This guide demonstrates how your OpenID Connect (OIDC) application can support multi-tenancy so that you can serve multiple tenants from a single application. Tenants can be distinct realms or security domains within the same OpenID Provider or even distinct OpenID Providers.

When serving multiple customers from the same application (e.g.: SaaS), each customer is a tenant. By enabling multi-tenancy support to your applications you are allowed to also support distinct authentication policies for each tenant even though if that means authenticating against different OpenID Providers, such as Keycloak and Google.

Please read the Using OpenID Connect to Protect Service Applications guide if you need to authorize a tenant using Bearer Token Authorization.

Please read the Using OpenID Connect to Protect Web Applications guide if you need to authenticate and authorize a tenant using OpenID Connect Authorization Code Flow.

Prerequisites

To complete this guide, you need:

  • Roughly 15 minutes

  • An IDE

  • JDK 11+ installed with JAVA_HOME configured appropriately

  • Apache Maven 3.8.4

  • A working container runtime (Docker or Podman)

  • Optionally the Quarkus CLI if you want to use it

  • Optionally Mandrel or GraalVM installed and configured appropriately if you want to build a native executable (or Docker if you use a native container build)

  • jq tool

Architecture

In this example, we build a very simple application which offers a single land page:

  • /{tenant}

The land page is served by a JAX-RS Resource and shows information obtained from the OpenID Provider about the authenticated user and the current tenant.

Solution

We recommend that you follow the instructions in the next sections and create the application step by step. However, you can go right to the completed example.

Clone the Git repository: git clone https://github.com/quarkusio/quarkus-quickstarts.git, or download an archive.

The solution is located in the security-openid-connect-multi-tenancy-quickstart directory.

Creating the Maven Project

First, we need a new project. Create a new project with the following command:

CLI
quarkus create app org.acme:security-openid-connect-multi-tenancy-quickstart \
    --extension=oidc,resteasy-jackson \
    --no-code
cd security-openid-connect-multi-tenancy-quickstart

To create a Gradle project, add the --gradle or --gradle-kotlin-dsl option.

For more information about how to install the Quarkus CLI and use it, please refer to the Quarkus CLI guide.

Maven
mvn io.quarkus.platform:quarkus-maven-plugin:999-SNAPSHOT:create \
    -DprojectGroupId=org.acme \
    -DprojectArtifactId=security-openid-connect-multi-tenancy-quickstart \
    -Dextensions="oidc,resteasy-jackson" \
    -DnoCode
cd security-openid-connect-multi-tenancy-quickstart

To create a Gradle project, add the -DbuildTool=gradle or -DbuildTool=gradle-kotlin-dsl option.

If you already have your Quarkus project configured, you can add the oidc extension to your project by running the following command in your project base directory:

CLI
quarkus extension add 'oidc'
Maven
./mvnw quarkus:add-extension -Dextensions="oidc"
Gradle
./gradlew addExtension --extensions="oidc"

This will add the following to your build file:

pom.xml
<dependency>
    <groupId>io.quarkus</groupId>
    <artifactId>quarkus-oidc</artifactId>
</dependency>
build.gradle
implementation("io.quarkus:quarkus-oidc")

Writing the application

Let’s start by implementing the /{tenant} endpoint. As you can see from the source code below it is just a regular JAX-RS resource:

package org.acme.quickstart.oidc;

import javax.inject.Inject;
import javax.ws.rs.GET;
import javax.ws.rs.Path;

import org.eclipse.microprofile.jwt.JsonWebToken;

import io.quarkus.oidc.IdToken;

@Path("/{tenant}")
public class HomeResource {

    /**
     * Injection point for the ID Token issued by the OpenID Connect Provider
     */
    @Inject
    @IdToken
    JsonWebToken idToken;

    /**
     * Returns the tokens available to the application. This endpoint exists only for demonstration purposes, you should not
     * expose these tokens in a real application.
     *
     * @return the landing page HTML
     */
    @GET
    public String getHome() {
        StringBuilder response = new StringBuilder().append("<html>").append("<body>");

        response.append("<h2>Welcome, ").append(this.idToken.getClaim("email").toString()).append("</h2>\n");
        response.append("<h3>You are accessing the application within tenant <b>").append(idToken.getIssuer()).append(" boundaries</b></h3>");

        return response.append("</body>").append("</html>").toString();
    }
}

In order to resolve the tenant from incoming requests and map it to a specific quarkus-oidc tenant configuration in application.properties, you need to create an implementation for the io.quarkus.oidc.TenantResolver interface.

