[ https://issues.jboss.org/browse/WFCORE-3077?page=com.atlassian.jira.plugi... ] Brian Stansberry commented on WFCORE-3077: ------------------------------------------ This one is odd. I can see a problem here, in that RemotingHttpUpgradeService.stop does this: injectedRegistry.getValue().removeProtocol(JBOSS_REMOTING); That will remove all upgrade handling for JBOSS_REMOTING, not just the handling added in the service instance's start() method. That's a problem if you have >1 http-connector for the same undertow listener, which seems to be what the test in question sets up (standard config has http-connector=http-remoting-connector and test adds http-connector=ejb3-tests, both with the same connector-ref.) So I would expect not doing a reload to break things because removing http-connector=ejb3-tests results in http-connector=http-remoting-connector no longer working until reload happens. But I don't see how it would result in http upgrade continuing to work until the reload happens. I'm not sure how the test setup with two http-connectors on the same undertow listener functions properly in the test anyway, since my reading of ChannelUpgradeHandler tells me it will always pick the one registered by the first RemotingHttpUpgradeService to add the protocol, since all such registrations will use an equivalent SimpleHttpUpgradeHandshake. -- This message was sent by Atlassian JIRA (v7.2.3#72005)