead of $host.

Solution: Edit Advanced - Custom Nginx Configuration like below.

location / {
  proxy_pass       $forward_scheme://$server:$port$request_uri;
#  The following line is the nginx default configuration, so there is no difference whether it is commented
#  proxy_set_header Host $proxy_host;
}

Further, you cannot override this proxy_set_header command by adding it to the advanced custom config in the UI because it simply appends the value to the existing header.

This is different from what I tested. The content in Advanced - Custom Nginx Configuration will replace config instead of appending

I highly recommend that developers add an option in the web page graphical configuration, proxy_set_header Host is followed by $host or $proxy_host. Give the users a choice.

I'm having the same problem. In my case the proxied app redirects to $HOST/login.html, whereas it has to redirect to $HOST/location/login.html for login to work.

And you can indeed work around it by just pasting the custom nginx configuration, but thats kind of besides the point.

The nginx default for Host is $proxy_host (https://docs.nginx.com/nginx/admin-guide/web-server/reverse-proxy/), probably precisely for this reason:

By default, NGINX redefines two header fields in proxied requests, “Host” and “Connection”, and eliminates the header fields whose values are empty strings. “Host” is set to the $proxy_host variable, and “Connection” is set to close.

IMHO the default should be respected, in the unlikely case that you truly want to overide Host with $host then you can always set up the location and add a custom header.

I deleted the line
proxy_set_header Host $host;
in /etc/nginx/conf.d/include/proxy.conf

Issue is now considered stale. If you want to keep it open, please comment 👍

I conducted a test using the custom location feature and concluded that the header configuration within the custom location cannot override the default header configurations provided by nginxproxymanager, such as:

perl
Copy code
proxy_set_header Host $host;
proxy_set_header X-Forwarded-Scheme $scheme;
proxy_set_header X-Forwarded-Proto $scheme;
proxy_set_header X-Forwarded-For $remote_addr;
proxy_set_header X-Real-IP $remote_addr;
proxy_pass http://192.168.1.250:8080;
Is there a method available to ensure that the headers in the custom configuration can override the default headers?

config：

nginx proxy file: