Up till now I was allowing the connection pool to do the caching of valid
connections once a server was found. Using this in house has been working
very well.

You can tune the URI 'timeout' option to be as low as possible for your
network - This is the amount of time the client will sift the multicast
traffic for valid servers.

You can also tune the socket timeout using the
'openejb.client.connection.socket.timeout' system property which has a
default of 500ms - If your server is returning an unreachable address (like
my IPv6 address) then this is how long it take to timeout.

However, I think you have a point with caching the URI so I have added this
feature in 'trunk' to the MulticastPulseClient.

Another 'todo' was to add an 'ignore' filter to the server
MulticastPulseAgent configuration to ignore specific hosts. At the moment
(and by design) the server sends ALL of it's public determined hosts which,
as you can see in the IPv6 address above, is not always desired.



--
View this message in context: 
http://openejb.979440.n4.nabble.com/slow-multicast-discovery-can-be-optimized-by-caching-server-tp4660726p4660784.html
Sent from the OpenEJB Dev mailing list archive at Nabble.com.

Reply via email to