package org.acme.quickstart.oidc;

import javax.enterprise.context.ApplicationScoped;

import io.quarkus.oidc.TenantResolver;
import io.vertx.ext.web.RoutingContext;

@ApplicationScoped
public class CustomTenantResolver implements TenantResolver {

    @Override
    public String resolve(RoutingContext context) {
        String path = context.request().path();
        String[] parts = path.split("/");

        if (parts.length == 0) {
            // resolve to default tenant configuration
            return null;
        }

        return parts[1];
    }
}

From the implementation above, tenants are resolved from the request path so that in case no tenant could be inferred, null is returned to indicate that the default tenant configuration should be used.

=== When a current tenant represents an OIDC web-app application, the current io.vertx.ext.web.RoutingContext will contain a tenant-id attribute by the time the custom tenant resolver has been called for all the requests completing the code authentication flow and the already authenticated requests, when either a tenant specific state or session cookie already exists. Therefore, when working with mulltiple OpenID Connect Providers, you only need a path specific check to resolve a tenant id if the RoutingContext does not have the tenant-id attribute set, for example: 
package org.acme.quickstart.oidc;

import javax.enterprise.context.ApplicationScoped;

import io.quarkus.oidc.TenantResolver;
import io.vertx.ext.web.RoutingContext;

@ApplicationScoped
public class CustomTenantResolver implements TenantResolver {

    @Override
    public String resolve(RoutingContext context) {
        String tenantId = context.get("tenant-id");
        if (tenantId != null) {
            return tenantId;
        } else {
            // Initial login request
            String path = context.request().path();
            String[] parts = path.split("/");

            if (parts.length == 0) {
                // resolve to default tenant configuration
                return null;
            }
            return parts[1];
        }
    }
}

===

=== If you also use Hibernate ORM multitenancy and both OIDC and Hibernate ORM tenant IDs are the same and must be extracted from the Vert.x RoutingContext then you can pass the tenant id from the OIDC Tenant Resolver to the Hibernate ORM Tenant Resolver as a RoutingContext attribute, for example: 
public class CustomTenantResolver implements TenantResolver {

    @Override
    public String resolve(RoutingContext context) {
        String tenantId = extractTenantId(context);
        context.put("tenantId", tenantId);
        return tenantId;
    }
}

===

Configuring the application

# Default Tenant Configuration
quarkus.oidc.auth-server-url=http://localhost:8180/realms/quarkus
quarkus.oidc.client-id=multi-tenant-client
quarkus.oidc.application-type=web-app

# Tenant A Configuration
quarkus.oidc.tenant-a.auth-server-url=http://localhost:8180/realms/tenant-a
quarkus.oidc.tenant-a.client-id=multi-tenant-client
quarkus.oidc.tenant-a.application-type=web-app

# HTTP Security Configuration
quarkus.http.auth.permission.authenticated.paths=/*
quarkus.http.auth.permission.authenticated.policy=authenticated

The first configuration is the default tenant configuration that should be used when the tenant can not be inferred from the request. This configuration is using a Keycloak instance to authenticate users.

The second configuration is the configuration that will be used when an incoming request is mapped to the tenant tenant-a.

Note that both configurations map to the same Keycloak server instance while using distinct realms.

You can define multiple tenants in your configuration file, just make sure they have a unique alias so that you can map them properly when resolving a tenant from your TenantResolver implementation.

Google OpenID Provider Configuration

In order to set-up the tenant-a configuration to use Google OpenID Provider, you need to create a project as described here.

