Thanks much, that worked great for me. I created quarkus.properties in Keycloak's conf/ directory, defined a new datasource inside (quarkus.datasource.myname.db-kind=postgresql, etc.), rebuilt Keycloak, and then from within my user storage provider, could retrieve a datasource like this:

import io.agroal.api.AgroalDataSource;
import io.quarkus.agroal.DataSource;
import io.quarkus.agroal.runtime.DataSources;
import io.quarkus.arc.Arc;
import java.lang.annotation.Annotation;
...
    public void someFunction(...) {
            agroalDataSource = Arc.container().instance(DataSources.class, new Annotation[0]).get().getDataSource("myname");                                                                                      
            conn = agroalDataSource.getConnection();
            // Now use the connection like any other JDBC connection
            // ...
    }

The hardest part was probably finding just the right stuff to import :)

Hey eggyknap

Great, that its working for you as well.
I found this solution rather ugly, so after checking the quarkus sourcecode it can be simplified by

import io.quarkus.agroal.runtime.DataSources;
....

    private javax.sql.DataSource createDatasource() {
        try {
            return (javax.sql.DataSource) DataSources.fromName("myName");
        }

That's considerably cleaner, thanks. Now, however, I get errors that my connection is already closed, when the enclosing JTA system tries to commit the transaction. This is the latest in a long series of disappointments trying to get this to work; for now I'm going back to our old, soon-to-be-unsupported authentication method, because I've got to get other stuff done than just messing around with Keycloak

Hi, as a reference: I did not use Agoral data source, as that requires an extra import.

I used this one

session.getProvider(JpaConnectionProvider.class).getEntityManager();

where the session is the KeycloakSession object in the constructor.

and then

this.query = this.em.createNamedQuery("QUERY_NAME", YOUREntity.class);

Hope this helps

Hey Joshua
even if its sounds a little weird, but I was facing the same issue with transactions.
My first try was defining the keycloak datasource (postgres) via environment variables as in every tutorial and add my second datasource (oracle db) with the conf file. This works ok but I ran in the transaction issues even though I don't use transaction. The oracle db is a second userfederation so read-only access.
Once I defined the postgres DB AND the oracle DB in the conf directory AND define the same postgres connection in the environment variables everythings worked as a charm.
Its a little weird, keycloak uses the connection to postgres (in my case) out of the conf directory during runtime and the connection in the environment variables to create the keycloak DB on the first startup.

Maybe something to try once you have a little time to experiment.

The solution from Simin works as well but only to get a connection the same database that keycloak is using for its own data imho