Configuring the hostname

The frontend endpoints are those accessible through a public domain and usually related to authentication/authorization flows that happen through the front-channel. For instance, when an SPA wants to authenticate their users it redirects them to the authorization_endpoint so that users can authenticate using their browsers through the front-channel.

By default, when the hostname settings are not set, the base URL for these endpoints is based on the incoming request so that the HTTP scheme, host, port, and path, are the same from the request. The default behavior also has a direct impact on how the server is going to issue tokens given that the issuer is also based on the URL set to the frontend endpoints. If the hostname settings are not set, the token issuer will also be based on the incoming request and also lack consistency if the client is requesting tokens using different URLs.

When deploying to production you usually want a consistent URL for the frontend endpoints and the token issuer regardless of how the request is constructed. In order to achieve this consistency, you can set either the hostname or the hostname-url options.

Most of the time, it should be enough to set the hostname option in order to change only the host of the frontend URLs:

When using the hostname option the server is going to resolve the HTTP scheme, port, and path, automatically so that:

However, if you want to set not only the host but also a scheme, port, and path, you can set the hostname-url option:

This option gives you more flexibility as you can set the different parts of the URL from a single option. Note that the hostname and hostname-url are mutually exclusive.

The backend endpoints are those accessible through a public domain or through a private network. They are used for a direct communication between the server and clients without any intermediary but plain HTTP requests. For instance, after the user is authenticated an SPA wants to exchange the code sent by the server with a set of tokens by sending a token request to token_endpoint.

By default, the URLs for backend endpoints are also based on the incoming request. To override this behavior, set the hostname-strict-backchannel configuration option by entering this command:

By setting the hostname-strict-backchannel option, the URLs for the backend endpoints are going to be exactly the same as the frontend endpoints.

When all applications connected to Keycloak communicate through the public URL, set hostname-strict-backchannel to true. Otherwise, leave this parameter as false to allow client-server communication through a private network.

The server exposes the administration console and static resources using a specific URL.

By default, the URLs for the administration console are also based on the incoming request. However, you can set a specific host or base URL if you want to restrict access to the administration console using a specific URL. Similarly to how you set the frontend URLs, you can use the hostname-admin and hostname-admin-url options to achieve that. Note that if HTTPS is enabled (http-enabled configuration option is set to false, which is the default setting for the production mode), the Keycloak server automatically assumes you want to use HTTPS URLs. The admin console then tries to contact Keycloak over HTTPS and HTTPS URLs are also used for its configured redirect/web origin URLs. It is not recommended for production, but you can use HTTP URL as hostname-admin-url to override this behaviour.

Most of the time, it should be enough to set the hostname-admin option in order to change only the host of the administration console URLs:

However, if you want to set not only the host but also a scheme, port, and path, you can set the hostname-admin-url option:

Note that the hostname-admin and hostname-admin-url are mutually exclusive.

To reduce attack surface, the administration endpoints for Keycloak and the Admin Console should not be publicly accessible. Therefore, you can secure them by using a reverse proxy. For more information about which paths to expose using a reverse proxy, see Using a reverse proxy.

In this example, the server is running behind a TLS termination proxy and publicly available from https://mykeycloak.

In this example, the server is running without a proxy and exposed using a URL using HTTPS.

It is highly recommended using a TLS termination proxy in front of the server for security and availability reasons. For more details, see Using a reverse proxy.

In this example, backend endpoints are exposed using the same URL used by the server so that clients always fetch the same URL regardless of the origin of the request.

In this example, the server is accessible using a port other than the default ports.

In this example, the server is running behind a proxy and both the server and the proxy are using their own certificates, so the communication between Keycloak and the proxy is encrypted. The reverse proxy uses the Forwarded header and does not set the X-Forwarded-* headers. We need to keep in mind that the proxy configuration options (as well as hostname configuration options) are not changing the ports on which the server actually is listening on (it changes only the ports of static resources like JavaScript and CSS links, OIDC well-known endpoints, redirect URIs, etc.). Therefore, we need to use HTTP configuration options to change the Keycloak server to internally listen on a different port, e.g. 8543. The proxy will be listening on the port 8443 (the port visible while accessing the console via a browser). The example hostname my-keycloak.org will be used for the server and similarly the admin console will be accessible via the admin.my-keycloak.org subdomain.