By tripeling the number of JMS producers/consumers I could again max out the CPU, however the messages rate is now still 40% lower than without having failover enabled and fewer JMS producers/consumers.
The only reason I want to enabled the failover option is to have the qpid jms client transparently reconnect to the broker in case of connection loss. I am not really interested in failing over to another broker. Is there another option that achieves transparent reconnect (but not necessarily failover) without the measured performance penality? July 5, 2018 6:13 PM, jo.stenb...@offilive.com wrote: > I am doing performance tests with Artemis 2.6.2 and QPID-JMS 0.34 client. I > can max-out the CPU of > test machines (client and broker) with my qpid-jms based performance client > that uses multiple > producers/consumers if I am configuring the qpid client without failover. If > I however change the > connect string from "amqps://server:5671?..." to > "failover:(amqps://server:5671)?failover.maxReconnectAttempts=-1&..." the > message rate of the qpid > client drops by 75% and the CPU usage on the sending machine stays at around > 5%. Is that an > expected behaviour? Can anyone else reproduce that? > > Thanks, > Jo --------------------------------------------------------------------- To unsubscribe, e-mail: users-unsubscr...@qpid.apache.org For additional commands, e-mail: users-h...@qpid.apache.org
