On 10/31/2014 5:33 PM, Marek Posolda wrote:
Hi Stan,

did you already tried to test with more auth-servers per Wildfly instance? I am seeing some potential issue in the fact that some stuff is static (at least KeycloakApplication.loadConfig() is static method and configProvider field in org.keycloak.Config is static too). It looks that this might need some refactoring (unless we want to share the same config among all auth-servers, which seems as bad solution to me)

I did not try myself yet to test with more auth-servers, just raising potential issue ;-)
Yes, I tested it.  Each server is its own deployment and thus runs under its own classloader.

Marek

On 29.10.2014 14:28, Stan Silvert wrote:
I've sent a PR for this:
https://github.com/keycloak/keycloak/pull/811

It's a pretty big change in the way the Auth Server is started when the KeyCloak subsystem is used.  The WAR is no longer dropped into the standalone/deployments directory.   This is especially helpful for domain deployments, but it makes standalone cleaner as well.  It will also be important for Feature Pack installation.

The main difference you will see right away with this PR is that the appliance dist now uses the subsystem to launch the Auth Server.

Here are some notes about how everything turned out.  Next, I'll update the documentation if there is no major rework that needs to be done after the PR is reviewed.



_______________________________________________
keycloak-dev mailing list
keycloak-dev@lists.jboss.org
https://lists.jboss.org/mailman/listinfo/keycloak-dev