Adding Reverse Proxies

This will cover the basic usage of NGINX as a reverse proxy; covering services on the URI path, not as a custom subdomain. See subdomain reverse proxy for setting up subdomains.

Use reference documentation and location block references for additional information.

location blocks should be placed in the server block.

Reference

Reference

Reference

Service Gotchas

Ensure the services running behind the proxy are in a known configuration:

  • Running on expected protocols (http or https) and ports.

  • Have firewalls setup to only allow traffic in/out of those ports to and from the proxy.

Trailing Slash Gotchas

In Short:

  • Services already using URI paths for the services should leave off trailing slashes in location and proxy_pass.

  • Services using no additional URI paths for services should use trailing slashes in location and proxy_pass.

Explanation:

If the proxy_pass directive is specified with a URI, then when a request is passed to the server, the part of a normalized request URI matching the location is replaced by a URI specified in the directive:

Trailing slash example.
location /name/ {
  proxy_pass http://app/remote/;
}
  • Assume request: http://proxy/name/path?params=1.

  • http://app/remote/ sees request as https://app/remote/path?params=1.

  • Essentially the matched URI path is removed and the rest is passed as though it was called from app’s page.

If proxy_pass is specified without a URI, the request URI is passed to the server in the same form as sent by a client when the original request is processed, or the full normalized request URI is passed when processing the changed URI:

No trailing slash example.
location /name/ {
    proxy_pass http://app/remote;
}
  • Assume request: http://proxy/name/path?params=1.

  • http://app/remote/ sees request as https://app/remote/name/path?params=1.

  • Essentially the URI path is concatenated to the end of the remote path.

Reference

Reference

Regex Versus Trailing Slashes

Most examples on the web use regex, but regex is generally slow, error prone and hard to read. In NGINX most regexes may be replaced with trailing slash to replace the matched URI path instead. It’s cleaner and easier to read, and usually covers the regex case.

Web Regex Example (bad) proxies /nzbget to https://nzbget:6791/.
location ~ ^/nzbget$ {
  return                302 $scheme://$host$request_uri/;
}
location ~ ^/nzbget($|./*) {
  rewrite               /nzbget/(.*) /$1 break;
  proxy_pass            https://nzbget:6791;
  include               /etc/nginx/conf.d/proxy.conf;
  proxy_set_header Host $host;
}
  • First rule regex matches nzbget and returns a 302 to the same URI with a trailing slash.

  • Second rule accepts /nzbget with parameters and proxies to the service.

  • This results in two proxy hits and two regex comprehensions; it’s also hard to understand what the regex is doing immediately.

Using trailing slashes (good) proxies /nzbget to https://nzbget:6791/.
location /nzbget/ {
  proxy_pass            https://nzbget:6791/;
  include               /etc/nginx/conf.d/proxy.conf;
  proxy_set_header Host $host;
}

Reference

Redirect Path to Base URI

For applications that serve https://app/.
location /gogs/ {
  proxy_pass https://gogs:3000/;
}

Note

Note trailing slashes.

Redirect Path to Service URI Path

For applications that serve https://app/path.
location /sonarr {
  proxy_pass http://sonarr:8989/sonarr;
  include /etc/nginx/conf.d/proxy-control.conf;
}

Note

Note no trailing slashes.

Custom Path for Service

Enable different paths to the same service.
location /tv {
  return     301 $scheme://$host/sonarr/;
}
location /sonarr {
  proxy_pass http://sonarr:8989/sonarr;
  include    /etc/nginx/conf.d/proxy-control.conf;
}

Note

tv will automatically redirect to sonarr.

Enable Websockets

Allow for apps requiring websockets to be used.
location /crashplan/ {
  proxy_pass                    https://crashplan:5800/;
  include                       /etc/nginx/conf.d/proxy-control.conf;

  location /crashplan/websockify {
    proxy_pass                  https://crashplan:5800/websockify/;
    include                     /etc/nginx/conf.d/proxy-control.conf;
    proxy_set_header Upgrade    $http_upgrade;
    proxy_set_header Connection $connection_upgrade;
  }
}
  • Upgrade and Connection must be used, and pass values through the websocket map to enable the connection upgrade or close the connection.

  • proxy_http_version 1.1 is required, but included in proxy-control.conf.

Rewrite Reponses with Subpath

Some applications are not URI Path aware and will re-write all responses behind the proxy using a static relative path or hostname; which will cause 404 errors and the app to break. Partially fixed using http_sub_module.

Note

Re-writing the proxy respones generally won’t fix a complicated application as there will be a large number of unknown responses that need to be re-written. Usually this is resolved using a sub-domain instead.

sub_filter      https://app:port/ https://reverse-proxy-server/subpath/;
sub_filter      'href="/' 'href="https://reverse-proxy-server/subpath/';
sub_filter_once off;
  • First rule rewrites responses from the app: https://app:port/page.html to https://reverse-proxy-server/subpath/page.html.

  • Second rules rewrites relative responses href="/other-page.html" to href="https://reverse-proxy-server/subpath/other-page.html".

Reference

Reference

Enable NGINX Start/Running with Backends Down

By design NGINX will prevent startup or running if upstream backends are down as it is intepreted to be a configuration error.

Services which are down may not resolve via DNS, and therefore will trigger this condition, requiring all services to be up for NGINX to function.

Another expression is a long running NGINX server where a backend has been restarted. This will mark the backend as bad and NGINX will no longer serve it even though the service may be running now.

By specifying an explicit IP no DNS lookup is required which prevents the service health check, allowing NGINX to start or run with backends down. This will show 502 errors when the service is down. Does not affect cert-based authentication setups.

Static IP for upstream service.
upstream my-backend {
  server {IP}:{PORT};
}

server {
  listen 443 ssl http2;
  server_name myservice.example.com myservice;

  location / {
    proxy_pass http://my-backend;
  }
}

Reference

Reference