Once you create the project and have your project’s client_id and client_secret, you can try to configure a tenant as follows:

# Tenant configuration using Google OpenID Provider
quarkus.oidc.tenant-b.auth-server-url=https://accounts.google.com
quarkus.oidc.tenant-b.application-type=web-app
quarkus.oidc.tenant-b.client-id={GOOGLE_CLIENT_ID}
quarkus.oidc.tenant-b.credentials.secret={GOOGLE_CLIENT_SECRET}
quarkus.oidc.tenant-b.token.issuer=https://accounts.google.com
quarkus.oidc.tenant-b.authentication.scopes=email,profile,openid

Starting and Configuring the Keycloak Server

To start a Keycloak Server you can use Docker and just run the following command:

docker run --name keycloak -e KEYCLOAK_ADMIN=admin -e KEYCLOAK_ADMIN_PASSWORD=admin -p 8180:8080 quay.io/keycloak/keycloak:{keycloak.version} start-dev

where keycloak.version should be set to 17.0.0 or higher.

You should be able to access your Keycloak Server at localhost:8180.

Log in as the admin user to access the Keycloak Administration Console. Username should be admin and password admin.

Now, follow the steps below to import the realms for the two tenants:

For more details, see the Keycloak documentation about how to create a new realm.

Running and Using the Application

Running in Developer Mode

To run the microservice in dev mode, use:

CLI
quarkus dev
Maven
./mvnw quarkus:dev
Gradle
./gradlew --console=plain quarkusDev

Running in JVM Mode

When you’re done playing with dev mode, you can run it as a standard Java application.

First compile it:

CLI
quarkus build
Maven
./mvnw clean package
Gradle
./gradlew build

Then run it:

java -jar target/quarkus-app/quarkus-run.jar

Running in Native Mode

This same demo can be compiled into native code: no modifications required.

This implies that you no longer need to install a JVM on your production environment, as the runtime technology is included in the produced binary, and optimized to run with minimal resource overhead.

Compilation will take a bit longer, so this step is disabled by default; let’s build again by enabling the native build:

CLI
quarkus build --native
Maven
./mvnw package -Dnative
Gradle
./gradlew build -Dquarkus.package.type=native

After getting a cup of coffee, you’ll be able to run this binary directly:

./target/security-openid-connect-multi-tenancy-quickstart-runner

Testing the Application

To test the application, you should open your browser and access the following URL:

If everything is working as expected, you should be redirected to the Keycloak server to authenticate. Note that the requested path defines a default tenant which we don’t have mapped in the configuration file. In this case, the default configuration will be used.

In order to authenticate to the application you should type the following credentials when at the Keycloak login page:

  • Username: alice

  • Password: alice

After clicking the Login button you should be redirected back to the application.

If you try now to access the application at the following URL:

You should be redirected again to the login page at Keycloak. However, now you are going to authenticate using a different realm.

In both cases, if the user is successfully authenticated, the landing page will show the user’s name and e-mail. Even though user alice exists in both tenants, for the application they are distinct users belonging to different realms/tenants.

Resolving Tenant Identifiers with Annotations

You can use the annotations and CDI interceptors for resolving the tenant identifiers as an alternative to using quarkus.oidc.TenantResolver. This can be done by setting the value for the key OidcUtils.TENANT_ID_ATTRIBUTE on the current RoutingContext.

Assuming your application supports two OIDC tenants (hr, and default) first you need to define one annotation per tenant ID other than default:

Proactive HTTP authentication needs to be disabled (quarkus.http.auth.proactive=false) for this to work. See Proactive Authentication section for further details.

 
@Inherited
@InterceptorBinding
@Retention(RetentionPolicy.RUNTIME)
@Target({ElementType.METHOD, ElementType.TYPE})
public @interface HrTenant {
}

Next, you’ll need one interceptor for each of those annotations:

@Interceptor
@HrTenant
public class HrTenantInterceptor {
    @Inject
    RoutingContext routingContext;

    @AroundInvoke
    Object setTenant(InvocationContext context) throws Exception {
        routingContext.put(OidcUtils.TENANT_ID_ATTRIBUTE, "hr");
        return context.proceed();
    }
}

Now all methods and classes carrying @HrTenant will be authenticated using the OIDC provider configured by quarkus.oidc.hr.auth-server-url, while all other classes and methods will still be authenticated using the default OIDC provider.

Programmatically Resolving Tenants Configuration

If you need a more dynamic configuration for the different tenants you want to support and don’t want to end up with multiple entries in your configuration file, you can use the io.quarkus.oidc.TenantConfigResolver.

