Sounds interesting. Will definitely have a look later. Don't forget to create the applicable jiras.
Curious on some of the background on what lead you down this path. Issues faced with multipoint and so on. Would be good for docs (recommending which to use when) and good for discussion. -David On Tuesday, June 12, 2012, AndyG wrote: > Added new 'multipulse' discovery, and a few fixes. > > Fixes: > ServiceDaemon.java - Must close the server socket even if it does not get a > lock. > ServiceManager.java - Revert 'conf.d' in path. If this is a 'tomee' > requirement then it needs to be checked? > EjbDaemon.java - Ensure streams are closed on failure. Log client induced > errors as warning (Should probably add error filtering). > SocketConnectionFactory.java - Use configured timeout variable. Also added > a > configurable default socket connection timeout of 150ms (Which is long > enough to connect to just about any server on the planet over TCP/IP these > days). > > MultiPulse: > For use where client and server can 'multicast' and a constant heartbeat is > not required or wanted. > The server listens on a defined multicast channel for a client request (A > single pulse). > A client sends a pulse (packet) on the defined channel. > The server responds with a crafted packet which 'can' contain all known > hosts (for multihomed systems use '0.0.0.0' or '[::]' for the ejbd > 'discovery' property), or just the defined value. > The client attempts connections to the server hosts until a valid > connection > is made, discarding any bad URIs. > > > > > -- > View this message in context: > http://openejb.979440.n4.nabble.com/Added-new-multipulse-discovery-and-a-few-fixes-tp4655565.html > Sent from the OpenEJB Dev mailing list archive at Nabble.com. >
