Thank you for your reply. The library in the application has been replaced to the latest 3.9.4 Release. The behaviour is different but it didn't resolve the problem.
I used the following connection URL: failover:(tcp://amq01:61616?soKeepAlive=true,tcp://amq02:61616?soKeepAlive=true)?randomize=true&backup=true There are 160-160 application processes (all are the same) on two different hosts (host01/02). I executed the following tests. Test 1 The expected number of application connections should be 160-160 from each host to amq01-amq02. This number was 158/160 to amq01/02 on host01 and 144/160 to amq01/02 on host02. The application instances that were connected to amq02 didn't have backup connections to amq01. Test 2 I shut down amq01. Result was 160 connections from both host to amq02 as expected. Test 3 I restarted amq01 and (after amq01 fully started) I shut down amq02. Result was 158/144 connections from host01/02 to amq01. The application instances that didn't initially have backup connections to amq01 didn't fail over to amq01. They didn't have connections to any ActiveMQ instance. I suppose they were attempting to reconnect to amq02 (I didn't trace it). Test 4 I restarted amq02. After a couple of seconds, number of connections were 158/2 from host01 and 144/16 from host02 to amq01/02. The application instances that didn't initially have backup connections to amq01 (only to amq02) reconnected to amq02. The original number of connections were 160-160 to amq02 which was no more the case. I repeated the test and I restarted everything not only once but the behaviour didn't change. I would appreciate your feedback. -- View this message in context: http://activemq.2283324.n4.nabble.com/ActiveMQ-CPP-backup-true-does-not-work-with-failover-transport-tp4728967p4729057.html Sent from the ActiveMQ - User mailing list archive at Nabble.com.