[ https://issues.jboss.org/browse/WFCORE-316?page=com.atlassian.jira.plugin... ] Brian Stansberry commented on WFCORE-316: ----------------------------------------- [~luck3y] I poked at this a bit and https://github.com/wildfly/wildfly-core/compare/master...bstansberry:WFCO... looks like the basis of a proper semantic. A slave always tries to contact the DC and falls back to the cached domain config only if it fails. In that case, post-boot it installs a service that begins periodically trying to discover the DC again. It kinda works, but 1) Once it joins domain management doesn't work right. All ops against the DC that would get propagated to the slave fail with an ArrayIndexOutOfBoundsException. 2) The persistence of domain.xml on the slave is broken; i.e. it won't persist changes back to domain.cached-remote.xml. It needs to because if the DC connection goes away and then the slave is reloaded it needs to reload to the last good state. 3) Interaction with --backup needs tweaking in general. IIRC we fail if both --backup and --cached-dc are set, and that no longer makes sense with this approach. 3) There are some test failures in AdminOnlyPolicyTestCase 4) Other stuff I'm sure I haven't thought about or noticed. ;) Still, unless 1) turns out to be some kind of really hard problem, this seems like something we can get done, so I'm going to schedule it. -- This message was sent by Atlassian JIRA (v6.4.11#64026)