I have an application which consists of a frontend and several backend services. The authentication is done via Keycloak. The workflow looks like this: The user logs into the frontend and gets a token from Keycloak. This token is sent to the backend with every request.
The following image explains the current architecture:
In Keycloak I have the following clients:
1. Frontend
2. Core Service
3. User Service
How can I validate calls between services now?
I would imagine something like a service account and these have the possibility to call each other independently from the bearer-token from the frontend. The problem is that both services can be called from the frontend as well as between each other.
Edit:
My API is written with NestJS.
And this is how I call the user-service in my core-service:
and this is my keycloak configuration for the the user-service:
At the moment I don't add anything to the request and I don't have any extra configuration on the interface. So I added the @Resource('user-service')
-Annotation to the Controller and the @Scope()
-Annotation to the Endpoint.
After that I don't get an error immediately and the endpoint is called.I can log that the logic is executed. But as response I still get a 401 Unauthorized Error.
Do I need to specify a scope or what do I need to add in the @Resource
-Annotation?
Edit 2:
I'll try to show you my current situation with many screenshots.
Here is your drawing again. For me, points 1-5 work and point 8 works even if I do not call another service.
That this works, I have the following configuration:
For the core service (gutachten-backend), I do not need to make any further configurations for this. I also have 2 different roles and I can specify them within the API.
Using Postman I send a request to the API and get the token from http://KEYCLOAK-SERVER_URL/auth/realms/REALM_NAME/protocol/openid-connect/token.
These are my 2 testing methods. I call the first one and it works. The following is logged. Means the token is validated received and I get Access:
Now I call the second method. This method calls the user-service.
This is my request in the core-service:
I do not add anything else to my request. Like a bearer token in the header.
The endpoint in the user service is just a test method which logs a message.
This is my configuration for the user service:
I have now tried something with resources, policies and permissions.
And analogously the client permission
@Resource('<name>')
and endpoints with @Scopes([<list>]))
Additionally, through a tutorial on setting up keyacloak in NestJS, I turned on the following config:This adds a global level resource guard, which is permissive. Only controllers annotated with @Resource and methods with @Scopes are handled by this guard.
For people who have the same problem in the future. The answer from @BenchVue helped a lot to understand the concept in general.
What was missing is that a token must also be added for each request between services. Namely the token of the client.
So before the request is sent, the following query takes place. This is the method to get the token for a client:
getAccessToken(): Observable<string> {
const header = {
headers: {
'Content-Type': 'application/x-www-form-urlencoded',
},
};
return this.httpService.post(
'{{keycloakurl}}/auth/realms/{{realm}}/protocol/openid-connect/token',
`grant_type=client_credentials&client_id={{clientId}}&client_secret={{clientSecret}}`,
header).pipe(
map((response) => {
return response.data.access_token as string;
}
));
}