SSL Termination with NGINX
How to manually configure Keeper Connection Manager SSL termination using NGINX
This documentation assumes that you already have an instance of NGINX properly configured for SSL/TLS, including the necessary private key and certificate, and that Guacamole has already been installed using Keeper Connection Manager. If you do not already have an instance of NGINX ready, please set up an instance of Nginx before proceeding. If you do not already have Guacamole installed, please see the installation instructions.
SSL termination is the recommended method of encrypting communication between users’ browsers and Guacamole, and involves configuring a reverse proxy like Nginx or Apache to handle strictly the SSL/TLS portion of the conversation with the Tomcat instance hosting Guacamole, handling encrypted HTTP externally while passing unencrypted HTTP to Tomcat internally.
Proxying Guacamole through NGINX
If Nginx has been configured for SSL/TLS, there should be a server
section within this configuration that defines the certificate and private key used by Nginx, and which requires Nginx to listen on the standard HTTPS port (443). To proxy Guacamole through Nginx such that Guacamole communication is encrypted, a new location
section will need to be added within this server
section:
where “HOSTNAME” is the hostname or IP address of the internal Guacamole server. This can also be replaced with "http://127.0.0.1:8080" in most cases.
While a typical proxy configuration for Nginx may only specify the proxy_pass and proxy_http_version directives, Guacamole requires additional configuration due to the nature of the application:
proxy_buffering off
disables buffering of packets sent to/from Guacamole. By default, Nginx will buffer communication between itself and the browser, effectively disrupting the stream of events and updates required for remote desktop. Without disabling buffering, the Guacamole connection will at best be slow, and at worst not function at all.The X-Forwarded-For header must be explicitly set to ensure that the IP addresses logged by Guacamole are correct. Without explicitly adding this header (and configuring Tomcat to trust this header), all connections will appear to come from the NGINX server.
The Upgrade and Connection headers are required parts of the WebSocket protocol. If omitted, WebSocket will not function correctly, and Guacamole will fall back to HTTP streaming, which is less efficient.
Applying the updated NGINX configuration
After the above changes have been made, NGINX must be reloaded to force rereading of its configuration files:
If you are using SELinux (the default on both CentOS and RHEL), you must also configure SELinux to allow HTTPD implementations like Nginx to establish network connections:
If Guacamole is not accessible through NGINX after the service has been reloaded, check the NGINX logs and/or journalctl to verify that the syntax of your configuration changes is correct. Such errors will result in NGINX refusing to reload its configuration, or refusing to start up entirely. If you do not see any errors from NGINX, verify that you have configured SELinux to allow NGINX to connect to the network and check the SELinux audit logs (/var/log/audit/audit.log
) for AVC denials.
Configuring Tomcat to trust "X-Forwarded-For
" from NGINX (manual deployment only)
X-Forwarded-For
" from NGINX (manual deployment only)This section applies only if you have manually deployed Guacamole under your own version of Tomcat. This will usually only be the case if:
You have chosen to manually deploy Guacamole under your own install of Apache Tomcat or JBoss, rather than use the provided version of Tomcat.
You are maintaining a deployment of Glyptodon Enterprise that was originally installed before the 2.5 release (2021-09-16).
If you deployed Guacamole automatically (the default and recommended method), this has already been configured for you within the bundled version of Tomcat.
For the client address sent by NGINX via "X-Forwarded-For
" to be correctly trusted as the true client address, you will need to add a "RemoteIpValve
" entry within /etc/tomcat/server.xml
. If this is not specified, the client address will be logged as the address of the internal proxy, which is not usually desirable.
The easiest way to add the required entry is to copy the example server.xml
file provided with the kcm
package, replacing the old /etc/tomcat/server.xml
:
The example server.xml file defines:
A single HTTP connector listening on port 8080.
A
RemoteIpValve
with all settings at their default values.
By default, the RemoteIpValve
will trust "X-Forwarded-For
" from all private networks (10.0.0.0/8
, 172.16.0.0/12
, 192.168.0.0/16
, 169.254.0.0/16
, and both IPv4 and IPv6 localhost). If you need this range to be narrowed, or if you have already made manual edits to server.xml
, you will need to make these changes manually.
Applying the updated Tomcat configuration
Once an appropriate RemoteIpValve
has been specified, Tomcat must be restarted to force rereading of server.xml
:
Last updated