failover is just used by the brokerURL property of the connection factory. Btw session and producers are cheap objects to keep around they are just a bit slow to create on demand. For sending the PooledConnectionFactory should help
On 12/11/2007, TOPPER_HARLEY <[EMAIL PROTECTED]> wrote: > > >"Its normally a glitch in the network - or that the broker died." : > I'm pretty sure the broker was OK since a new client could get a connection > & session and create a temp queue. I previously registered another query ( > http://www.nabble.com/Temp-queue-deleted-when-thread-interrupted-tf4393533s2354.html#a12527277 > here ) regarding using VM style and temp queues getting deleted so this is > why we began using tcp style on server side..... > > >"This is an inefficient way of using JMS BTW...you should try pool > producers if you can.": > > Yes I was reading that but was wondering what is a ballpark figure to start > with for pooling e.g. 10 sessions & 10 producers (I realise its not possible > to accurately w/o application knowledge but a best guess would be > appreciated since they are heavy objects to create...?) > > >"Failover causes the client to reconnect to the broker if the socket > dies.": > Regarding where the failover syntax is needed, is this required in the > transportConnector part of the activeMQ.xml or server side, or just the > client i.e. as part of the > org.apache.activemq.ActiveMQConnectionFactory.brokerURL for the > SimpleMessageListenerContainer? > > /Tom > > > > James.Strachan wrote: > > > > On 12/11/2007, TOPPER_HARLEY <[EMAIL PROTECTED]> wrote: > >> >"It could just be the socket has been dropped.": > >> Is socket dropping a common occurence or simply related to the quality of > >> the network one is running on (my TCP level knowledge isn't great). > > > > Its normally a glitch in the network - or that the broker died. > > > >> Would > >> vm: style for the brokerURL prevent this since it would be insice JVM? > > > > Yes > > > > > >> This is our setup: > >> We have a number of predefined topics which we use when broadcasting > >> updates > >> and number temp queues for direct request/responses communication (one > >> for > >> each of our clients). We maintain one connection on the server side and > >> as > >> we send data from different threads, we cache a threadlocal session > >> object > >> (since ActiveMQSession is for single thread use). For each individual > >> message send, we create and destroy a MessageProducer based on the > >> destination as follows: > >> > >> try { > >> // Thread local retrieval > >> final Session session = this.getSession(); > >> producer = session.createProducer(destination); > >> producer.send(message); > >> } finally { > >> if (producer != null) { > >> try { > >> producer.close(); > >> } catch (final JMSException e) { > >> e.printStackTrace(); > >> } > >> } > >> } > > > > This is an inefficient way of using JMS BTW... > > > > http://activemq.apache.org/how-do-i-use-jms-efficiently.html > > > > you should try pool producers if you can. > > > > > >> >"You night wanna enable failover...": > >> We have only been assigned one port for the broker (specified obviously > >> as > >> part of the brokerURL property) so I dont know if failover can assist us: > > > > Failover causes the client to reconnect to the broker if the socket dies. > > > > > >> my > >> understanding of failover is that you need more than one URI and AMQ > >> switches to another one if one broker goes down: > > > > Not true - it works fine with a single broker URI > > > > -- > > James > > ------- > > http://macstrac.blogspot.com/ > > > > Open Source SOA > > http://open.iona.com > > > > > > -- > View this message in context: > http://www.nabble.com/Unable-to-create-MessageProducer-tf4782623s2354.html#a13709793 > Sent from the ActiveMQ - User mailing list archive at Nabble.com. > > -- James ------- http://macstrac.blogspot.com/ Open Source SOA http://open.iona.com