This interface allows you to dynamically create tenant configurations at runtime:

package io.quarkus.it.keycloak;

import javax.enterprise.context.ApplicationScoped;
import java.util.function.Supplier;

import io.smallrye.mutiny.Uni;
import io.quarkus.oidc.OidcTenantConfig;
import io.quarkus.oidc.TenantConfigResolver;
import io.vertx.ext.web.RoutingContext;

@ApplicationScoped
public class CustomTenantConfigResolver implements TenantConfigResolver {

    @Override
    public Uni<OidcTenantConfig> resolve(RoutingContext context, TenantConfigResolver.TenantConfigRequestContext requestContext) {
        String path = context.request().path();
        String[] parts = path.split("/");

        if (parts.length == 0) {
            // resolve to default tenant configuration
            return null;
        }

        if ("tenant-c".equals(parts[1])) {
            // Do 'return requestContext.runBlocking(createTenantConfig());'
            // if a blocking call is required to create a tenant config
            return Uni.createFromItem(createTenantConfig());
        }

        // resolve to default tenant configuration
        return null;
    }

    private Supplier<OidcTenantConfig> createTenantConfig() {
        final OidcTenantConfig config = new OidcTenantConfig();

        config.setTenantId("tenant-c");
        config.setAuthServerUrl("http://localhost:8180/realms/tenant-c");
        config.setClientId("multi-tenant-client");
        OidcTenantConfig.Credentials credentials = new OidcTenantConfig.Credentials();

        credentials.setSecret("my-secret");

        config.setCredentials(credentials);

        // any other setting support by the quarkus-oidc extension

        return () -> config;
    }
}

The OidcTenantConfig returned from this method is the same used to parse the oidc namespace configuration from the application.properties. You can populate it using any of the settings supported by the quarkus-oidc extension.

Tenant Resolution for OIDC 'web-app' applications

Several options are available for selecting the tenant configuration which should be used to secure the current HTTP request for both service and web-app OIDC applications, such as:

  • Check URL paths, for example, a tenant-service configuration has to be used for the "/service" paths, while a tenant-manage configuration - for the "/management" paths

  • Check HTTP headers, for example, with a URL path always being '/service', a header such as "Realm: service" or "Realm: management" can help selecting between the tenant-service and tenant-manage configurations

  • Check URL query parameters - it can work similarly to the way the headers are used to select the tenant configuration

All these options can be easily implemented with the custom TenantResolver and TenantConfigResolver implementations for the OIDC service applications.

However, due to an HTTP redirect required to complete the code authentication flow for the OIDC web-app applications, a custom HTTP cookie may be needed to select the same tenant configuration before and after this redirect request because:

  • URL path may not be the same after the redirect request if a single redirect URL has been registered in the OIDC Provider - the original request path can be restored but after the the tenant configuration is resolved

  • HTTP headers used during the original request are not available after the redirect

  • Custom URL query parameters are restored after the redirect but after the tenant configuration is resolved

One option to ensure the information for resolving the tenant configurations for web-app applications is available before and after the redirect is to use a cookie, for example:

package org.acme.quickstart.oidc;

import java.util.List;

import javax.enterprise.context.ApplicationScoped;

import io.quarkus.oidc.TenantResolver;
import io.vertx.core.http.Cookie;
import io.vertx.ext.web.RoutingContext;

@ApplicationScoped
public class CustomTenantResolver implements TenantResolver {

    @Override
    public String resolve(RoutingContext context) {
        List<String> tenantIdQuery = context.queryParam("tenantId");
        if (!tenantIdQuery.isEmpty()) {
            String tenantId = tenantIdQuery.get(0);
            context.addCookie(Cookie.cookie("tenant", tenantId));
            return tenantId;
        } else if (context.cookieMap().containsKey("tenant")) {
            return context.getCookie("tenant").getValue();
        }

        return null;
    }
}

Disabling Tenant Configurations

Custom TenantResolver and TenantConfigResolver implementations may return null if no tenant can be inferred from the current request and a fallback to the default tenant configuration is required.

If it is expected that the custom resolvers will always infer a tenant then the default tenant configuration is not needed. One can disable it with the quarkus.oidc.tenant-enabled=false setting.

Note that tenant specific configurations can also be disabled, for example: quarkus.oidc.tenant-a.tenant-enabled=false.

Configuration Reference

References