[ https://jira.jboss.org/jira/browse/JBREM-1189?page=com.atlassian.jira.plu... ] David Lloyd commented on JBREM-1189: ------------------------------------ A Connection represents multiple clients; the connection must remain open to drain all existing requests. If the connection is dropped, all outstanding requests have an indeterminate result. Just unbinding the server port isn't enough because that only affects the creation of new Connections. To drain new Clients, we must keep the connection up but refuse new client requests coming across. -- This message is automatically generated by JIRA. - If you think it was sent incorrectly contact one of the administrators: https://jira.jboss.org/jira/secure/Administrators.jspa - For more information on JIRA, see: http://www.atlassian.com/software/jira

[ https://jira.jboss.org/jira/browse/JBREM-1189?page=com.atlassian.jira.plu... ] David Lloyd commented on JBREM-1189: ------------------------------------ In order to ensure smooth operation of the test suite, I've added an "awaitClosed()/awaitClosedUninterruptibly()" method pair to coordinate clean shutdown. Maybe all that's needed is one more method - requestClose(), something like that - on Connection which does the graceful shutdown? Or maybe close() should be graceful by default with an additional closeImmediate() method. I think the first might be better though. Then the pattern would look something like this: connection = do connect; try { ... do work ... connection.requestClose(); connection.awaitClosed(); } finally { // make sure it is shut down either way... safeClose(connection); // this might be needed as well: connection.awaitClosedUninterruptibly(); } -- This message is automatically generated by JIRA. - If you think it was sent incorrectly contact one of the administrators: https://jira.jboss.org/jira/secure/Administrators.jspa - For more information on JIRA, see: http://www.atlassian.com/software/jira

[ https://jira.jboss.org/jira/browse/JBREM-1189?page=com.atlassian.jira.plu... ] David Lloyd commented on JBREM-1189: ------------------------------------ Immediate shutdown would be for doing a hard shutdown of the container or system for example, or when the wait time for close to finish is more expensive than the loss of connection, or when you've initiated a graceful close and some time has elapsed and you just want the connection to go away. IoFuture is a possibility. Cancel wouldn't work (obviously). It's not clear though how results would be handled if two threads called closeAsync(). With close() only one thread would get a result; but if the expected semantics of closeAsync() were to be to await the close finish, that rule would be broken, and close exceptions might be delivered to multiple threads. I guess overall I don't like it - it would be the only case anywhere that more than one IoFuture can exist for the same event that I can think of. The correct semantics would be very hard to figure out. awaitClose() has no side effects other than simply waiting for the resource to be closed. -- This message is automatically generated by JIRA. - If you think it was sent incorrectly contact one of the administrators: https://jira.jboss.org/jira/secure/Administrators.jspa - For more information on JIRA, see: http://www.atlassian.com/software/jira